id stringlengths 7 26 | text_type stringclasses 2
values | original_text stringlengths 25 7.75k | cot_response stringclasses 1
value | domain stringclasses 31
values | grpo_v5_3 stringlengths 310 815 | grpo_v7_1 stringlengths 371 841 | grpo_v7_2 stringlengths 389 844 | grpo_v7_3 stringlengths 342 872 | prompt listlengths 2 2 | ground_truth stringclasses 2
values | prompt_len int64 157 1.89k |
|---|---|---|---|---|---|---|---|---|---|---|---|
deepseek-v3.2_1955 | ai | Public administration represents a multifaceted discipline dedicated to enhancing governmental operations through improved management and policy. It has been defined in several ways, including: the oversight of public initiatives; the practical implementation of political decisions that shape daily civic life; and the ... | null | GovernmentPublic | {
"rationale": "First, the passage uses polished, generic encyclopedic language and balanced parallel lists that are common in AI-generated text. Second, it lacks concrete examples or a distinct voice and includes mildly tautological phrasings (e.g., 'the academic examination... and the analysis of existing policies'... | {
"rationale": "First, the passage uses highly polished, generic encyclopedic phrasing and balanced parallelism (e.g., 'including...') which is common in AI-generated text. Second, it lacks concrete examples or personal nuance and relies on broad definitions and synonyms, suggesting algorithmic summarization rather t... | {
"rationale": "First, the passage uses very polished, generic encyclopedic language with balanced parallel lists and no personal voice, which is common in AI-generated text. Second, it lacks concrete examples or idiosyncratic phrasing and repeats transitional phrases like 'including' and 'in addition,' suggesting al... | {
"rationale": "First, the passage uses very polished, generic encyclopedic language with balanced parallelism and lacks personal voice or concrete examples, which is common in AI-generated text. Second, it relies on broad, stock definitions (e.g., 'multifaceted discipline' and list-like definitions) rather than orig... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 252 |
deepseek-v3.2_909 | ai | Gendered behavior refers to the fundamental expressions of gender socialization and identity manifested throughout an individual's developmental stages. It encompasses the ways individuals comport themselves in accordance with socially constructed male or female gender norms (Au & Rudolfsdottir, 2021). A prominent gend... | null | EducationMaterial | {
"rationale": "First, the prose is polished and generic with smooth transitions and balanced paragraphs but lacks concrete sensory detail, which is common in AI-generated text. Second, there are subtle factual oddities (e.g., 'conception' used figuratively) and parenthetical citations that appear plausibly fabricate... | {
"rationale": "First, the passage uses polished, generic academic phrasing and smooth transitions while avoiding concrete sensory detail, which is common in AI-generated text. Second, it relies on broad generalizations and repeated transitional markers ('Consequently', 'Therefore') suggesting algorithmic coherence o... | {
"rationale": "First, the passage uses polished, generic academic phrasing and smooth transitions while avoiding concrete sensory detail, which is common in AI-generated text. Second, it relies on broad generalizations and repeated transitional markers ('Consequently', 'Therefore') suggesting algorithmic smoothing r... | {
"rationale": "First, the passage shows highly polished, generic academic phrasing with smooth transitions and few concrete sensory details, which is common in AI-generated text. Second, there are mechanical repetitions (e.g., 'Consequently' used multiple times) and balanced, neutral generalizations suggesting algor... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 511 |
deepseek-v3.2_3900 | ai | As part of the Nainital-Cape Survey, eight variable stars of the $\delta\,$Scuti classification have been identified, exhibiting pulsation periods spanning from several minutes to a few hours. To interpret these detected pulsational characteristics, we have conducted detailed non-adiabatic linear stability analyses on ... | null | TechnicalWriting | {
"rationale": "First, the paragraph uses polished, generic academic phrasing and smooth transitions while making broad claims without providing raw data or methodological detail, a common trait of AI-generated scientific text. Second, there are subtle oddities—e.g., the unusual 'delta\\,Scuti' spacing and the use of... | {
"rationale": "First, the paragraph uses polished, generic academic phrasing and smooth transitions while making broad claims without providing equations or detailed methodological specifics, a common trait of AI-generated text. Second, there are subtle oddities—redundant hedging ('strong correspondence') and unifor... | {
"rationale": "First, the paragraph uses polished, generic academic phrasing and smooth transitions while making broad claims without providing equations or detailed evidence, a common trait of AI-generated text. Second, there are subtle factual oddities—listing 'HD\\$\\mathrm{118660}$' et al. as variables and asser... | {
"rationale": "First, the paragraph uses highly polished, generic scientific language with smooth transitions and lacks specific methodological details or data, a common trait of AI-generated text. Second, there are stylistic oddities (e.g., 'To interpret...' followed by empty clauses) and unusually balanced, comple... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 351 |
deepseek-v3.2_924 | ai | While the specific term "mentoring" may be relatively recent in ecclesiastical vocabulary, the practice itself has a long and established history within Christian communities. For years, churches have informally facilitated spiritual growth through various relationships—such as spiritual coaching, direction, disciplesh... | null | EducationMaterial | {
"rationale": "First, the prose is highly polished and generic, using balanced transitions and broad generalizations rather than providing concrete, idiosyncratic details typical of human scholarly writing. Second, the text repeatedly paraphrases and restates arguments (e.g., multiple mentions of 'Christian mentorin... | {
"rationale": "First, the passage uses highly polished, generic academic phrasing and smooth transitions while making broad claims without concrete data, a pattern common in AI-generated text. Second, it repeatedly paraphrases a single source (Mallison 1998) and supplies plausible but shallow citations and balanced ... | {
"rationale": "First, the passage uses highly polished, generic academic phrasing and smooth transitions while making broad, unsubstantiated claims, a pattern common in AI-generated text. Second, it repeatedly paraphrases a single citation (Mallison 1998) in exhaustive, balanced terms without close reading or specif... | {
"rationale": "First, the passage uses highly polished, generic academic phrasing and balanced transitions with few concrete details, suggesting algorithmic generation rather than a human writer's distinct voice. Second, it repeatedly paraphrases a single concept (Christian mentoring/Christian coaching) and includes... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 679 |
deepseek-v3.2_2210 | ai | A bureaucratic error made me Satan. I didn't ask for this. Nobody would.
I just want my old, boring life back.
**RRRRRNT RRRRRRNT RRRRRRNT**
I slam the alarm. Two weeks of waking up under a blood-red stone ceiling. Two weeks of a bed made of human bones. Two weeks of brushing my teeth with blood. My ex said I wasn't... | null | LiteratureCreativeWriting | {
"rationale": "First, the piece uses generic, trope-driven plot points and flat sensory detail while relying on abrupt transitions, which is common in AI-generated text. Second, there are mechanical signs like inconsistent capitalization (RRRRRNT), excessive use of short, declarative sentences, and oddly formal head... | {
"rationale": "First, the passage uses generic, shock-horror tropes and flat sentence-level pacing rather than deep sensory detail, which is common in AI outputs. Second, there are mechanical patterns (repeated 'Two weeks...' openings, abrupt transitions) and oddly formalized emotional beats like 'Nobody would' that... | {
"rationale": "First, the passage uses generic, trope-driven vignettes (bureaucratic error, sudden promotion) and lacks deep sensory detail, which is common in AI outputs. Second, there are mechanical patterns like repeated sentence openings ('First, ...'), excessive use of bolding for emphasis, and abrupt transitio... | {
"rationale": "First, the passage uses generic, shock-horror tropes and flat sentence-level pacing rather than showing nuanced emotion, which is common in AI outputs. Second, there are mechanical signs like excessive parallelism (\"Dull. Simple. Easy.\") and abrupt transitions between exposition and shock that sugge... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 360 |
deepseek-v3.2_725 | ai | Circularly shifts an integer's digits right by a given amount, returning a string.
**Key Logic:**
* If the shift exceeds the digit count, the digits are simply reversed.
* Otherwise, digits are moved from the end to the front.
**Examples:**
* `circular_shift(12, 1)` → `"21"`
* `circular_shift(12, 2)` → `"12"` | null | Code | {
"rationale": "First, the text is very polished and generic, using clear sectional headings and short explanatory sentences common in AI-generated outputs. Second, it makes a slightly imprecise claim ('digits are simply reversed' if shifted beyond the count) which matches AI tendency to state plausible but not fully... | {
"rationale": "First, the text is very generic and formulaic, using clear headings and bulleted examples without personal nuance. Second, it shows uniform sentence length and neutral phrasing common in AI-generated outputs rather than idiosyncratic human wording. Moreover, it lacks contextual details or stylistic ma... | {
"rationale": "First, the text is very polished and generic, using clear sectional headings (Key Logic, Examples) and short, uniform sentences common in AI outputs. Second, it provides a mechanical, exhaustive explanation ('If the shift exceeds... the digits are simply reversed') and gives minimal illustrative examp... | {
"rationale": "First, the text is very polished and generic, offering high-level logic and examples without personalization, which is common in AI-generated outputs. Second, there is mechanical repetition (two bullet points identical) and lack of contextual nuance or error handling that a human might include when di... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 242 |
deepseek-v3.2_577 | ai | Studies indicate that the alcohol industry uses corporate social marketing—a type of CSR focused on public welfare—to deflect attention from its harmful activities. These campaigns, which appear to advocate for responsible consumption, also work to normalize drinking as a common social behavior. | null | Business | {
"rationale": "First, the passage is very polished and generic, using high-level claims ('Studies indicate') without citing specifics, a common trait of AI-generated text. Second, it relies on neutral, balanced phrasing and smooth transitions ('also work to normalize drinking'), suggesting algorithmic sentence const... | {
"rationale": "First, the passage is very polished and uses generic, high-level claims (e.g., 'studies indicate') without providing specifics or citations, a common trait of AI-generated text. Second, its neutral, balanced tone and lack of personal voice or anecdotal detail suggest algorithmic summarization rather t... | {
"rationale": "First, the passage is very polished and uses generic, high-level claims ('Studies indicate...') without citing specifics, a common trait of AI-generated text. Second, it repeats common buzzwords (CSR, social marketing, responsible consumption) in quick succession and maintains a neutral, even tone typ... | {
"rationale": "First, the text is very polished and generic, using balanced, hedging language ('studies indicate', 'appear to') which is common in AI-generated outputs. Second, it lacks concrete details, examples, or attribution and stays at a high level with neutral tone, suggesting algorithmic summarization rather... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 198 |
deepseek-v3.2_2272 | ai | [ CW ] Compose a poetic work wherein a persona harbors a clandestine truth, poised for disclosure. The fog derides our hesitation with the taunts of a simpleton. The odor of remorse carries a vileness surpassing even that of the grave, and in its wake, a being stirs within my depths, gradually ascending toward the ligh... | null | LiteratureCreativeWriting | {
"rationale": "First, the prose is highly polished and abstract with balanced clauses and frequent transitional phrases, a pattern common in AI-generated creative text. Second, it relies on generic dramatic devices (fog derides, monstrosity rising) and lacks concrete sensory grounding or idiosyncratic detail typical... | {
"rationale": "First, the prose is highly polished and uses many formal transitions while avoiding concrete sensory detail, which is common in AI-generated text. Second, there are subtle oddities—repetitive abstract phrasings ('secret', 'monstrosity') and unusual lexical choices like 'inter you'—suggesting algorithm... | {
"rationale": "First, the prose is highly polished and abstract with balanced, adjective-heavy sentences and few concrete sensory specifics, which is common in AI-generated text. Second, there are mechanical repetitions (e.g., 'I shall...') and unusual lexical choices like 'inter you' that suggest algorithmic assemb... | {
"rationale": "First, the prose is unusually polished and abstract with excessive balancing and synonymic padding (e.g., 'buried... alongside'), a common trait of AI-generated text. Second, there are mechanical repetitions ('this... this') and lack of concrete sensory grounding despite vivid descriptions, suggesting... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 524 |
deepseek-v3.2_918 | ai | Email has totally changed how we talk to each other. It lets us connect with anyone—coworkers, family, friends, or clients—instantly, without the hassle of meeting up in person.
But it’s interesting to see how email itself has reshaped our communication. For one, it’s massively expanded our reach. Conversations don’t ... | null | EducationMaterial | {
"rationale": "First, the prose is generic and evenly balanced with smooth transitions and few personal details, a pattern common in AI-generated text. Second, there are telltale markers like overgeneralized claims, lack of concrete examples beyond a single year, and a parenthetical citation 'CNBC, 2018' used casual... | {
"rationale": "First, the prose is generic and evenly balanced with smooth transitions and lacks personal anecdotes or concrete examples, a pattern common in AI-generated text. Second, there are mechanical signs like repetitive sentence openings ('First... second... moreover') and a parenthetical citation 'CNBC, 201... | {
"rationale": "First, the piece uses generic, evenly balanced exposition and smooth transitions with few concrete personal details, a pattern common in AI-generated text. Second, it includes a suspicious citation ('CNBC, 2018') and makes broad generalizations about texting and email that suggest hallucination rather... | {
"rationale": "First, the piece uses generic, evenly balanced exposition and smooth transitions with few concrete personal details, a pattern common in AI-generated text. Second, it includes a suspicious citation ('CNBC, 2018') and makes broad, polished generalizations rather than specific lived experience, suggesti... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 451 |
deepseek-v3.2_3879 | ai | Imagine a computer humming a secret song. Not through wires or Wi-Fi, but through silent tremors in the very desk it sits upon. This is the realm of air-gap covert channels—where isolated computers whisper data through light, heat, or sound. Now, we introduce a new verse: the seismic whisper.
Every computer has a hidd... | null | TechnicalWriting | {
"rationale": "First, the passage uses polished, generic academic phrasing and smooth transitions while avoiding concrete experimental details (no datasets, metrics, or citations), which is common in AI-generated summaries. Second, it repeats high-level motifs ('vibrational covert channel', 'score of countermeasures... | {
"rationale": "First, the passage uses polished, high-level exposition with smooth transitions and few concrete technical details (e.g., no mention of sampling rates, noise budgets, or empirical evidence), which is common in AI-generated summaries. Second, it relies heavily on rhetorical devices ('Imagine...', 'Now,... | {
"rationale": "First, the piece uses polished, promotional academic phrasing ('In this paper, we unveil...') and lacks concrete experimental details, which is common in AI-generated summaries. Second, there are conceptual slippages (saying 'seismic whisper' for a vibration) and smooth transitions indicative of model... | {
"rationale": "First, the passage uses polished, generic academic phrasing and smooth transitions (e.g., 'In this paper, we unveil...') which is common in AI-generated outputs. Second, it lacks concrete experimental details and includes vague claims ('results ... will be presented'), suggesting template-driven gener... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 393 |
deepseek-v3.2_3952 | ai | Geographic routing uses node positions to guide routing and is widely studied in sensor networks. However, its practical applicability remains a major challenge. Prior work often relies on broad, unverifiable assumptions about idealized networks or employs costly graph planarization methods. This raises key questions: ... | null | TechnicalWriting | {
"rationale": "First, the passage has a polished, generic academic tone with balanced transitions and few concrete technical specifics, which is common in AI-generated abstracts. Second, it makes high-level claims (e.g., 'establish the four foundational principles') without presenting evidence or novel details, sugg... | {
"rationale": "First, the passage uses polished, generic academic phrasing and makes high-level claims (e.g., 'establish...four foundational principles') without presenting concrete results or evidence, which is common in AI-generated abstracts. Second, it repeats stock transitions ('This raises key questions', 'In ... | {
"rationale": "First, the passage uses polished, generic academic phrasing and makes high-level claims without concrete experimental details, which is common in AI-generated abstracts. Second, it repeats transitional phrases ('However', 'In this paper') and balances clauses in a mechanically even way typical of AI o... | {
"rationale": "First, the passage uses polished, generic academic phrasing and lacks concrete experimental details, which is common in AI-generated abstracts. Second, it makes sweeping claims (e.g., 'four foundational principles') without elaboration and repeats stock transitions suggesting template-driven generatio... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 303 |
deepseek-v3.2_1954 | ai | A splash of soap and water can be a shield, holding back a tide of invisible threats. Sometimes, the cure itself becomes the barrier—stopping a single spark from becoming a wildfire, whether it leaps from person to person or seeps into our food and water. | null | GovernmentPublic | {
"rationale": "First, the passage uses polished, generalized metaphors and balanced clauses ('splash... tide... contagion') which is common in AI-generated text. Second, it lacks concrete details or personal grounding and relies on stock persuasive phrases ('holding back...', 'cure itself becomes the barrier'), sugg... | {
"rationale": "First, the passage uses polished, generalized metaphors (soap as shield, cure itself becomes barrier) and lacks concrete sensory detail or personal grounding, which is common in AI-generated text. Second, its balanced, evenly paced sentences and neutral tone suggest algorithmic smoothing rather than a... | {
"rationale": "First, the passage uses polished, generic metaphors (soap as shield, cure itself becomes barrier) and lacks concrete sensory detail, which is common in AI-generated text. Second, its balanced, evenly paced sentences and neutral tone suggest algorithmic smoothing rather than a distinct human voice. Mor... | {
"rationale": "First, the passage uses polished, generalized metaphors (soap as shield, cure as barrier) and lacks concrete sensory detail or personal grounding, which often indicates AI generation. Second, its balanced, neutral tone and smooth transitions suggest algorithmic smoothing rather than spontaneous human ... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 204 |
deepseek-v3.2_2702 | ai | Governor Scott Walker has named August 13th "Purple Hearts for Healing" Day for a young Wisconsin girl. Her family thanked him for honoring "our little hero." The girl was attacked by two of her classmates, who told police they stabbed her 19 times as a "dedication" to Slenderman, a creepy fictional character from onli... | null | NewsArticle | {
"rationale": "First, the passage mixes sensational facts with smooth, neutral exposition in a way common to AI summarization rather than a distinct human voice. Second, it shows mild factual inconsistencies (e.g., Governor Scott Walker naming a specific day for a real event) and uses generic transitional phrases li... | {
"rationale": "First, the piece shows abrupt topic shifts (a 'purple hearts' proclamation followed by a detailed news-style report) and uses generic transitional phrasing common in AI outputs. Second, it mixes plausible facts with clear factual inaccuracies and conflations (e.g., mischaracterizing the victim as 'inc... | {
"rationale": "First, the piece shows mechanical, evenly balanced news-like paragraphs with generic transitions and minimal sensory detail, which is common in AI-generated text. Second, it contains clear factual inaccuracies and conflations (e.g., misattributing names and events) suggesting hallucination rather than... | {
"rationale": "First, the piece shows mechanical, evenly balanced transitions and generic phrasing (e.g., 'raising money...local businesses are pitching in') suggesting algorithmic summarization rather than a human narrator's distinctive voice. Second, it mixes sensational facts with smooth, neutral exposition in a ... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 476 |
deepseek-v3.2_549 | ai | A 2000 study reviewed existing econometric research on the link between social and financial performance. It concluded that the conflicting results of earlier studies—which had reported positive, negative, and neutral financial impacts—stemmed from flawed empirical methods. The authors argued that when properly specifi... | null | Business | {
"rationale": "First, the passage contains domain-specific detail (a 2000 study reviewing econometric literature) and cites specific theoretical perspectives (critical theorists' 'lofty' or 'unrealistic expectations') in a way typical of human academic writing. Second, its prose shows varied sentence lengths, nuance... | {
"rationale": "First, the passage uses precise, domain-specific phrasing (e.g., 'A 2000 study' and 'window-dressing') and cites specific critiques, which is typical of human academic writing. Second, its structure is purposefully nuanced with balanced transitions ('The authors... argued... Critics...'), showing edit... | {
"rationale": "First, the passage uses very polished, generic academic phrasing and smooth transitions without concrete specifics (no year cited for the 2000 study), which is common in AI-generated text. Second, it presents balanced viewpoints (authors argue-neutral; critics-window-dressing) in a templated way rathe... | {
"rationale": "First, the passage uses very polished, generic academic phrasing (e.g., 'reviewed existing econometric research... stemmed from flawed empirical methods') which is characteristic of AI-generated text. Second, it lacks concrete details, numbers, or citations and relies on balanced, evenly structured pa... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 276 |
deepseek-v3.2_1624 | ai | It's chocolate—what more can I say? This assortment has the perfect variety of everything our family loves: Kit Kats, Reese's, Take 5, and more. With a family of six, it goes fast in our house! | null | FoodCusine | {
"rationale": "First, the copy is very polished and uses generic, widely-applicable phrases ('It's chocolate... what more can I say', 'perfect variety') common in AI-generated marketing text. Second, it lacks concrete personal details about taste or context and relies on simple, balanced sentence structures typical ... | {
"rationale": "First, the text uses generic, polished phrasing ('It's chocolate... what more can I say') and lacks concrete sensory detail, which is common in AI outputs. Second, there is a mild inconsistency—'With a family of six' yet 'in our house!'—suggesting formulaic stitching rather than natural conversational... | {
"rationale": "First, the text uses generic, polished promotional language ('It's chocolate... what more can I say') common in AI outputs. Second, it lacks personal specifics and contains repetitive positive cues ('perfect variety', 'goes fast in our house'), suggesting algorithmic generation. Moreover, there is a m... | {
"rationale": "First, the text uses generic, polished promotional language ('with a family of six, it goes fast in our house!') suggesting template-driven generation. Second, it lacks personal specifics and contains repetitive positive adjectives ('perfect variety', 'what more can I say'), common in AI outputs. More... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 199 |
deepseek-v3.2_263 | ai | So, what was Romanticism? It was this huge wave of creativity—think art, books, music, and big ideas—that started in Europe around the late 1700s. It really hit its stride from about 1800 to 1850. | null | ArtCulture | {
"rationale": "First, the passage uses generic, evenly balanced phrasing ('So, what was...') and high-level summarization common in AI outputs. Second, it lacks concrete details, examples, or a distinct personal voice and instead lists categories (art, books, music) in a mechanical way. Moreover, the tone is neutral... | {
"rationale": "First, the reply uses generic, evenly balanced phrasing ('So, what was...') and broad, well-known facts (Romanticism, dates) which is common in AI outputs. Second, it lacks personal detail, nuance, or anecdote and relies on safe, generalizing vocabulary ('think art, books, music, and big ideas'), sugg... | {
"rationale": "First, the reply is very brief, uses generic high-level phrases ('This huge wave...') and lacks concrete examples or nuance, which is common in AI-generated text. Second, it repeats common patterns (a hook plus a tidy two-sentence summary) suggesting algorithmic generation rather than a personalized h... | {
"rationale": "First, the text is very brief, generic, and uses neutral explanatory phrasing common in AI outputs. Second, it lacks personal voice, specific examples, or nuance and relies on safe generalities ('think art...'), suggesting algorithmic generation. Moreover, the polished, evenly balanced structure and a... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 208 |
deepseek-v3.2_2577 | ai | The flag has been raised above the City Chambers in George Square as a gesture of solidarity with the people of France. This follows the killing of twelve individuals at the offices of the satirical magazine *Charlie Hebdo* by suspected Islamist extremists on Wednesday. Glasgow undertook a similar action last month by ... | null | NewsArticle | {
"rationale": "First, the prose is polished and generic with smooth transitions and few sensory details, a common trait of AI-generated copy. Second, there are factual incongruities (e.g., mentioning 'the flag of Pakistan' for a French flag-raising) and stock transitional phrases ('This follows...') suggesting algor... | {
"rationale": "First, the prose is generic and evenly balanced with smooth transitions and lacks sensory detail or a distinct human voice, which is common in AI-generated copy. Second, there are factual inconsistencies (mentioning 'the flag of Pakistan' for the Charlie Hebdo flag) and some odd phrasings ('thoughts a... | {
"rationale": "First, the prose is generic and evenly balanced with smooth transitions and stock phrases (e.g., 'thoughts and prayers') rather than distinctive human detail. Second, there are factual inconsistencies—mentioning 'the killing of twelve individuals' then linking to last month's Pakistan action—which sug... | {
"rationale": "First, the prose is very polished and generic with balanced transitions and few sensory details, suggesting algorithmic generation rather than a human reporter's immediate reporting. Second, there is repetitive, formulaic phrasing (multiple 'He stated...' clauses) and lack of specific timestamps or un... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 390 |
deepseek-v3.2_2364 | ai | You work the line at a cybernetic implant plant, where the favorite pastime among coworkers is planting defective parts on each other. "Goddamn it, Ryan," Jerry grumbled, fishing his finger from his tea along with a soggy bourbon. The digit had felt strange all day—though, to be fair, it hadn't been his original for ye... | null | LiteratureCreativeWriting | {
"rationale": "First, the prose is polished and even with generic worldbuilding and tidy transitions, which is common in AI-generated text. Second, there are mechanical repetitions (e.g., 'Just him... Just him') and abrupt plot conveniences (the lift/zip mechanism) suggesting algorithmic smoothing rather than lived ... | {
"rationale": "First, the prose is polished and evenly paced with generic worldbuilding and few sensory specifics, which often indicates algorithmic generation. Second, there are mechanical signs (repetitive transitional phrases, balanced sentence lengths, lack of distinctive stylistic fingerprints) common in AI out... | {
"rationale": "First, the prose is polished and evenly paced with generic worldbuilding and few sensory specifics, which is common in AI-generated text. Second, there are mechanical repetitions (e.g., 'Just him... Just him') and abrupt transitions suggesting algorithmic smoothing rather than human stylistic variatio... | {
"rationale": "First, the prose is polished and evenly paced with generic worldbuilding and few sensory specifics, which often indicates algorithmic generation. Second, there are mechanical repetitions (e.g., 'Just him, a magazine') and abrupt transitions suggesting template-driven output rather than a human's nuanc... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 637 |
deepseek-v3.2_2521 | ai | In this research, both free testosterone and androstenedione showed a statistically significant relationship with sexual desire in the overall group of women. The measure ADT-G was not more strongly linked to sexual desire than blood androgen levels, indicating it is not a better assessment tool than measuring circulat... | null | MedicalText | {
"rationale": "First, the passage uses specific domain terminology (free testosterone, androstenedione, ADT-G) and makes a nuanced empirical claim about measurement tools, which matches human academic writing. Second, it shows small stylistic idiosyncrasies ('in this research' without 'the') and varied sentence leng... | {
"rationale": "First, the passage uses specific domain terminology (free testosterone, androstenedione, ADT-G) and makes a nuanced empirical claim about measurement tools, which matches human academic writing. Second, its brevity, direct comparison to blood androgen levels, and lack of excessive hedging or generic f... | {
"rationale": "First, the passage uses specific domain terminology (free testosterone, androstenedione, ADT-G) and makes a nuanced empirical claim about comparative link strength, which suggests subject-matter knowledge typical of human authors. Second, its concise, paper-like tone and lack of excessive hedging or g... | {
"rationale": "First, the passage uses concise, domain-specific language (free testosterone, androstenedione, ADT-G) suggesting familiarity with human scientific writing rather than generic AI filler. Second, it presents a specific empirical contrast (not more strongly linked... indicating...) implying comparative d... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 216 |
deepseek-v3.2_3814 | ai | The Lambek calculus serves as a prominent logical framework for analyzing natural language syntax. While its initial formulation captured many complex syntactic phenomena, these were confined to context-free boundaries. To handle finer linguistic details, various enhancements to the Lambek calculus have been proposed. ... | null | TechnicalWriting | {
"rationale": "First, the passage uses polished, generic academic phrasing and makes high-level claims without providing concrete technical detail, a common trait of AI-generated text. Second, it includes a plausible but potentially fabricated citation (Morrill and Valentin 2015) and presents abstract, balanced stat... | {
"rationale": "First, the passage uses polished, generic academic phrasing and makes high-level claims without providing concrete technical detail, a common trait of AI-generated text. Second, there are subtle oddities—redundant transitions ('To handle finer linguistic details...') and unusual spacing ('non‑standard... | {
"rationale": "First, the paragraph uses polished, generic academic phrasing and makes high-level claims without providing concrete evidence, a common trait of AI-generated text. Second, there are subtle factual oddities—e.g., attributing a 2015 contribution to Morrill and Valentin and asserting undecidability along... | {
"rationale": "First, the paragraph uses polished, generic academic phrasing and makes high-level claims without providing concrete evidence or examples, a common trait of AI-generated text. Second, there are subtle oddities—redundant transitions ('To handle finer linguistic details...') and unusual spacing ('non‑st... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 302 |
deepseek-v3.2_903 | ai | Let's take a look at this Bosch dishwasher manual. Right away, you notice it's packed with helpful illustrations. It shows you exactly how to add rinse aid, load your dishes, and take out the baskets—super handy for anyone who isn't a dishwasher expert.
The manual is clearly written with the user in mind. Most of the ... | null | EducationMaterial | {
"rationale": "First, the passage uses generic, evenly balanced praise and lacks concrete specifics (no model numbers, pages, or personal anecdotes), which is common in AI-generated text. Second, its polished, neutral tone, repetitive sentence patterns, and smooth transitions suggest algorithmic smoothing rather tha... | {
"rationale": "First, the passage uses very generic, evenly balanced descriptions and lacks concrete specifics (no model numbers, pages, or personal experience), which is common in AI-generated text. Second, its polished, neutral tone, repetitive sentence patterns, and smooth transitions suggest algorithmic smoothin... | {
"rationale": "First, the tone is uniformly promotional and generic with smooth transitions and few concrete specifics, which is common in AI-generated text. Second, it uses repetitive sentence patterns ('You'll...','It...') and balanced paragraph structure suggesting algorithmic smoothing rather than a distinct hum... | {
"rationale": "First, the text uses very generic, evenly balanced praise and lacks concrete specifics (no page numbers, examples, or personal experience), which is common in AI-generated copy. Second, its polished, neutral tone, repetitive positive adjectives, and systematic structure suggest algorithmic generation ... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 387 |
deepseek-v3.2_3713 | ai | In the first round, both the women's and men's tournaments have six groups of four teams each. The top two teams from every group move on automatically, and a few more can earn a spot through a lucky loser playoff. | null | Sports | {
"rationale": "First, the passage is brief, conversational, and uses specific, plausible tournament details (women's and men's tournaments, six groups of four teams) in a way typical of human explanatory writing. Second, it employs casual phrasings like 'a few more can earn a spot' and uses a concrete example ('top ... | {
"rationale": "First, the text is brief, conversational, and uses casual phrasing like 'In the first round' which is common in human forum answers. Second, it shows small stylistic quirks (e.g., 'a few more can earn...') suggesting spontaneous human wording rather than overly polished AI output. Moreover, it lacks r... | {
"rationale": "First, the text is brief, conversational, and uses casual phrasing like 'In the first round' which is common in human forum answers. Second, it shows small stylistic variation ('the women's and men's tournaments') and lacks repetitive AI patterns or excessive hedging. Moreover, it includes a specific ... | {
"rationale": "First, the text is very brief, generic, and lacks personalization or contextual detail, traits common in AI-generated explanatory replies. Second, its clear, evenly balanced structure and neutral tone—introducing size, grouping, then progression—match patterns often seen in AI outputs. Moreover, there... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 196 |
deepseek-v3.2_2370 | ai | The clock is ticking from the first incision. For patients over seventy, an emergency laparotomy is a perilous passage where survival cannot be taken for granted. To change the tide, we must reimagine the journey. It begins with swift recognition and a surge of skilled hands at the ready, guided by crystal-clear protoc... | null | MedicalText | {
"rationale": "First, the passage uses polished, generic transitional phrases and abstract, generalized statements (e.g., 'imagine then', 'this is how we turn...') which is common in AI-generated text. Second, it lacks concrete clinical detail and instead relies on broad, promotional metaphors ('harbor'), suggesting... | {
"rationale": "First, the passage uses polished, generic transitional phrases (e.g., 'To change the tide', 'Imagine then') and lacks concrete clinical detail, which is common in AI-generated text. Second, it relies on broad abstractions ('centralized centers', 'casting a safety net') rather than specific mechanisms ... | {
"rationale": "First, the passage uses polished, generic transitional phrases (e.g., 'To change the tide', 'Imagine then') and lacks concrete clinical detail, which is common in AI-generated text. Second, it relies on broad abstractions ('centralized centers', 'casting a safety net') rather than specific mechanisms ... | {
"rationale": "First, the passage uses polished, generic transitional phrases (imagine then, this is how we turn...) and lacks concrete clinical detail, suggesting algorithmic generalization. Second, it relies on balanced, parallel constructions and abundant positive adjectives ('crystal-clear protocols', 'harbor') ... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 270 |
deepseek-v3.2_2480 | ai | Sleep issues are widespread among adults in Japan. While methods like anchoring vignettes can adjust for differences in how people report symptoms, they are not essential for effectively addressing sleep and energy concerns. Instead, policies and clinical care should prioritize factors such as older age, gender-based d... | null | MedicalText | {
"rationale": "First, the passage uses polished, generic academic phrasing and makes broad policy recommendations without supporting evidence, a common trait of AI-generated text. Second, it includes a possibly invented term ('anchoring vignettes') and offers balanced, evenly weighted points (age, gender, comorbidit... | {
"rationale": "First, the passage uses polished, generic academic phrasing and makes broad policy recommendations without supporting data, a common trait of AI-generated text. Second, it includes a possibly invented term ('anchoring vignettes') and presents balanced, evenly weighted points (differences in reporting,... | {
"rationale": "First, the passage uses polished, generic academic phrasing and a checklist-like list of recommendations (age, gender, comorbidity, exercise) which is common in AI-generated summaries. Second, it lacks concrete data or nuance and makes broad claims ('are widespread...') without supporting detail, sugg... | {
"rationale": "First, the passage uses polished, generic academic phrasing ('policies and clinical care') and lacks personal voice or concrete data, which is common in AI-generated text. Second, it makes broad, balanced claims (e.g., 'not essential...adjust for differences') without supporting evidence and repeats k... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 229 |
deepseek-v3.2_3055 | ai | A gasp, a hand to the heart.
Behold the vista the window frames—a far cry from the promised picture.
A shared, sinking realization: the view is a profound and utter disappointment. | null | PersonalCommunication | {
"rationale": "First, the passage uses polished, abstract phrasings ('a shared, sinking realization'') and balanced sentence structures common in AI-generated text. Second, it lacks concrete sensory detail or personal idiosyncrasy and relies on generic dramatic cues, suggesting algorithmic generation. Moreover, the ... | {
"rationale": "First, the passage uses polished, abstract phrasing ('a shared, sinking realization') and lacks concrete sensory detail, which is common in AI-generated text. Second, there is mechanical parallelism (a line followed by 'Behold...') and some redundancy ('A gasp, a hand to the heart'), suggesting algori... | {
"rationale": "First, the passage uses polished, abstract phrasing ('a shared, sinking realization') and lacks concrete sensory detail, which is common in AI-generated text. Second, there is mechanical parallelism (a gasp... Behold... A shared...) and balanced sentence lengths suggesting algorithmic patterning rathe... | {
"rationale": "First, the passage uses very polished, abstract phrasing ('a shared, sinking realization') and lacks concrete sensory detail, which often indicates AI generation. Second, there is mechanical parallelism (multiple lines beginning with 'A...') and slightly stilted transitions suggesting algorithmic patt... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 190 |
deepseek-v3.2_3105 | ai | Did you attend the performance of "Ally Mabel" last evening?
Yes, I did. Did you?
Indeed, it was quite humorous.
Yes, I thoroughly enjoyed it. The scene with the judge was particularly excellent, was it not? | null | PersonalCommunication | {
"rationale": "First, the dialogue is unusually polished and evenly balanced with generic, polite responses lacking distinct personal voices, a pattern common in AI-generated text. Second, there are small stylistic oddities (for example, 'particularly excellent' meaning 'especially') and a lack of sensory detail or ... | {
"rationale": "First, the dialogue is unusually polished and evenly balanced with generic positive lines (e.g., 'thoroughly enjoyed' and 'particularly excellent') rather than showing distinct personal voices. Second, there is mechanical back-and-forth repetition ('Did you... Yes, I did') and a slightly stilted turn-... | {
"rationale": "First, the dialogue is unusually polished and evenly balanced with generic positive lines ('thoroughly enjoyed', 'particularly excellent') suggesting algorithmic generation. Second, there is mechanical repetition of phrases (multiple 'Yes, I...' responses) and a lack of distinct personal voice or sens... | {
"rationale": "First, the exchange is unusually polished and evenly balanced with generic, polite responses lacking personal details, which is common in AI-generated text. Second, there is mechanical repetition (multiple 'Yes, I...' lines) and a lack of disfluent or idiosyncratic phrasing suggesting human spontaneit... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 200 |
deepseek-v3.2_709 | ai | def find_rightmost_vowel_between_consonants(word):
"""Returns the rightmost vowel in the word that is surrounded by consonants.
Only vowels not at the start or end of the word are considered.
If no such vowel exists, returns an empty string. The search is case-sensitive.
"""
if len(word) < 3:
... | null | Code | {
"rationale": "First, the code shows a polished, idiomatic Python style with a doctest-like docstring and use of set('aeiouAEIOU') which is common in AI-generated examples. Second, there is excessive clarity ('case-sensitive', 'only vowels not at the start or end') and a mechanical, exhaustive approach (checking i-1... | {
"rationale": "First, the code is very polished and uses common coding-pattern patterns (set('aeiouAEIOU'), range(len(word)-2, 0,-1)) and a neat docstring style often seen in AI-generated examples. Second, there is excessive symmetry and completeness—returning an empty string, listing required cases, and insisting o... | {
"rationale": "First, the code is very polished and uses common coding-pattern patterns (set for vowels, slice -2..0 reverse loop) suggesting template-like generation. Second, there is excessive clarity and formalism (docstring, case-sensitivity, explicit bounds check) typical of AI outputs rather than casual human ... | {
"rationale": "First, the code is very polished and generic, using common patterns (set('aeiouAEIOU'), range(len(word)-2, 0, -1)) and a neat docstring — styles often seen in AI-generated answers. Second, there is excessive symmetry and completeness (returning '' when no match, case-sensitive flag, bounded index rang... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 291 |
deepseek-v3.2_2206 | ai | Each day, the world collectively selects one person to die. The individual receiving the highest number of daily votes is terminated. This was humanity’s final creation. Societal disintegration commenced within a week. Every aspiration and ideal that had surrounded its invention disintegrated once it was released onlin... | null | LiteratureCreativeWriting | {
"rationale": "First, the prose is polished, evenly balanced, and uses generic transitional phrases (e.g., 'Thus', 'Eventually') and tidy paragraphing common in AI-generated text. Second, it lacks idiosyncratic sensory detail or deep emotional nuance and instead stacks named events and transitions in a mechanical wa... | {
"rationale": "First, the prose shows nuanced stylistic choices—specific details (Eastern Time at noon), precise casualty counts, and a satirical social commentary—that read like intentional human creative control. Second, there are subtle rhetorical moves (a sardonic observer joke, use of 'so' instead of 'but') and... | {
"rationale": "First, the prose is polished and uses balanced, generalized statements alongside occasional awkward phrasings (e.g., 'The individual receiving the highest number of daily votes is terminated'), which is common in AI outputs. Second, the piece relies heavily on explicit exposition ('the truth was desol... | {
"rationale": "First, the prose is polished, evenly balanced, and uses generic transitional phrases (e.g., 'Thus,' 'Eventually') in a way common to AI-generated text. Second, it lacks idiosyncratic sensory detail or personal voice and instead stacks well-known tropes (Kardashians, victims falling by vote) suggesting... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 602 |
deepseek-v3.2_937 | ai | We don't see the factory floor in the gleam of a new phone. Marx warned of this societal sleight of hand: commodities become enchanted objects, their value spun from the invisible hours of labor woven into them. This fetishism—devoid of any sensual meaning—eclipses human need, making products the masters and exchanges ... | null | EducationMaterial | {
"rationale": "First, the prose is highly polished and uses generic transitional phrases (e.g., 'This fetishism...') and balanced, generalized statements common in AI-generated text. Second, it lacks concrete sensory detail or idiosyncratic personal markers and instead relies on abstract, evenly weighted arguments s... | {
"rationale": "First, the passage uses polished, generalized philosophical phrasing and smooth transitions without concrete details or specific evidence, which is common in AI-generated text. Second, it relies on abstract abstractions (e.g., 'potent cipher', 'haunted marketplace') and balanced, neutral tone rather t... | {
"rationale": "First, the passage uses polished, generalized philosophical phrasing and smooth transitions without concrete sensory detail, which is common in AI-generated text. Second, it relies on abstract abstractions (e.g., 'potent cipher', 'haunted marketplace') and balanced, neutral paragraphs suggesting algor... | {
"rationale": "First, the passage uses polished, generalized philosophical language and smooth transitions without concrete sensory detail, which is common in AI-generated text. Second, it relies on abstract metaphors (commodified objects as 'ghostly dance') and balanced, neutral framing rather than idiosyncratic pe... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 392 |
deepseek-v3.2_2856 | ai | It’s strange, but all of this reminds me: he never really had the permission to do that. He didn’t earn it. He’s had plenty of time, but no—other things always mattered more. Honestly, it would take him twice as long as he’s even known me to deserve that right.
I know this probably doesn’t make sense to anyone else. ... | null | OnlineContent | {
"rationale": "First, the passage uses polished, generic emotional statements and balanced sentence lengths without strong sensory detail, a common trait of AI-generated text. Second, it relies on abstract generalizations ('It’s strange...') and hedging ('Honestly... Maybe that’s for the best') rather than concrete ... | {
"rationale": "First, the passage uses generic, evenly balanced phrasing and lacks concrete sensory detail or a distinct personal voice, which often indicates AI generation. Second, there are mechanical patterns (repeated 'he' openings, hedging like 'Maybe that's for the best') and polished neutrality common in AI o... | {
"rationale": "First, the passage uses generic, evenly balanced phrasing and lacks concrete sensory detail or a distinct personal voice, which often indicates AI generation. Second, there are mechanical patterns (repetition of 'even known me' and hedging like 'Maybe that's for the best') common in AI outputs. Moreov... | {
"rationale": "First, the passage uses generic, evenly balanced phrasing and lacks concrete sensory detail or a distinct personal voice, which often indicates algorithmic generation. Second, there are mechanical repetitions (e.g., 'Maybe that's for the best') and hedging ('Honestly...') common in AI outputs aiming f... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 233 |
deepseek-v3.2_2221 | ai | So you wake up and the weirdest thing happens: you're suddenly inside your favorite video game. The catch? You still need a job, something that fits this new world. But you've got zero new skills—just whatever you were back home. So what's an average day like? Well, I was a reporter, so that's all I am here. Don't even... | null | LiteratureCreativeWriting | {
"rationale": "First, the piece uses generic, evenly balanced prose and stock transitional phrases ('that first day', 'lately') with few sensory specifics, which is common in AI-generated text. Second, it relies on trope-driven plot progression and safe, neutral voice rather than distinctive personal details or idio... | {
"rationale": "First, the piece uses very even pacing, generic transitional phrases ('That first day...'), and lacks sensory specificity, which are common in AI-generated text. Second, it relies on trope-driven plot progression and repeats simple sentence patterns ('Every. Game.'), suggesting algorithmic smoothing r... | {
"rationale": "First, the piece uses very even, generic pacing and stock transitional phrases ('that first day', 'that first night') suggesting algorithmic smoothing rather than a distinct human voice. Second, it relies on broad exposition and lacks sensory specificity or deep emotional nuance, instead offering tidy... | {
"rationale": "First, the piece uses very even, generic pacing and lacks sensory detail or deep characterization, which is common in AI-generated text. Second, it relies heavily on transitional phrases ('that first day', 'lately') and repeats structures ('I...'), suggesting algorithmic patterning rather than human i... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 485 |
deepseek-v3.2_1907 | ai | Cities in Europe and Asia often have really solid public transit systems that serve their older, dense urban areas. But over in the Americas and other newer cities, things are more spread out, and public transport isn't always as widespread or connected. If you're curious about this stuff, the International Association... | null | GovernmentPublic | {
"rationale": "First, the reply uses generic, balanced phrasing and smooth transitions without personal anecdotes or concrete specifics, a common trait of AI-generated text. Second, it includes a plausible but unsourced factual reference to the UITP and a tidy sign-off ('go-to global group') suggesting template-like... | {
"rationale": "First, the reply uses generic, evenly balanced phrasing and lacks concrete specifics or personal detail, a common trait of AI-generated text. Second, it includes a plausible but unsourced factual reference (the UITP) and offers broad, neutral guidance—patterns often seen in AI responses. Moreover, the... | {
"rationale": "First, the reply uses generic, evenly balanced phrasing and smooth transitions without personal detail, a common trait of AI-generated text. Second, it includes a plausible but unsourced factual reference (the UITP) and offers broad, neutral guidance, suggesting template-like generation. Moreover, the... | {
"rationale": "First, the reply uses generic, balanced phrasing and lacks personal detail or concrete examples beyond a single organization, which is common in AI-generated text. Second, it contains smooth transitions and neutral, polished tone without errors or idiosyncrasies, suggesting algorithmic consistency rat... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 235 |
deepseek-v3.2_4151 | ai | Bound states in the continuum (BICs) have been a prominent research focus in optics and photonics for the past decade. Studying quasi-BICs in simple structures, where their effects are most pronounced, is of particular interest. Dielectric cylinders serve as a prime example, and quasi-BICs have been extensively investi... | null | TravelTourism | {
"rationale": "First, the passage uses polished, generic academic phrasing and makes high-level claims without providing experimental detail or citations, a common trait of AI-generated text. Second, it repeats key phrases ('quasi-BIC', 'strong-coupling/weak-coupling') in a mechanical pattern and presents balanced, ... | {
"rationale": "First, the passage uses polished, generic academic phrasing and makes broad claims without providing numerical data or experimental detail, a common trait of AI-generated text. Second, there are subtle conceptual imprecisions (e.g., invoking 'three waves' in the far field and treating avoided crossing... | {
"rationale": "First, the passage uses polished, generic academic phrasing and lacks experimental details or citations, suggesting algorithmic summarization. Second, it repeats key phrases ('quasi-BIC', 'strong-coupling/weak-coupling') in a mechanical way and presents balanced, evenly structured paragraphs common in... | {
"rationale": "First, the passage uses very polished, generic academic phrasing and lacks experimental details or numerical data, suggesting algorithmic summarization. Second, it repeats key phrases ('quasi-BIC', 'strong-coupling/weak-coupling') in a mechanical way and presents balanced, high-level statements typica... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 386 |
deepseek-v3.2_3711 | ai | The earliest recorded competition at the ancient Olympic Games was the stadium race, a sprint of roughly 192 meters. This athletic event dates back to at least 776 BC, when its first victors were documented. | null | Sports | {
"rationale": "First, the passage is concise, factual, and uses specific numeric detail ('roughly 192 meters') and a dated origin (776 BC), matching typical human-authored encyclopedia text. Second, its simple, natural sentence constructions and lack of excessive hedging or repetitive AI patterns suggest human autho... | {
"rationale": "First, the passage is very brief and generic, offering high-level facts ('dates back to at least 776 BC') without nuance or sourcing, which is common in AI-generated text. Second, there is mild redundancy (saying 'This athletic event' after introducing 'The earliest recorded competition...') and a mec... | {
"rationale": "First, the passage is very brief, neutral, and uses generic transitional phrases ('This athletic event...'), a pattern common in AI-generated text. Second, it makes a precise but unsupported specific claim (stadium race ≈192 meters) without sourcing, suggesting model overgeneralization rather than inf... | {
"rationale": "First, the passage is very brief, neutral, and generic, lacking personal voice or contextual detail, which often indicates AI generation. Second, it uses smooth, evenly balanced sentence structures and polished transitions common in machine outputs rather than exhibiting human idiosyncratic phrasing. ... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 196 |
deepseek-v3.2_31 | ai | We characterize the singular values of the linear transformation in a standard 2D multi-channel convolutional layer, enabling their efficient computation. This characterization further yields an algorithm for projecting a convolutional layer onto an operator-norm ball. As an effective regularizer, this projection impro... | null | AcademicResearch | {
"rationale": "First, the paragraph is unusually polished and uses generic academic phrasings ('characterize... yields an algorithm') while making a specific technical claim about projecting convolutions, which is common in AI-generated abstracts. Second, it lacks citations and repeats key phrases ('This...'), sugge... | {
"rationale": "First, the paragraph is unusually polished and uses smooth transitions while making broad claims (e.g., 'improves performance' and project improves performance) which is common in AI-generated summaries. Second, it blends correct technical phrases ('singular values', 'operator-norm ball') with a speci... | {
"rationale": "First, the paragraph blends technical jargon (e.g., 'singular values', 'operator-norm ball') with a sudden, impressive empirical claim about CIFAR-10 in a way typical of AI-generated summaries that stitch together sources. Second, there is a lack of detailed attribution or nuanced explanation—results ... | {
"rationale": "First, the text makes unusually strong and rapid claims (a 6.2% to 5.3% reduction in test error from projecting singular values) which is atypical for scientific writing and suggests overgeneralization common in AI-generated outputs. Second, the style is polished, generic, and lacks supporting detail ... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 240 |
deepseek-v3.2_1309 | ai | The Company currently lacks a formal procedure to evaluate climate-related risks and their potential financial impact on its operations. However, as noted previously, the Company's disaster recovery plan does include measures to address climate-related disruptions and ensure business continuity. Should a significant cl... | null | Environmental | {
"rationale": "First, the passage uses polished, generic corporate language and high-level assurances without concrete details, which is common in AI-generated text. Second, it shows repetitive transitional phrasing ('as noted previously', 'Should a significant climate event occur') suggesting template-like generati... | {
"rationale": "First, the passage uses highly polished, generic corporate phrasing (e.g., 'as noted previously', 'maintains sufficient capital and surplus') which is common in AI-generated text. Second, it is balanced and evenly structured with no specific details, examples, or idiosyncratic language, suggesting alg... | {
"rationale": "First, the passage uses highly polished, generic corporate language and smooth transitions without concrete specifics, which is common in AI-generated text. Second, it repeats phrases like 'the Company' and relies on hedging ('as noted previously', 'maintains sufficient capital and surplus') suggestin... | {
"rationale": "First, the passage uses highly polished, generic corporate language and smooth transitions without concrete specifics, which is common in AI-generated text. Second, it repeats phrases like 'the Company' and relies on hedging ('as noted previously', 'maintains sufficient capital and surplus') suggestin... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 217 |
deepseek-v3.2_2897 | ai | Diane: Orange
Derek: Red
Kelly: Red
Matthew: Orange
Kristen: Orange
Joanne: Yellow
Mary Clare: Sky blue
Brett: Blue
Emily: Yellow
Dave: Sepia
Bryan: Bright yellow
Chris: Orange
Brandon: Green
Me: Brown
(List created to remember everyone's chosen color.) | null | OnlineContent | {
"rationale": "First, the passage lists names and colors in a mechanically even, exhaustive way with little stylistic variation, which is common in AI-generated outputs. Second, there is odd redundancy (multiple 'Orange' entries) and a parenthetical aside ('List created to remember everyone's chosen color.') that re... | {
"rationale": "First, the passage shows mechanical, evenly balanced sentence lengths and a lack of connective transitions, which is common in AI outputs. Second, there is a notable absence of personalization or contextual detail (no dialogue, minimal backstory) suggesting algorithmic generation rather than human con... | {
"rationale": "First, the passage shows mechanical, evenly balanced sentence lengths and a lack of connective transitions, which is common in AI-generated text. Second, there is repetitive naming and minimal contextual linking (e.g., 'List created to remember everyone's chosen color'), suggesting algorithmic list ge... | {
"rationale": "First, the passage presents a mechanically ordered list of names and colors with perfect parallelism and no back-and-forth dialogue, which is common in AI outputs. Second, there is unusually little natural conversational nuance or hesitation among the speakers and a lack of sensory or contextual detai... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 224 |
deepseek-v3.2_889 | ai | For repatriating cremated remains to France, the required point of contact is the Consulate General of France in Los Angeles. An appointment must be scheduled at least two days in advance to arrange the shipment and sealing of the urn.
The family will need to provide a specific set of documents, including:
* A U.S. ... | null | EducationMaterial | {
"rationale": "First, the passage has a very polished, generic instructional tone with exhaustive lists and few personal details, which is common in AI-generated text. Second, it shows repetitive parallel structures (multiple enumerated points starting with 'The...') and lacks anecdotal or contextual nuance that a h... | {
"rationale": "First, the passage has a very polished, generic informational tone with clear enumerations and few personal touches, which is common in AI-generated text. Second, it includes oddly specific but potentially fabricated details (e.g., 'airline... cargo hold' for repatriation and a lone name 'Julie Duhaut... | {
"rationale": "First, the text has a very polished, generic informational tone with well-structured lists and few personal touches, which is common in AI-generated outputs. Second, there are subtle oddities—redundant phrasing ('airlines like Air France-KLM') and a possibly fabricated name ('Julie Duhaut-Bedos')—sugg... | {
"rationale": "First, the text has a highly polished, neutral tone with systematic listing and few personal touches, suggesting algorithmic generation. Second, it includes specific but generic details (Consulate General in Los Angeles, Air France-KLM, named contact) without deeper contextualization, a common AI patt... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 370 |
deepseek-v3.2_852 | ai | Defining what it means to be human is not straightforward. Some equate humanity with personal autonomy and control over one's own life. Others believe it lies in the trappings of civilization—wearing clothes, holding a job, and engaging in activities beyond the animal realm. Determining how people wish to live is equal... | null | EducationMaterial | {
"rationale": "First, the prose is polished, generic, and evenly balanced with smooth transitions and few concrete sensory details, a pattern common in AI-generated text. Second, it uses abstract summarizing phrases (e.g., 'Overall... profoundly inspiring') and lacks distinct personal idiosyncrasies or vivid anecdot... | {
"rationale": "First, the prose is very even, generic, and polished with balanced transitions and few sensory specifics, which is common in AI-generated text. Second, there are subtle factual slippages (a general 'On Being' mention paired with a vague citation of Thich Nhat Hanh) suggesting synthesis rather than fir... | {
"rationale": "First, the prose is very even, generic, and polished with balanced transitions and few sensory specifics, traits common in AI-generated text. Second, there are mild factual slippages (a general description of Thich Nhat Hanh rather than a deep engagement) and repetitive transitional phrasing suggestin... | {
"rationale": "First, the prose is very even, generic, and polished with balanced transitions, lacking sensory detail or a distinct personal voice, which is common in AI-generated text. Second, there are mechanical signposts (explicit linking phrases like 'thus' and 'overall') and safe, neutral phrasing suggesting a... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 443 |
deepseek-v3.2_3433 | ai | When the apostles in Jerusalem heard that Samaria had accepted the word of God, they sent Peter and John to them. Upon their arrival, they prayed for the Samaritans to receive the Holy Spirit, for it had not yet come upon any of them; they had only been baptized in the name of the Lord Jesus. Then Peter and John laid t... | null | Religious | {
"rationale": "First, the passage uses generic, evenly balanced phrasing and a neutral explanatory tone common in AI outputs. Second, it lacks personal voice, nuance, or contextual detail beyond basic facts, suggesting algorithmic summarization rather than human reflection. Moreover, sentence structures are simple a... | {
"rationale": "First, the passage uses generic, evenly balanced phrasing and a neutral explanatory tone common in AI outputs. Second, it lacks personal voice, nuance, or contextual detail (no date or situational grounding) and repeats simple transitions, suggesting algorithmic generation. Moreover, sentence length a... | {
"rationale": "First, the passage is highly generic and uses smooth transitional phrases (When... , Upon their arrival..., Then...) which is common in AI-generated text. Second, it lacks personal voice, concrete details, or stylistic idiosyncrasies and repeats simple sentence patterns suggesting algorithmic generati... | {
"rationale": "First, the passage is highly generic and formulaic, using neutral, evenly balanced sentences common in AI outputs. Second, it lacks personal voice, anecdotal detail, or stylistic idiosyncrasy suggesting a human author, instead showing polished repetition of structures (e.g., 'upon their arrival... the... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 233 |
deepseek-v3.2_1470 | ai | So, about that downside you asked about: it really comes down to two things.
If you take out a loan and close your cards, it can hurt your score. That's because your credit utilization on that loan instantly hits 100%, and you're replacing several older accounts with one brand new one, which shortens your overall cred... | null | Finance | {
"rationale": "First, the reply has a polished, generic explanatory tone with balanced transitions and no personal anecdotes, which is common in AI-generated text. Second, it uses repeated transitional phrases ('if...','first', 'one last tip') and presents information in a checklist-style manner typical of AI output... | {
"rationale": "First, the reply has a very polished, evenly balanced tone with clear signposting ('So, about that...') and generic transitional phrases, which is common in AI-generated text. Second, it uses smooth transitions and lacks personal anecdotes or concrete examples beyond general guidance, suggesting algor... | {
"rationale": "First, the reply has a very polished, evenly balanced tone with clear signposting ('So, about that...') and generic transitional phrases, which is common in AI-generated text. Second, it uses smooth transitions and lacks personal anecdotes or concrete examples beyond broad generalizations, suggesting ... | {
"rationale": "First, the reply has a very polished, evenly balanced tone with clear signposting ('If... If... Either way') and lacks personal anecdotes or concrete numbers, suggesting algorithmic generation. Second, it uses generic transitional phrases and repeats safe, general advice (balance, credit utilization, ... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 373 |
deepseek-v3.2_3922 | ai | While perpendicular MgO-based Magnetic Tunnel Junctions are prime contenders for powering next-generation STT memory, they face a stubborn hurdle: switching current density remains stubbornly above the critical 10⁶ A/cm² mark. A breakthrough emerged in 2012, when Wang et al. demonstrated that an electric field could as... | null | TechnicalWriting | {
"rationale": "First, the passage uses polished, generic academic phrasing ('dive deeper', 'unveiling the intricate micromagnetic landscape') and makes broad claims without providing concrete experimental details, a common trait of AI-generated text. Second, there are stylistic signs of machine generation—balanced s... | {
"rationale": "First, the passage uses polished, high-level academic phrasing and smooth transitions without providing concrete experimental details, which is common in AI-generated text. Second, it makes broad claims (e.g., 'our simulations reveal...') while omitting data or methodological specifics, a typical AI p... | {
"rationale": "First, the passage uses polished, generic academic phrasing ('dive deeper', 'unveiling the intricate micromagnetic landscape') and lacks concrete experimental details, which is common in AI-generated text. Second, there are stylistic signs of machine generation—repetitive word choices ('stubborn'), ba... | {
"rationale": "First, the passage uses polished, generic academic phrasing ('dive deeper', 'unveiling the intricate micromagnetic landscape') and lacks concrete experimental details, which is common in AI-generated text. Second, there are stylistic signs of machine generation—balanced but formulaic transitions, smoo... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 293 |
deepseek-v3.2_580 | ai | Good CSR policies help a company stay on the right side of the law. They make it easier to follow new rules, avoid expensive lawsuits, and get everyone in the company on the same page about doing the right thing. | null | Business | {
"rationale": "First, the passage is brief, conversational, and uses idiomatic phrasing ('on the right side of the law' and 'stay on the same page'), which is common in human-authored advice. Second, it offers practical, experience-based suggestions (follow new rules, avoid expensive lawsuits) without excessive hedg... | {
"rationale": "First, the passage is short, conversational, and uses plain language with a personal tone (e.g., 'stay on the right side of the law'), which is common in human-authored tips. Second, it offers practical, experience-based advice without excessive hedging or repetitive filler patterns often seen in AI o... | {
"rationale": "First, the passage is short, conversational, and uses plain language with a clear purpose-driven structure, which is common in human-authored advice. Second, it gives practical, experience-based examples (stay on the right side of the law, avoid expensive lawsuits) rather than generic, overly polished... | {
"rationale": "First, the text is short, conversational, and uses plain language with a personal tone (e.g., 'stay on the right side of the law'), which often indicates human authorship. Second, it offers practical, experience-based advice without excessive hedging or mechanical repetition common in AI outputs, sugg... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 194 |
deepseek-v3.2_681 | ai | Determines if a list is sorted in ascending order. Returns False if any number appears more than twice. | null | Code | {
"rationale": "First, the text is extremely brief and uses conversational, memory-based wording ('appears more than twice') typical of human note-taking. Second, it lacks common AI patterns like excessive hedging, repetitive transitional phrases, or overly balanced sentence structures and instead offers a compact, p... | {
"rationale": "First, the text is extremely brief and uses generic, evenly balanced clauses common in AI-generated outputs. Second, it lacks contextual detail and personal phrasing, giving a templated feel rather than a human conversational response. Moreover, the clean grammar and neutral tone match frequent AI com... | {
"rationale": "First, the text is extremely brief and uses generic, evenly balanced clauses common in AI-generated outputs. Second, it lacks contextual detail or personal phrasing and shows repetitive sentence patterns suggesting algorithmic generation. Moreover, the neutral, mechanical tone and absence of stylistic... | {
"rationale": "First, the text is extremely brief and lacks contextual detail, common in AI-generated outputs. Second, it uses a generic, evenly balanced clause 'Determines if... Returns False...' suggesting template-like generation. Moreover, there is no personal voice or nuance and the language is mechanically cle... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 170 |
deepseek-v3.2_3902 | ai | The strength and vertical distribution of atmospheric turbulence is a key factor determining the performance of optical and infrared telescopes, yet it remains challenging to measure. We describe a new technique that uses a sequence of short-exposure images of a star field from a small telescope. By analyzing the diffe... | null | TechnicalWriting | {
"rationale": "First, the passage contains domain-specific details (Markov Chain Monte Carlo optimization, AST3 on Dome A) and a coherent experimental/analysis narrative typical of human-authored scientific abstracts. Second, its sentence variety, purposeful use of transitions, and absence of common AI repetition or... | {
"rationale": "First, the passage uses polished, dense technical language and many buzzwords (e.g., 'Markov Chain Monte Carlo optimization', 'structure functions') in a way common to AI-generated abstracts. Second, it makes broad, generic claims ('We present verification results...') without presenting data or speci... | {
"rationale": "First, the paragraph packs many advanced technical terms (structure functions, Markov Chain Monte Carlo, AST3 at Dome A) in a smooth, polished manner without deep empirical support, which is common in AI-generated text. Second, there are subtle factual oddities (a brief mention of 'demonstration... wi... | {
"rationale": "First, the text uses highly polished, abstract academic phrasing ('key factor determining the performance... remains challenging to measure') suggesting algorithmic generation rather than spontaneous human prose. Second, it is unusually balanced and generic, lacking personal anecdotes or concrete expe... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 309 |
deepseek-v3.2_3766 | ai | Our investigation peers into the stellar nurseries of S231-S235, listening to their whispers in the rare 'quasi-thermal' lines of ammonia and cyanoacetylene, and to the more urgent flares of methanol and water masers. These regions are embedded within the vast molecular cloud G174+2.5. From this cosmic reservoir, we fi... | null | TechnicalWriting | {
"rationale": "First, the passage uses polished, generic academic phrasing and smooth transitions while avoiding concrete methodological detail, a common trait of AI-generated text. Second, it contains plausible but unsourced specifics (object names, redshifts, precise temperatures) presented without context or nuan... | {
"rationale": "First, the passage uses polished, high-level academic phrasing with smooth transitions but lacks methodological detail, which is common in AI-generated summaries. Second, it contains plausible-sounding specifics (object names, tentative detections, quoted temperatures) without citations or nuance, sug... | {
"rationale": "First, the passage uses polished, high-level academic phrasing with smooth transitions but lacks methodological detail, which is common in AI-generated summaries. Second, it makes unusually specific yet unverified claims (e.g., detection of NH₃ and HC₃N in named objects and a precise 36.2 GHz line) su... | {
"rationale": "First, the passage uses highly polished, generic academic phrasing ('peeks into the stellar nurseries', 'whispers in the rare...') suggesting algorithmic smoothing rather than idiosyncratic human detail. Second, it packs many specific-sounding discoveries (WB89 673, 36.2 GHz line, temperature ~16–30 K... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 414 |
deepseek-v3.2_2215 | ai | A kid gets snatched. The monsters under the bed and in the closet are furious—they swear they'll find her. I watch. I stay close. But I never touch.
Tonight was different. She was just... gone. No trace. No sound. No little heartbeat.
Her mom is crying. Police lights flash outside. Blue. Red. Blue. Red.
Her bed was ... | null | LiteratureCreativeWriting | {
"rationale": "First, the passage uses tight, trope-driven pacing and generic sensory cues (smells, colors, transitions) instead of deep, idiosyncratic detail, which is common in AI-generated text. Second, it relies on repetitive sentence patterns ('I...') and mechanical listing of events rather than distinct voice ... | {
"rationale": "First, the passage uses tight, trope-driven pacing and frequent, staccato sentence openings ('I') which is common in AI-generated microfiction. Second, it relies on list-like sensory tokens (Blue. Red. ... Cigarettes, booze) and abrupt transitions rather than deep character development, suggesting alg... | {
"rationale": "First, the passage relies on short, declarative sentences, repeated motifs (e.g., 'Blue. Red.') and mechanical pacing that are common in AI-generated text. Second, it lacks deep sensory grounding and shows abrupt transitions while using generic emotional statements ('kid gets snatched', 'desperation')... | {
"rationale": "First, the passage relies on short, declarative sentences, repeated motifs (e.g., 'Blue. Red.') and mechanical pacing that often appear in AI-generated text. Second, it lacks deep sensory grounding and shows abrupt transitions while using generic emotional statements ('kid gets snatched', 'desperation... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 354 |
deepseek-v3.2_1223 | ai | Here at REPUBLIC, we're actively working to manage the risks that climate change could bring. Here’s a look at what we’re doing:
* We use trusted industry tools to assess risks from hurricanes, wind, tornadoes, and earthquakes.
* We keep a close eye on our claims and where our property risks are concentrated.
* ... | null | Environmental | {
"rationale": "First, the passage uses very polished, generic corporate phrasing and a balanced list structure common in AI-generated text. Second, it lacks concrete specifics (no percentages, dates, names of tools or companies) and relies on high-level buzzwords like 'catastrophe models' and 'reinsurance', suggesti... | {
"rationale": "First, the passage uses very polished, generic corporate phrasing and a balanced list structure common in AI-generated text. Second, it lacks concrete specifics (no percentages, dates, names of tools or companies) and relies on high-level buzzwords like 'catastrophe models' and 'reinsurance', suggesti... | {
"rationale": "First, the passage uses very polished, evenly balanced language and a checklist-style structure common in AI-generated copy. Second, it lacks concrete specifics (no numbers, dates, or unique examples) and relies on generic reassuring phrases like 'trusted industry tools' and 'run several catastrophe m... | {
"rationale": "First, the text uses very polished, generic corporate language and a checklist-style structure common in AI-generated copy. Second, it lacks concrete specifics (no numbers, dates, or examples) and repeats safe phrases like 'we use trusted industry tools' suggesting template-driven generation. Moreover... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 311 |
deepseek-v3.2_1160 | ai | I rented this DVD on a whim, knowing nothing about it. I was surprised to see Michael Madsen in a film I’d never heard of, but it was free, so I gave it a shot.
My feeling is that Mr. Madsen would probably prefer to forget he was ever involved in this mess.
It proves that even talented actors can’t salvage a terrible... | null | Entertainment | {
"rationale": "First, the review uses generic, evenly balanced criticisms and lacks concrete sensory details (no specific scenes or dialogue), a common trait of AI-generated text. Second, its polished, neutral tone, predictable transitions, and balanced sentence lengths suggest algorithmic generation rather than a d... | {
"rationale": "First, the review uses generic, evenly balanced criticisms and lacks concrete sensory details (no specific scenes or dialogue), which is common in AI outputs. Second, its polished grammar, neutral tone, and formulaic transitions ('That's the reason...') suggest algorithmic generation rather than a dis... | {
"rationale": "First, the review uses generic, evenly balanced criticisms and lacks concrete scene-specific details, which is common in AI-generated text. Second, there are repetitive transitions ('first', 'second', 'moreover') and polished neutrality suggesting algorithmic smoothing rather than a distinct human voi... | {
"rationale": "First, the review uses very generic, evenly balanced phrasing and lacks concrete sensory details (no scenes, dialogue, or specific examples), which is common in AI-generated text. Second, it repeats stock transitions ('that's the reason...') and maintains a neutral, polished tone without personal idio... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 326 |
deepseek-v3.2_1657 | ai | The package was advertised as containing large salted peanuts, yet the contents consisted of small, unsalted nuts. It is unclear whether this discrepancy was a simple mistake or a deliberate attempt to mislabel the item as "Jumbo." | null | FoodCusine | {
"rationale": "First, the text is generic and uses neutral, balanced phrasing ('It is unclear...') which is common in AI outputs. Second, it lacks specific details or personal perspective (no brand, timing, or cause), suggesting algorithmic summarization rather than lived experience. Moreover, the sentence structure... | {
"rationale": "First, the reply uses neutral, generic phrasing ('It is unclear...') and lacks personal detail or sensory nuance, which is common in AI outputs. Second, its structure is balanced and formal with clear signposting (yet; it is unclear...) suggesting algorithmic sentence planning rather than spontaneous ... | {
"rationale": "First, the reply is generic and uses neutral, balanced phrasing ('It is unclear...'), which is common in AI outputs. Second, it lacks personal detail or sensory context and offers only hedging language, suggesting algorithmic generation. Moreover, there is mechanical repetition of connectors and balan... | {
"rationale": "First, the reply is generic and lacks personal detail or sensory context, which is common in AI-generated text. Second, its balanced, formal tone and use of hedging ('It is unclear...') are typical of AI's tendency to be neutral and noncommittal. Moreover, the lack of specific examples or emotional co... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 196 |
deepseek-v3.2_3856 | ai | We took a look at a one-dimensional spin-glass model where the spins have a lot of components and their interactions weaken with distance. We also checked out a thinned-out version of the model, but it ended up behaving quite differently from the original.
At absolute zero, we figured out the "defect energy" by compar... | null | TechnicalWriting | {
"rationale": "First, the prose is smooth, generic, and uses balanced transitional phrases with few concrete experimental details, a pattern common in AI-generated summaries. Second, there are subtle factual oddities (e.g., 'θ = 3/4 – σ' for a one-dimensional spin-glass and unusual thresholds like σ = 5/8) suggestin... | {
"rationale": "First, the text uses very even, generic phrasing and lacks experimental detail or nuance (no data, no derivation steps), which is common in AI-generated summaries. Second, there are odd or imprecise technical claims—e.g., 'θ = 3/4 – σ' for a spin-glass transition and the specificσ threshold—suggesting... | {
"rationale": "First, the writing is very polished and generic with smooth transitions and few concrete experimental details, suggesting algorithmic summarization rather than firsthand reporting. Second, there are subtle oddities—e.g., the specific-looking 'θ = 3/4 – σ' relation and the unusual 'σ = 5/8' threshold—t... | {
"rationale": "First, the writing is very polished and generic with smooth transitions and lacks experimental detail or data, suggesting algorithmic summarization. Second, there are odd phrasings (e.g., 'thinned-out version... but it ended up behaving quite differently') and neat, plausible-but-inaccurate numeric cl... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 343 |
deepseek-v3.2_1000 | ai | Hey, if the "Zombie Chronicles" crew ever sees this, here's some friendly advice:
1. In a twist-ending movie, maybe don't put close-ups of EVERY death in the opening credits? Kinda spoils the surprise...
2. I get you were on a super tight budget, and honestly, you pulled off some miracles with it. But next time, mayb... | null | Entertainment | {
"rationale": "First, the reply uses generic, evenly balanced critiques and a polished, neutral tone common in AI outputs rather than a distinct personal voice. Second, it includes a plausible but unverified-specific name and maintains consistent sentence lengths and structure, patterns often seen in machine-generat... | {
"rationale": "First, the reply uses generic, evenly balanced critique phrases (e.g., 'you pulled off some miracles') and lacks vivid personal details, which is common in AI outputs. Second, there are mechanical patterns—repeated hedging ('maybe'), formulaic transitions ('But next time...'), and a tidy numbered list... | {
"rationale": "First, the reply uses generic, evenly balanced critique phrases ('kinda spoils the surprise', 'big props') and lacks concrete scene-specific details, which is common in AI outputs. Second, there are mechanical patterns—repetitive hedging ('maybe'), abrupt topic shifts, and a canned closing ('Only for ... | {
"rationale": "First, the reply uses generic, evenly balanced critique phrases (\"kinda spoils the surprise\", \"big props...\") and lacks vivid personal details, which is common in AI outputs. Second, there are mechanical patterns—repeated 'honestly', abrupt topic shifts, and a formulaic numbered list—that suggest ... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 349 |
deepseek-v3.2_1263 | ai | We're doing our part to save energy by cutting back on electricity use whenever we can. Our office and computer gear—like workstations, printers, and copiers—are set to power-saving mode. And please help out by turning off the lights in any empty offices or workspaces you see! | null | Environmental | {
"rationale": "First, the passage uses generic, polished phrasing ('doing our part', 'set to power-saving mode') and lacks specific details, a common trait of AI-generated copy. Second, its balanced sentence lengths, neutral tone, and absence of personalization or sensory detail suggest algorithmic generation rather... | {
"rationale": "First, the passage uses generic, polished phrasing ('doing our part... save energy') and lacks personal details or sensory specifics, which is common in AI outputs. Second, its structure is highly balanced and formulaic (setup, list, call-to-action) and it makes broad generalizations rather than offer... | {
"rationale": "First, the message uses generic, polished phrasing ('doing our part... save energy') and lacks personal details or sensory specifics, which is common in AI outputs. Second, its structure is highly balanced and formulaic (setup, list, appeal) and it makes broad, neutral claims rather than drawing on li... | {
"rationale": "First, the text uses highly generic, polished phrasing ('doing our part', 'set to power-saving mode') suggesting algorithmic generation rather than a distinct personal voice. Second, it lacks specific details or anecdotes and relies on safe, universally acceptable cues (offices, workspaces) which is c... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 209 |
deepseek-v3.2_983 | ai | I'm honestly surprised anyone liked this movie. The plot is a total mess and the characters are completely wooden. Michael Bowen just spends the whole time spying on everyone and getting them all wrong. Nobody has a single interesting or redeeming quality. This isn't edgy, imaginative, or clever. It's not even straight... | null | Entertainment | {
"rationale": "First, the review uses highly generic, evenly balanced negative phrases (\"total mess\", \"completely wooden\", \"not even straightforward\") and lacks concrete scene-level details, which is common in AI-generated text. Second, its structure is formulaic—setup, critique, alternative suggestions—with s... | {
"rationale": "First, the review uses highly generic, evenly balanced negative statements and lacks concrete scene-specific details, which is common in AI outputs. Second, its polished grammar, repetitive transitional phrases ('not even straightforward—it's just tedious'), and stock suggestions (switch to PBS, walk ... | {
"rationale": "First, the review uses highly generic, evenly balanced negative phrases (\"total mess\", \"completely wooden\", \"not even straightforward\") and lacks concrete scene-specific details, which is common in AI outputs. Second, its structure is formulaic—setup, critique, alternative suggestions—with polis... | {
"rationale": "First, the review uses highly generic, evenly balanced negative phrases (\"total mess\", \"completely wooden\", \"not even straightforward\") suggesting algorithmic generation rather than a distinct personal voice. Second, it lacks concrete scene-specific details and relies on stock transitions (\"If ... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 318 |
deepseek-v3.2_3567 | ai | From exhibition to institution: the game first dazzled Olympic crowds on the courts of St. Louis in 1904, a vibrant demonstration of skill. It wasn't until the 1936 Berlin Games, however, that men's basketball officially took its place on the medal podium, forever changing the Olympic landscape. | null | Sports | {
"rationale": "First, the passage uses polished, generic phrasing ('dazzled...', 'vibrant demonstration') and smooth transitional markers, patterns common in AI-generated text. Second, it makes broad, uncited historical claims (men's basketball officially on the medal podium in 1936) without nuance or sourcing, a ty... | {
"rationale": "First, the passage uses polished, generic phrasing ('dazzled...', 'vibrant demonstration') and lacks concrete details or sources, which is common in AI-generated text. Second, it makes a broad factual claim (men's basketball officially taking its place on the medal podium in 1936) without nuance or su... | {
"rationale": "First, the passage uses polished, generic transitional phrases ('From exhibition... It wasn't until...') and balanced, high-level statements, which are common in AI-generated text. Second, it lacks concrete details (no names, dates beyond a single decade) and offers broad generalizations rather than s... | {
"rationale": "First, the passage uses very polished, generic phrasing (e.g., 'vibrant demonstration of skill' and 'forever changing the Olympic landscape') which is common in AI-generated text. Second, it lacks concrete details, personal voice, or specific evidence and instead offers broad, balanced statements sugg... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 216 |
deepseek-v3.2_3733 | ai | Artistic gymnastics was introduced at the inaugural Olympic Games in Athens in 1896 and has been included in every subsequent edition. Initially, the program encompassed disciplines not traditionally considered artistic, such as climbing and acrobatics. | null | Sports | {
"rationale": "First, the passage is very concise, evenly balanced, and uses generic transitional phrasing ('was introduced... and has been included'), a pattern common in AI-generated text. Second, it offers broad, surface-level facts with no sourcing or nuance (e.g., 'initially... not traditionally considered') su... | {
"rationale": "First, the passage is very brief and uses neutral, encyclopedic phrasing common in AI outputs. Second, it makes a broad factual claim ('introduced at the inaugural Olympic Games... in 1896') without nuance or sourcing, which is typical of AI generalizations. Moreover, there is mild redundancy (mention... | {
"rationale": "First, the passage is very brief, evenly balanced, and uses generic encyclopedic phrasing common in AI outputs. Second, it lacks personal voice, nuance, or specific sourcing beyond a single year, suggesting algorithmic summarization rather than human-authored depth. Moreover, there is mild redundancy ... | {
"rationale": "First, the text is very brief, neutral, and generic, resembling AI-generated explanatory output. Second, it uses smooth, balanced transitions and lacks personal voice or anecdotal detail, suggesting algorithmic generation. Moreover, there is mild redundancy ('in every subsequent edition' after noting ... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 198 |
deepseek-v3.2_1971 | ai | Under current law, the Debtor must complete the Plan to receive a discharge (11 U.S.C. § 1141(d)(5)). Therefore, if the Debtor fails to perform, Oshuntola would suffer minimal prejudice from the Plan's confirmation. The Court interprets 11 U.S.C. § 108 as not extending this period. Although Oshuntola did not object to ... | null | LegalDocument | {
"rationale": "First, the passage uses generic, evenly balanced legal phrasing and smooth transitions with few case-specific details, which is common in AI-generated text. Second, there are stylistic signs like repetitive sentence openings ('First...'), lack of jurisdictional nuance, and a dangling 'modification' wi... | {
"rationale": "First, the passage uses generic, evenly balanced legal phrasing and smooth transitions that are common in AI-generated text. Second, it lacks case-specific detail and includes a dangling 'modification ends with the phrase:' marker suggesting truncated generation rather than a finished human draft. Mor... | {
"rationale": "First, the passage uses generic, evenly balanced legal phrasing and smooth transitions that are common in AI-generated text rather than showing distinctive human argumentation. Second, there are factual/temporal inconsistencies (citing § 1141(d)(5) and referencing objection waivers) suggesting lack of... | {
"rationale": "First, the passage uses generic, evenly balanced legal phrasing and lacks case-specific detail or original analysis, which is common in AI-generated text. Second, there are mechanical transitions ('Therefore', 'Although', 'Historically') and a tendency to summarize rather than deeply argue, suggesting... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 289 |
deepseek-v3.2_135 | ai | Multipliers are the biggest power and space hogs in digital deep neural networks. We trained a bunch of top-tier neural networks (called Maxout networks) on three popular datasets: MNIST, CIFAR-10, and SVHN. We tried them with three different number formats: floating point, fixed point, and dynamic fixed point. For eac... | null | AcademicResearch | {
"rationale": "First, the passage has a polished, generic explanatory tone with smooth transitions and lacks concrete experimental details or citations, which is common in AI-generated summaries. Second, it repeats simple ideas (e.g., 'you really don't need high precision') and uses balanced, neutral phrasing typica... | {
"rationale": "First, the passage has a polished, generic explanatory tone with smooth transitions and lacks concrete experimental details or citations, which is common in AI-generated summaries. Second, it repeats simple ideas (e.g., 'you really don't need high precision') and uses illustrative numeric specifics ('... | {
"rationale": "First, the passage has a polished, generic explanatory tone with smooth transitions and lacks concrete experimental details or citations, which is common in AI-generated summaries. Second, it repeats simple ideas (e.g., 'you really don't need high precision') and uses illustrative numeric specifics ('... | {
"rationale": "First, the passage is generic and high-level with polished transitions and lacks experimental detail or citations, which is common in AI-generated summaries. Second, it repeats simple ideas (e.g., 'you really don't need high precision') and uses balanced, neutral phrasing suggesting algorithmic genera... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 286 |
deepseek-v3.2_598 | ai | Pursuant to the Dodd–Frank Wall Street Reform and Consumer Protection Act of 2010, the Volcker Rule mandates a degree of institutional separation between investment banking and commercial banking services. | null | Business | {
"rationale": "First, the sentence is concise, factual, and uses specific legal phrasing (Dodd–Frank Wall Street Reform and Consumer Protection Act of 2010, Volcker Rule) in a way typical of human-authored encyclopedia text. Second, it exhibits natural stylistic choices — for example the use of 'Pursuant to' and a d... | {
"rationale": "First, the sentence is highly polished, neutral, and encyclopedic in tone, which is common in AI-generated text. Second, it uses precise legal phrasing ('the Dodd–Frank Wall Street Reform and Consumer Protection Act of 2010') but provides no expansion or nuance, suggesting formulaic generation. Moreov... | {
"rationale": "First, the sentence is very polished, evenly balanced, and uses generic encyclopedic phrasing common in AI outputs. Second, it lacks personal voice, specific contextual cues, or minor stylistic idiosyncrasies typical of human writing and instead provides a neutral, surface-level summary. Moreover, the... | {
"rationale": "First, the sentence is very polished, generic, and encyclopedic in tone, which often characterizes AI-generated text. Second, it lacks personalization, nuance, or minor flaws common in human writing and instead uses balanced, evenly distributed clauses suggesting algorithmic generation. Moreover, the ... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 189 |
deepseek-v3.2_2731 | ai | A legal curtain has been drawn around a pivotal digital lockpick, and a coalition of news organizations is demanding it be lifted. In ongoing court proceedings, they argue the FBI’s secrecy lacks justification, stressing their request steers clear of true national security risks. The focus is on a shadowy transaction: ... | null | NewsArticle | {
"rationale": "First, the prose is polished and generic, relying on balanced, expository clauses and stock transitions rather than distinctive human idiosyncrasies. Second, it uses plausible but vague specifics (e.g., 'ten wrong passcode attempts' with a $1.3 million price tag) and smooth rhetorical framing common t... | {
"rationale": "First, the prose is polished and generic, relying on balanced, expository sentence patterns common in AI-generated summaries rather than distinctive human stylistic fingerprints. Second, it uses plausible but vague specifics (e.g., 'ten wrong passcode attempts' with a $1.3 million price tag) and avoid... | {
"rationale": "First, the prose is polished and generic, relying on balanced, expository clauses and stock transitions rather than distinctive human idiosyncrasies. Second, there are subtle factual inconsistencies (e.g., calling Syed Rizwan Farook one of the San Bernardino attackers) and vague sourcing patterns comm... | {
"rationale": "First, the prose is polished, generic, and uses balanced, neutral clauses common in AI-generated summaries rather than distinctive human reporting voice. Second, it lacks concrete sourcing or attribution (no dates or links) and relies on vague attributions like 'the FBI turned to an unnamed third part... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 426 |
deepseek-v3.2_2749 | ai | Engineers from Network Rail and Thames Water are on site after a hole opened up in the track at Forest Hill on Monday, causing major disruption expected to last until Wednesday. Southern, Thameslink, and London Overground services are all affected.
Thames Water confirmed a sewer beneath the tracks had collapsed, descr... | null | NewsArticle | {
"rationale": "First, the piece uses a neutral, evenly balanced news tone with smooth transitions and few unique details, producing a generic summary rather than original reporting. Second, it includes specific but plausibly fabricated details (exact tonnage, exact times) and makes broad causal links (Forest Hill co... | {
"rationale": "First, the piece uses a polished, even-toned news-wire style with generic transitions and few sensory details or direct quotes beyond one brief attribution, suggesting algorithmic summarization. Second, it shows repetitive phrasings (multiple 'working ... to...') and balanced paragraph structures comm... | {
"rationale": "First, the piece uses a polished, evenly balanced news-wire tone with generic transitions and few sensory details or direct quotes, suggesting algorithmic summarization rather than original reporting. Second, there are factual inconsistencies (e.g., mentioning 'approximately 50 tonnes of ballast' with... | {
"rationale": "First, the piece uses a highly polished, even-toned news-wire style with generic transitions and few sensory details or direct quotes beyond one line each, suggesting algorithmic summarization. Second, it repeats phrases (e.g., 'working ... as fast as possible') and balances clauses in a way common to... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 538 |
deepseek-v3.2_1685 | ai | This dog food is excellent for canines of all ages. I have a three-year-old dog and a puppy, and both maintain soft coats and experience minimal illness. The convenience of Amazon Prime shipping further enhances the value of this product. | null | FoodCusine | {
"rationale": "First, the text uses generic, polished praise and lacks concrete details (no brand name, specifics about the product), which is common in AI-generated copy. Second, its balanced sentence lengths, neutral tone, and smooth transitions suggest algorithmic generation rather than a distinct personal voice.... | {
"rationale": "First, the text uses generic, polished phrasing ('excellent for canines of all ages', 'further enhances the value') common in AI outputs. Second, it lacks concrete sensory details or personal anecdotes beyond a single age and breed mention, suggesting templated generation. Moreover, the balanced, neut... | {
"rationale": "First, the text uses generic, polished praise and lacks concrete specifics (no brand name, no sensory details), which is common in AI-generated copy. Second, its structure is formulaic—setup, benefit, personal touch, endorsement—and transitions are smooth in a way AI often produces when generating rev... | {
"rationale": "First, the text is very generic and lacks concrete details (no brand name, no specific symptoms), suggesting template-like generation. Second, its polished positive tone and balanced structure—claim, personal detail, benefit of prime shipping—are common in AI outputs. Moreover, there is mild repetitio... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 197 |
deepseek-v3.2_2548 | ai | Onlay prostheses provide superior postoperative mechanical alignment compared to all-polyethylene Inlay designs, largely due to their thicker construct. Given this advantage and their higher survivorship rates, Onlay prostheses are the preferred first option for medial UKR. | null | MedicalText | {
"rationale": "First, the passage uses polished, generic medical language and smooth transitions without providing supporting data or citations, a common trait of AI-generated text. Second, it makes strong clinical claims (e.g., 'largely due' to thicker construct and higher survivorship) without nuance or evidence, ... | {
"rationale": "First, the paragraph is very polished and uses generic transitional phrases ('Given this advantage...'), which is common in AI-generated text. Second, it makes broad claims (e.g., 'superior ... compared to all-polyethylene') without supporting details or data, suggesting a templated output. Moreover, ... | {
"rationale": "First, the sentence is very polished and uses generic transitional phrases ('Given this advantage...'), a pattern common in AI-generated text. Second, it makes strong claims (superior mechanical alignment, thicker construct) but provides no supporting details or data, suggesting formulaic generation r... | {
"rationale": "First, the sentence is very polished and generic, using balanced transitional phrases ('Given this advantage') which is common in AI-generated text. Second, it makes broad claims (superior mechanical alignment, higher survivorship) without supporting details or data, suggesting a templated, high-level... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 203 |
deepseek-v3.2_1179 | ai | At REPUBLIC, we take climate change seriously, just like any other risk. Things like stronger or more frequent hurricanes, tornadoes, and storms can definitely impact our business.
So, how do we handle it? Our team is always talking about new and existing risks. We bring our key people together to model and review the... | null | Environmental | {
"rationale": "First, the passage uses generic, polished corporate phrasing and smooth transitions without concrete specifics, which is common in AI-generated text. Second, it repeats high-level concepts (risk modeling, dedicated group, top leadership) and lacks personal anecdotes or idiosyncratic detail typical of ... | {
"rationale": "First, the passage uses generic corporate phrasing and smooth transitions without concrete specifics (no data, names, or examples), which is common in AI-generated text. Second, its balanced, evenly structured sentences and lack of personal nuance or small inconsistencies suggest algorithmic generatio... | {
"rationale": "First, the passage uses generic, polished corporate phrasing and smooth transitions without concrete specifics (no data, names, or examples), which is common in AI-generated text. Second, it repeats high-level patterns ('we...', 'our team...', 'bottom line:') in a balanced, uniform sentence structure ... | {
"rationale": "First, the passage uses very polished, generic corporate language and lacks concrete specifics (no data, examples, or unique anecdotes), which is common in AI-generated text. Second, its structure is highly balanced and repetitive with clear transitions ('So', 'For the big risks...'), suggesting algor... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 293 |
deepseek-v3.2_2348 | ai | Once a year, Shia LaBeouf volunteers for a shift at a suicide hotline. One afternoon, after returning from lunch and logging into his system, he received his first call.
"Suicide Prevention Hotline," he answered.
A distraught young woman named Lucy was on the line, weeping as she described her traumatic home life and... | null | LiteratureCreativeWriting | {
"rationale": "First, the prose is generic and expository with even pacing and minimal sensory detail, which is common in AI-generated text. Second, there are mechanical repetitions (multiple 'Suicide Prevention Hotline' openings), abrupt topic shifts, and oddly specific yet shallow details (Shia screaming, collapsi... | {
"rationale": "First, the prose is generic and expository with abrupt transitions and flat dialogue, which is common in AI-generated text. Second, there are mechanical inconsistencies (a suicide hotline call abruptly shifts to a comedic liquor-store rant) and odd stylistic choices like heavy use of bolding and sudde... | {
"rationale": "First, the prose is generic and expository with abrupt transitions and mechanical dialogue tags (e.g., 'he answered', sudden shifts), which is common in AI outputs. Second, there are odd factual mismatches and incongruous juxtapositions (Shia LaBeouf portrayed as volunteering at a hotline while also h... | {
"rationale": "First, the prose is unnaturally even and expository with abrupt scene shifts (intro, abuse, reaction, joke) and lacks sensory detail, a common trait of AI-generated text. Second, there are mechanical repetitions and odd transitions—linking a real person volunteering to a hotline abruptly to a crude in... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 565 |
deepseek-v3.2_2728 | ai | Heinz Christian Strache, the leader of the party, cited the handling of postal ballots as one of many procedural irregularities. "We are not merely reacting to a loss," he asserted. "This is fundamentally about safeguarding the democratic principles upon which our system rests." Their candidate was narrowly defeated by... | null | NewsArticle | {
"rationale": "First, the passage exhibits polished, neutral summarization with smooth transitions and generalized linking phrases (e.g., 'ultimately', 'initially') that are common in AI-generated text. Second, it uses multiple plausible-sounding specifics while remaining vague on sources and includes generic filler... | {
"rationale": "First, the passage exhibits polished, generic news-language with balanced transitions and hedging (e.g., 'initially... ultimately') rather than deep, original reporting, suggesting algorithmic summarization. Second, it includes plausible-sounding specifics but lacks attribution and shows mild repetiti... | {
"rationale": "First, the passage exhibits polished, generic news-language with smooth transitions and broad generalizations rather than deep, specific reporting, suggesting algorithmic summarization. Second, there are factual and stylistic inconsistencies (a 150-page submission versus brief mention, oddly precise v... | {
"rationale": "First, the passage exhibits polished, generic news-language with balanced transitions and hedging rather than deep, specific reporting, suggesting algorithmic summarization. Second, it repeats stock phrases (e.g., 'ultimately... ultimately') and supplies plausible-but-general details (exact percentage... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 673 |
deepseek-v3.2_1802 | ai | The civil service refers to a branch of government predominantly made up of career officials selected based on professional qualifications, not political appointment or election. Their employment generally continues across changes in political administration. | null | GovernmentPublic | {
"rationale": "First, the passage has a concise, encyclopedia-like tone with specific phrasings ('refers to' and 'not political appointment or election') typical of human-edited sources. Second, its focused single-sentence structure and use of a hyphenated subordinate clause reflect careful human editorial style rat... | {
"rationale": "First, the passage is concise and uses specific, contextually accurate details (civil service, career officials, professional qualifications) suggesting a knowledgeable human author. Second, its natural, economical phrasing and lack of excessive hedging or repetitive filler are more typical of human w... | {
"rationale": "First, the sentence is concise, well-structured, and uses precise domain terminology (civil service, career officials, professional qualifications), suggesting a knowledgeable human author. Second, there is no excessive hedging or repetitive filler common in many AI outputs, and the tone is natural ra... | {
"rationale": "First, the sentence is brief, well-structured, and uses natural academic phrasing ('refers to...'), suggesting a human writing a gloss or encyclopedia entry. Second, it lacks common AI markers like excessive hedging, mechanical repetition, or unnatural transitions and instead shows purposeful specific... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 187 |
deepseek-v3.2_4006 | ai | From Polyakov loop correlators, we extract the screening masses within the deconfined phase of (3+1)-dimensional SU(3) pure gauge theory at finite temperature in the vicinity of the deconfinement transition, considering two distinct channels defined by angular momentum and parity. The ratio of these masses is then comp... | null | TravelTourism | {
"rationale": "First, the passage uses highly polished, generic academic phrasing and a dense stack of advanced terms (Polyakov loop, deconfined phase, (3+1)-dimensional SU(3) pure gauge theory, three-state Potts model) in a way typical of AI-generated text that composes plausible but superficial content. Second, it... | {
"rationale": "First, the passage uses very polished, high-level academic phrasing but makes unusually broad and generic claims without providing concrete data or specific evidence, which is common in AI-generated text. Second, there are conceptual mismatches—for example comparing 'three-dimensional, three-state Pot... | {
"rationale": "First, the passage uses very polished, high-level academic phrasing but makes vague or imprecise connections (e.g., comparing ratios of mass spectra to 'three-state Potts model' quantum numbers), which is common in AI-generated text. Second, it repeats generic transitions ('we investigate... and we co... | {
"rationale": "First, the passage uses very polished, high-level academic phrasing but makes vague or imprecise connections (e.g., comparing ratios of mass spectra to 'three-state Potts model' in broken symmetry), which is common in AI-generated text. Second, it lacks concrete data, specific numerical results, or cl... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 305 |
deepseek-v3.2_2244 | ai | Write a story or poem called "This Side Of My Skin." This skin of mine has a story to tell. You can read the lines and guess at the history—the past that's written on a person with a face, with lips for breathing and eyes for seeing.
Some people might think it's ugly. Maybe they've never had to heal from something ter... | null | LiteratureCreativeWriting | {
"rationale": "First, the prose is polished, evenly paced, and uses generic transitional phrases (e.g., 'Some people might think...'), which is common in AI-generated text. Second, it lacks vivid sensory specifics and relies on abstract moralizing ('It’s a humbling thing') rather than concrete personal detail, sugge... | {
"rationale": "First, the piece uses generic, evenly paced language and repeated transitional phrases (e.g., 'Maybe...') which is common in AI-generated text. Second, it lacks deep sensory specifics and relies on abstract moralizing ('It’s a humbling thing'), suggesting algorithmic generalization rather than lived e... | {
"rationale": "First, the piece uses polished, evenly paced prose and common transitional phrases (e.g., 'Some people might think...') which is typical of AI-generated text aiming for neutrality. Second, it lacks vivid sensory specifics and shows balanced, generic imagery ('past that's written on a person with a fac... | {
"rationale": "First, the piece uses very even, polished prose with repeated transitional phrases ('Maybe...') and lacks vivid sensory specifics, which often indicates algorithmic generation. Second, there are mechanical repetitions (multiple mentions of 'this skin' and tidy emotional framing) suggesting template-dr... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 364 |
deepseek-v3.2_38 | ai | Machine learning models often exhibit sensitivity to carefully crafted input alterations. These subtle modifications, known as adversarial perturbations, are designed to mislead a model's output while remaining nearly indistinguishable from the original to a human observer. This work implements three distinct technique... | null | AcademicResearch | {
"rationale": "First, the passage contains domain-specific details (adversarial perturbations, PCA decomposition, supplementary materials) and a coherent academic-abstract structure typical of human-authored papers. Second, its phrasing shows purposeful variation, purpose, and a focused rhetorical flow rather than e... | {
"rationale": "First, the passage uses precise, domain-specific terminology (adversarial perturbations, PCA, supplementary materials) and makes concrete methodological claims typical of human academic writing. Second, its structure—clear problem statement, specific technique description, and mention of supplementary... | {
"rationale": "First, the passage uses precise, domain-specific terminology (adversarial perturbations, PCA, supplementary materials) and makes concrete methodological claims typical of human academic writing. Second, its structure—clear problem statement, specific technique description, and mention of additional de... | {
"rationale": "First, the passage uses precise, domain-specific terminology (adversarial perturbations, PCA, supplementary materials) suggesting familiarity with academic writing. Second, its structure—clear problem statement, specific methodological detail (least significant principal components), and reference to ... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 279 |
deepseek-v3.2_928 | ai | The Supreme Court upheld Static Control’s claim that Lexmark violated the Lanham Act’s false-advertising provision. To resolve the case, the Court first interpreted the statute’s meaning using traditional statutory construction principles. It rejected framing the question as one of prudential standing, noting that the ... | null | EducationMaterial | {
"rationale": "First, the passage uses polished, generic legal phrasing and a neat two-part structure without case-specific factual detail, a common trait of AI-generated summaries. Second, it makes broad generalizations (e.g., 'historically included...') and avoids original argumentation or citation beyond a single... | {
"rationale": "First, the passage uses highly polished, generic legal phrasing and clear signposting (e.g., 'To resolve the case...') without providing concrete factual detail, which is common in AI-generated summaries. Second, it repeats general standards ('these principles... provided a clearer framework') and lac... | {
"rationale": "First, the passage uses very polished, generic legal language and a formulaic structure (intro, two-part test, conclusion) common in AI-generated summaries. Second, it lacks citations and makes broad, generalized assertions about statutes and standing rather than providing original doctrinal depth or ... | {
"rationale": "First, the passage uses very polished, generic legal language and a mechanical structure (intro, two-part test, conclusion) common in AI-generated summaries. Second, it lacks citations and offers high-level, evenly balanced statements without deep doctrinal nuance, suggesting algorithmic generalizatio... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 427 |
deepseek-v3.2_3184 | ai | Stopped in after a Sunday walk in the park. The food and coffee were excellent, the service was friendly, and I love the vibe here. | null | ProductReview | {
"rationale": "First, the text is very generic and uses stock positive phrases ('excellent', 'friendly', 'love the vibe') without concrete details, which is common in AI outputs. Second, its short, evenly balanced sentences and lack of personal nuance or sensory specifics suggest algorithmic generation rather than a... | {
"rationale": "First, the reply is very brief and generic, offering only positive impressions without sensory detail or personal context. Second, its polished grammar and neutral tone are common in AI-generated outputs rather than idiosyncratic human speech. Moreover, there is a lack of hesitation or conversational ... | {
"rationale": "First, the reply is very brief and generic, offering only positive generalities rather than specific details about food, drink, or experience. Second, its polished grammar and neutral tone are common in AI outputs trained to produce clean, standalone reviews. Moreover, there is a lack of sensory detai... | {
"rationale": "First, the text is extremely brief, generic, and lacks personal details or sensory specifics, which is common in AI-generated outputs. Second, its polished positive tone and balanced structure ('The food and coffee were excellent...') match patterns often produced by language models. Moreover, there i... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 180 |
deepseek-v3.2_443 | ai | The Human Resources department functions as the central administrative entity responsible for executing comprehensive human resource management within an organization. Its purview encompasses a wide array of employment-related duties, including ensuring adherence to labor legislation and employment standards, conductin... | null | Business | {
"rationale": "First, the passage uses highly polished, generic corporate language and balanced parallel structures without concrete examples, which is common in AI-generated text. Second, it repeats abstract phrases ('purview', 'fundamentally') and maintains flawless grammar and neutrality, suggesting algorithmic g... | {
"rationale": "First, the passage uses highly polished, generic corporate language and balanced sentence structures without concrete examples or personal nuance, which is common in AI-generated text. Second, it repeats abstract phrases ('purview', 'fundamentally') and lists roles in a mechanical manner suggesting al... | {
"rationale": "First, the passage uses highly polished, generic corporate language and balanced parallelism (e.g., 'Its purview encompasses...') which is common in AI-generated text. Second, it lacks concrete specifics, examples, or personal nuance and instead offers broad, evenly weighted definitions suggesting alg... | {
"rationale": "First, the passage uses highly polished, generic corporate language and balanced parallelism without concrete examples, which is common in AI-generated text. Second, it lacks personal voice or idiosyncratic phrasing and repeats neutral adjectives (e.g., 'comprehensive', 'meticulous') suggesting algori... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 289 |
deepseek-v3.2_3986 | ai | RIS-aided millimeter wave systems improve robustness and coverage. This paper demonstrates that RIS can also enhance localization as a communication by-product. We use sparse reconstruction for high-resolution channel estimates, which are then mapped to position. However, the large arrays in these systems make sparse r... | null | TravelTourism | {
"rationale": "First, the paragraph is highly polished and uses generic academic phrasings ('This paper demonstrates...', 'Results from a realistic 3D indoor scenario') while lacking concrete experimental details or numbers, a common trait of AI-generated abstracts. Second, it repeats key phrases ('RIS', 'sparse rec... | {
"rationale": "First, the paragraph is highly polished and uses generic academic phrasing (e.g., 'This paper demonstrates...') while lacking concrete experimental details or numbers, which is common in AI-generated abstracts. Second, there are subtle oddities—'as a communication by-product' and using 'LoS' rather th... | {
"rationale": "First, the paragraph is highly polished and uses generic academic phrasing ('This paper demonstrates...', 'Results...confirm...'), a common trait of AI-generated abstracts. Second, it lacks concrete experimental details (no numbers, datasets, or specific parameter settings) and makes broad, rapid clai... | {
"rationale": "First, the paragraph is highly polished and generic, making broad claims (e.g., 'significantly improved localization accuracy') without presenting data or specifics, which is common in AI-generated abstracts. Second, it repeats buzzwords ('sparse reconstruction', 'multidimensional orthogonal matching ... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 292 |
deepseek-v3.2_708 | ai | def modp(n: int, p: int):
"""Compute (2^n) mod p efficiently.
>>> modp(3, 5)
3
>>> modp(1101, 101)
2
>>> modp(0, 101)
1
>>> modp(3, 11)
8
>>> modp(100, 101)
1
"""
result = 1
for _ in range(n):
result = (2 * result) % p
return result | null | Code | {
"rationale": "First, the code is very polished with type hints, uniform docstring examples, and no personal idiosyncrasies, which is common in AI-generated outputs. Second, there is a subtle choice like using result = 1 for caching instead of directly returning 1 and eliminating the else branch, a pattern often see... | {
"rationale": "First, the code is very polished and uses type hints and a balanced docstring with multiple example lines, a pattern common in AI-generated answers. Second, there is a subtle inconsistency—the name 'modp' is used for both modular exponent and this loop approach—which is a frequent sign of AI producing... | {
"rationale": "First, the code is very polished and uses type hints and doctest examples in a way common to AI-generated answers. Second, there is a subtle inconsistency—asking for modular exponent but implementing a simple loop—suggesting pattern-based generation rather than practical human coding. Moreover, the ch... | {
"rationale": "First, the code is very polished and generic, using type hints and a balanced docstring with multiple example lines, which is common in AI-generated snippets. Second, there is excessive uniformity in sentence length and style with no personalizing comments or idiosyncratic phrasing, suggesting algorit... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 283 |
deepseek-v3.2_3442 | ai | When you cross the Jordan River and settle in the land the LORD your God is giving you as your permanent home, and when He grants you peace from all the enemies surrounding you so that you live in security, then He will designate a special place for His name to reside. To that place you must bring everything I command:... | null | Religious | {
"rationale": "First, the passage uses polished, generic devotional language and smooth transitions with few concrete details, which is common in AI-generated text. Second, there are mild redundancies (e.g., 'special place for His name to reside') and balanced parallel lists ('your... your...') patterns often produc... | {
"rationale": "First, the passage uses polished, generic devotional language and smooth transitions that are common in AI-generated text. Second, there are subtle inconsistencies (e.g., 'I command' followed by broad communal listing) suggesting algorithmic generalization rather than precise biblical nuance. Moreover... | {
"rationale": "First, the passage has a polished, generic devotional tone with balanced transitions and few concrete details, which is common in AI-generated text. Second, there are subtle oddities—redundant clauses ('so that you live in security') and unusual capitalization of 'LORD'—suggesting algorithmic assembly... | {
"rationale": "First, the passage has a polished, generic devotional tone with balanced transitions and few concrete details, which is common in AI-generated text. Second, there is mechanical repetition (notably 'there') and parallel lists ('your... your...') suggesting algorithmic smoothing rather than natural huma... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 290 |
deepseek-v3.2_2116 | ai | The defendant argues that the sentencing judge improperly considered two unproven factors: first, a belief in her guilt on the more serious aggravated battery charges despite the jury's conviction only on the lesser simple battery offenses; and second, alleged misconduct involving contact with a victim after the incide... | null | LegalDocument | {
"rationale": "First, the passage uses highly polished, generic legal phrasing and broad assertions ('It is a well-established principle...') without case-specific factual development, a common sign of AI-generated text. Second, it cites multiple authorities in a formulaic, evenly balanced manner and includes plausi... | {
"rationale": "First, the passage uses highly polished, generic legal phrasing and broad assertions ('It is a well-established principle...') without case-specific factual development, a common trait of AI-generated text. Second, the structure is formulaic—stating the issue, citing precedents in order, and ending wi... | {
"rationale": "First, the passage uses highly polished, generic legal phrasing and broad assertions ('It is a well-established principle...') without case-specific factual development, a common trait of AI-generated text. Second, there are mechanical signs like excessive parallelism (two 'first... second...' lists),... | {
"rationale": "First, the passage uses highly polished, generic legal language and broad assertions ('It is a well-established principle...') without case-specific factual development, a common trait of AI-generated text. Second, the structure is overly neat and exhaustive—listing potential grounds, citing hypotheti... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 405 |
deepseek-v3.2_414 | ai | Fintech, a portmanteau of "financial technology," describes companies that use innovative technology to compete with traditional methods in delivering financial services. Its core pillars are often summarized as the "ABCDs": Artificial intelligence, Blockchain, Cloud computing, and big Data. Technologies like mobile ba... | null | Business | {
"rationale": "First, the passage has a clear, organized explanatory structure with specific examples (mobile banking, investing apps) and a cited origin ('portmanteau... financial technology'), suggesting a human summarizing known material. Second, its tone is neutral and textbook-like without excessive hedging or ... | {
"rationale": "First, the passage has a polished, generic explanatory tone with balanced transitions and no personal voice, which is common in AI-generated text. Second, it uses formulaic structures (a brief definition followed by 'Its core pillars...') and lists examples in a mechanical way, suggesting algorithmic ... | {
"rationale": "First, the passage has a polished, generic explanatory tone with balanced transitions and no personal voice, which is common in AI-generated text. Second, it uses formulaic structures (a brief definition followed by 'Its core pillars...') and lists examples in a mechanical way rather than offering nua... | {
"rationale": "First, the passage has a clear, concise structure and a personal tone (e.g., 'often summarized as the ABCDs') suggesting a human writer. Second, it uses specific but lightly touched examples (mobile banking, cryptocurrency) rather than overgeneralizing, which is common in human informal exposition. Mo... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 231 |
deepseek-v3.2_2626 | ai | A nearly £39 million stimulus package found its destination through the Welsh government's budget agreement with the Liberal Democrats. Finance Minister Jane Hutt stated this allocation would aid in "generating immediate benefits" economically. However, Plaid Cymru labeled the sum "completely inadequate," while the Con... | null | NewsArticle | {
"rationale": "First, the passage uses a polished, even journalistic tone with balanced paragraph transitions and generic summarizing phrases (e.g., 'According to the government...'), a common trait of AI-generated text. Second, it contains multiple specific but plausibly fabricated details — precise numbers without... | {
"rationale": "First, the prose is very even and encyclopedic with balanced paragraph transitions and few stylistic idiosyncrasies, which is common in AI-generated text. Second, there are subtle inconsistencies (the text cites '£39 million' then later '£38.9 million') and some mildly stilted phrasings ('found its de... | {
"rationale": "First, the piece uses a very even, polished news tone with balanced paragraph transitions and lacks concrete timestamps or named sources, suggesting algorithmic summarization. Second, there are mechanical repetitions (multiple mentions of allocations and initiatives) and some mildly awkward phrasings ... | {
"rationale": "First, the piece uses a very even, polished news tone with balanced paragraph transitions and lacks concrete timestamps or named sources, suggesting algorithmic summarization. Second, there is mechanical repetition (multiple 'These...') and some mildly stilted phrasing ('found its destination through ... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 729 |
deepseek-v3.2_569 | ai | Financial technology firms encompass both emerging startups and established financial or technology enterprises that aim to either supplant or improve upon the financial services offered by incumbent institutions. | null | Business | {
"rationale": "First, the sentence is highly generic, evenly balanced, and uses polished academic phrasing without concrete details or personal nuance, which is common in AI-generated text. Second, it exhibits mild redundancy (e.g., 'both... and' followed by 'either... or') and balanced clause structures typical of ... | {
"rationale": "First, the sentence is highly generic and uses balanced, formal phrasing ('both... and...') which is common in AI-generated text. Second, it lacks concrete details, examples, or a distinct voice and repeats safe constructions ('either... or'), suggesting algorithmic neutrality. Moreover, the polished,... | {
"rationale": "First, the sentence is highly polished and generic, using balanced clauses and formal vocabulary common in AI-generated text. Second, it lacks concrete details, examples, or a distinct personal voice, suggesting algorithmic generalization rather than human nuance. Moreover, there is mild redundancy (e... | {
"rationale": "First, the sentence is highly polished, generic, and uses balanced parallel structures (e.g., 'both...and...that aim...') which is common in AI-generated text. Second, it lacks concrete specifics, examples, or a distinct voice, suggesting algorithmic generalization rather than human experiential writi... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 180 |
deepseek-v3.2_2410 | ai | Research indicates that dexamethasone is a viable substitute for prednisone in managing mild to moderate pediatric asthma flare-ups, offering advantages in ease of use, taste, and affordability. Further studies are required to assess its effectiveness for children requiring hospitalization. | null | MedicalText | {
"rationale": "First, the sentence is highly polished, generic, and uses common transitional phrases ('Research indicates', 'Further studies are required') without providing specific evidence or novel insight, which is typical of AI-generated summaries. Second, it makes broad claims ('Research indicates...') but off... | {
"rationale": "First, the text is highly polished and uses generic, balanced phrasing common in AI outputs (e.g., 'Research indicates...') rather than providing specific data or citations. Second, it lacks personal nuance or stylistic idiosyncrasies typical of human writing and instead relies on formulaic transition... | {
"rationale": "First, the text is very polished and uses generic, high-level phrasing ('Research indicates...') without specific data or citation, which is common in AI-generated outputs. Second, it relies on neutral, balanced transitions ('Further studies are required...') and lacks personal voice or nuance, sugges... | {
"rationale": "First, the text is very polished, generic, and uses balanced transitional phrases (e.g., 'Research indicates...', 'Further studies are required'), which is common in AI-generated summaries. Second, it lacks personal voice, specific data, citations, or nuance and relies on broad terms like 'research in... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 202 |
deepseek-v3.2_3664 | ai | In 2007, FIBA submitted a proposal to the International Olympic Committee (IOC) to include 3x3 basketball in the programme for the Youth Olympic Games Singapore 2010. This inaugural event marked the discipline's first official appearance, and it has been a fixture in all subsequent Youth Olympic Games. On 9 June 2017, ... | null | Sports | {
"rationale": "First, the passage has an encyclopedic, citation-style tone with precise dates and specific details that match human-edited reference text. Second, its coherent chronological structure, use of subordinate clauses, and lack of typical AI overgeneralization or excessive hedging suggest human authorship.... | {
"rationale": "First, the passage has a neutral, encyclopedia-like tone with precise dates and specific details (e.g., 9 June 2017, IOC Executive Board, Olympic Games Tokyo 2020), suggesting synthesis from a factual source. Second, its structure is tightly organized with smooth transitions and lacks common AI marker... | {
"rationale": "First, the passage has a very neutral, encyclopedic tone with tightly structured factual sentences and no personal voice, which is common in AI-generated text. Second, there is mild redundancy (e.g., 'marked the discipline's first official appearance' followed by 'the first new discipline...') and a l... | {
"rationale": "First, the passage has a very polished, neutral, and encyclopedic tone with tightly structured factual statements, which is common in AI-generated text. Second, it lacks personal voice, anecdotes, or minor stylistic inconsistencies and repeats transitional phrases (e.g., 'This...'), suggesting algorit... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 322 |
deepseek-v3.2_1943 | ai | Western European, Byzantine, and Islamic societies, which mostly followed medical traditions from Hippocrates and Galen, also focused on prevention. They created health plans by checking out the local environment—looking at the landscape, winds, and sunlight, as well as testing the water and food sources for both peopl... | null | GovernmentPublic | {
"rationale": "First, the passage has a conversational, idiomatic tone (e.g., 'checking out the local environment') and specific cultural details (Western European, Byzantine, Islamic societies) that suggest a human writer. Second, it presents a compact, nuanced overview with varied sentence lengths and no excessive... | {
"rationale": "First, the passage has a concise, conversational tone with a specific detail ('testing the water and food sources') suggesting personal knowledge rather than generic filler. Second, its structure is varied and compact, with clear transitions like 'also' and a smooth paragraph flow common in human writ... | {
"rationale": "First, the paragraph is very brief, neutral, and uses generic transitional phrases ('also'), a pattern common in AI-generated text. Second, it offers high-level summaries without specific details, citations, or personal nuance, suggesting algorithmic generalization rather than human reflection. Moreov... | {
"rationale": "First, the passage is very polished, generic, and lacks personal voice or concrete details, suggesting algorithmic generation. Second, it uses balanced transitions and neutral summarizing language ('also') common in AI outputs rather than idiosyncratic human phrasing. Moreover, there is mild redundanc... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 213 |
deepseek-v3.2_3891 | ai | A growing spectrum of application scenarios demands the efficient and timely derivation of sophisticated insights from semantically enriched data streams, a need particularly prominent within the realms of the Web and the Internet of Things (IoT). Frequently, this process of knowledge extraction necessitates the applic... | null | TechnicalWriting | {
"rationale": "First, the passage uses highly polished, generic academic phrasing and makes sweeping claims without providing concrete experimental details, a common trait of AI-generated text. Second, it repeatedly uses grandiose adjectives ('groundbreaking', 'substantially superior') and smooth transitions in a wa... | {
"rationale": "First, the passage uses highly polished, generic academic phrasing and makes sweeping claims without supporting specifics, a common trait of AI-generated text. Second, it repeatedly concatenates buzzwords (e.g., 'groundbreaking evaluation methodology', 'substantially superior processing speeds') and n... | {
"rationale": "First, the passage uses highly polished, generic academic phrasing and makes sweeping claims without supporting details, a common trait of AI-generated text. Second, it repeatedly concatenates buzzwords (e.g., 'groundbreaking evaluation methodology', 'substantially superior processing speeds') and nam... | {
"rationale": "First, the passage uses highly polished, generic academic phrasing and excessive filler ('substantially', 'frequently') while avoiding concrete experimental details, a common trait of AI-generated text. Second, it makes sweeping claims (e.g., 'groundbreaking evaluation methodology' and speedups over C... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 388 |
deepseek-v3.2_460 | ai | Market segmentation is based on a core principle: distinct customer groups need tailored marketing strategies. This means customizing the offer, price, promotion, and distribution for each segment. | null | Business | {
"rationale": "First, the passage is very generic and uses balanced, textbook-like phrasing ('is based on a core principle', 'This means...'), a common pattern in AI-generated explanatory text. Second, it lacks concrete examples, personal voice, or idiosyncratic detail and repeats simple structures ('need...need') s... | {
"rationale": "First, the passage is very brief, generic, and uses clear, evenly balanced sentence structures common in AI-generated explanatory text. Second, it lacks personal voice, nuance, or concrete examples and repeats key phrases (e.g., 'segment'/'customizing') suggesting algorithmic generalization. Moreover,... | {
"rationale": "First, the text is very brief, evenly structured, and uses generic textbook phrasing common in AI outputs. Second, it lacks personal voice, specific examples, or nuanced explanation and repeats key terms (distinct/customer groups/need tailored marketing) suggesting algorithmic generation. Moreover, th... | {
"rationale": "First, the text is extremely brief, well-structured, and generic, traits common in AI-generated explanatory outputs. Second, it lacks personalization, concrete examples, or nuance and repeats simple vocabulary (e.g., 'segment'/'customized'), suggesting algorithmic templating. Moreover, the neutral, ev... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 184 |
deepseek-v3.2_993 | ai | Essentially, this film is just a feature-length Ramones performance, sporadically interrupted by moments of sheer idiocy and profound tedium. What sort of educational institution is this supposed to be?<br /><br />If you are not an ardent devotee of the Ramones, I must stress—DO NOT, under any circumstances, squander y... | null | Entertainment | {
"rationale": "First, the passage uses highly polished, generic negative phrases (e.g., 'sporadically interrupted by moments of sheer idiocy and profound tedium') and lacks concrete sensory details, a common trait of AI-generated text. Second, its balanced, formal sentence structures and canned transitions ('What so... | {
"rationale": "First, the passage uses highly polished, generic negative phrases (\"sporadically interrupted by moments of sheer idiocy and profound tedium\") and lacks concrete sensory details, which is common in AI-generated text. Second, there are mechanical markers like excessive hedging ('essentially', 'complet... | {
"rationale": "First, the passage uses highly polished, generic negative phrases (\"sporadically interrupted by moments of sheer idiocy and profound tedium\") and lacks concrete sensory details, which is common in AI-generated text. Second, there are mechanical markers like excessive hedging ('essentially', 'complet... | {
"rationale": "First, the passage uses highly polished, generic negative phrases (\"sporadically interrupted by moments of sheer idiocy and profound tedium\") and lacks concrete sensory details, a common trait of AI-generated text. Second, there is mechanical repetition and excessive hedging ('completely insufferabl... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 283 |
deepseek-v3.2_581 | ai | Consumers should personally check the social responsibility claims made by companies. The available records, audits, and reports offer a way for people to confirm that the products they buy are ethically and sustainably produced. | null | Business | {
"rationale": "First, the passage uses a concise, conversational tone with a personal recommendation ('Consumers should personally check...') which is typical of human-authored opinion pieces. Second, it offers specific, plausible details (social responsibility claims, records, audits, reports) without overgeneraliz... | {
"rationale": "First, the short, conversational tone and use of first-person 'Consumers' suggest a human author writing from direct experience. Second, the passage offers specific, pragmatic advice without overgeneralizing or adding excessive hedging, which is common in human casual writing. Moreover, there are no o... | {
"rationale": "First, the short, conversational tone and use of first-person 'Consumers' suggest a human author writing from direct experience. Second, the passage is specific and succinct rather than exhibiting AI's common overgeneralization or excessive hedging. Moreover, there are no repetitive patterns or mechan... | {
"rationale": "First, the short, conversational tone and use of first-person 'Consumers' suggest a human author writing from direct experience. Second, the passage is clear, specific (mentions checks of social responsibility claims, records, audits, and reports) and lacks common AI over-explanation or excessive hedg... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 190 |
deepseek-v3.2_742 | ai | Given a string, this function determines whether the length of the string is a prime integer, outputting True if it is prime and False otherwise.
Illustrative cases:
- prime_length('Hello') yields True
- prime_length('abcdcba') yields True
- prime_length('kittens') yields True
- prime_length('orange') yields ... | null | Code | {
"rationale": "First, the prose is very polished and formulaic with explicit signposting (Given..., Illustrative cases...) and balanced examples, which is common in AI-generated explanatory text. Second, it uses generic high-level descriptions ('the implementation first...') and lacks personal phrasing or idiosyncra... | {
"rationale": "First, the prose is very polished and formulaic with explicit signposting (Given..., Illustrative cases...) and lacks personal phrasing or idiosyncrasies, which often indicates AI generation. Second, the examples are perfectly balanced and generic (various lengths including a case with no trailing spa... | {
"rationale": "First, the prose is very polished and formulaic with explicit signposting (Given..., Illustrative cases...) and lacks personal phrasing or idiosyncratic errors, which often indicates AI generation. Second, the examples are perfectly balanced and generic (various lengths including a case with no traili... | {
"rationale": "First, the prose is very polished and generic, offering a high-level explanatory style common in AI-generated text. Second, it lacks personal phrasing or idiosyncratic errors and repeats structural patterns (e.g., '...yields ... yields'), suggesting algorithmic uniformity. Moreover, it makes an exhaus... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 294 |
deepseek-v3.2_1731 | ai | The variety packs are absolutely delicious!
Each morning, I enjoy a serving, and at a mere thirty cents per meal, it truly baffles me why these aren't a staple in every household across the globe.
The maple and brown sugar flavor is outstanding, with apples and cinnamon coming in a close second, followed by the clas... | null | FoodCusine | {
"rationale": "First, the text uses highly polished, generic positive adjectives and balanced sentence patterns typical of AI-generated promotional copy. Second, it lacks concrete personal details (no specific brand, timeframe, or sensory notes beyond general satisfaction) and repeats safe, broad claims, suggesting ... | {
"rationale": "First, the text uses highly polished, generic positive adjectives and repetitive sentence patterns (e.g., 'This...') common in AI-generated outputs. Second, there are odd phrasings ('observing as it magically doubles in size before my eyes') and mechanical transitions suggesting algorithmic assembly r... | {
"rationale": "First, the text uses excessive, generic praise and repetitive adjectives (e.g., 'epitome', 'astonishingly affordable') which is common in AI-generated outputs. Second, it lacks concrete sensory details beyond basic flavor names and includes odd mechanical instructions ('boil water... into a small pot'... | {
"rationale": "First, the text uses excessive, generic praise and mechanical transitions (e.g., 'epitome of convenience...') which is common in AI-generated outputs. Second, it lacks concrete sensory details about texture or specific taste and instead relies on broad positive statements, suggesting algorithmic gener... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 335 |
deepseek-v3.2_1327 | ai | As a group of car and home insurance companies, we've always known that weather and climate can cause serious losses. That's why we've always built our products, pricing, and safety nets with a close eye on the risk of big, weather-related claims. We constantly track and model these trends to help guide our decisions.
... | null | Environmental | {
"rationale": "First, the passage uses generic, polished corporate phrasing and smooth transitions without concrete personal details, a common trait of AI-generated text. Second, it includes plausible but vague specifics (e.g., 'we got approval to use a tool called Fireline' and a broad list of states) suggesting fi... | {
"rationale": "First, the passage uses very polished, generic corporate phrasing and repeats safe, high-level statements (e.g., 'we always...') rather than providing deep specifics, which is common in AI-generated text. Second, it includes plausible but vague references ('a tool called Fireline' and a 2017 approval)... | {
"rationale": "First, the passage uses very polished, generic corporate language and repeats high-level phrases ('we always...') in a way common to AI-generated text. Second, it lacks concrete, time-specific details (no year beyond 'recently' and no attribution) and includes plausible-but-general facts, suggesting a... | {
"rationale": "First, the passage uses very polished, generic corporate language and repeats phrases ('we've always...') in a way common to AI-generated text. Second, it lacks deep specifics or anecdotes and includes oddly broad geographic lists (e.g., Gulf Coast and lower-Atlantic states) suggesting template-driven... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 507 |
deepseek-v3.2_2779 | ai | Just heading out with the family today—should be fun, it always is. Still need to tidy my room if I want that new furniture, though. Mom’s a pro at setting up those little bargains. Ugh, fine. Okay, I’m out. | null | OnlineContent | {
"rationale": "First, the text uses short, generic sentences and common transitional phrases ('Still... though', 'Ugh, fine. Okay, I’m out') which are typical of AI-generated outputs. Second, it lacks deep sensory detail or personal nuance and instead offers balanced, neutral statements about routine and planning, s... | {
"rationale": "First, the passage uses short, generic sentences and lacks deep sensory detail or personal reflection, which is common in AI outputs. Second, there are mechanical transitions ('Ugh, fine. Okay, I'm out.') and balanced, neutral phrasing suggesting algorithmic smoothing rather than spontaneous human idi... | {
"rationale": "First, the text uses short, generic sentences and smooth transitions that are common in AI-generated outputs. Second, there are odd perspective shifts (a casual 'Ugh, fine. Okay, I’m out.' after starting 'with the family') suggesting lack of coherent situational grounding typical of AI. Moreover, it l... | {
"rationale": "First, the text uses very short, evenly balanced sentences and generic transitional phrases (e.g., 'Still need to...') which is common in AI-generated outputs. Second, there's a lack of deep sensory detail or personalization and some mechanical transitions ('Ugh, fine. Okay, I'm out.'), suggesting alg... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 204 |
deepseek-v3.2_1101 | ai | This cinematic endeavor follows the harrowing ordeal of a prominent actress filming on location in Africa, who is violently abducted and transported deep into the unforgiving jungle to be held captive for a substantial ransom. In a desperate bid for her rescue, the film's producer enlists a specialized individual to re... | null | Entertainment | {
"rationale": "First, the passage uses highly polished, generic critical phrasing and balanced sentence structures that are common in AI-generated text. Second, it shows repetitive transitions, mild contradictions (e.g., 'abruptly' followed by 'soporific') and occasional factual oddities suggesting algorithmic assem... | {
"rationale": "First, the passage uses highly polished, generic critical phrasing and excessive hedging (e.g., 'abundantly', 'unflinching') while avoiding concrete sensory detail, a common trait of AI-generated text. Second, there are mechanical repetitions and odd transitions (a sudden shift from critique to remark... | {
"rationale": "First, the passage exhibits highly polished, generic critical language and even pacing with few concrete sensory details, suggesting algorithmic generation rather than a human reviewer's idiosyncratic voice. Second, there are mechanical repetitions (multiple 'It...') and occasional factual/information... | {
"rationale": "First, the passage exhibits excessive, mechanical hedging and abstract generalities (e.g., 'abundant, unflinching depictions') rather than concrete sensory detail, a common trait of AI-generated text. Second, there are abrupt topic shifts and oddly specific but irrelevant details (VideoAsia/Terror Tal... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 600 |
deepseek-v3.2_2829 | ai | Authenticity is paramount. It is not worthwhile to compromise one's true self to gain approval. People should appreciate others for who they are, not for efforts made to cultivate their favor. | null | OnlineContent | {
"rationale": "First, the passage uses generic, balanced phrasing and lacks personal detail or concrete examples, a common trait of AI-generated text. Second, its neat three-sentence structure and polished transitions suggest algorithmic optimization for clarity rather than human idiosyncratic voice. Moreover, there... | {
"rationale": "First, the text is very brief and uses generic, balanced phrasing ('authenticity is paramount', 'not for efforts made to cultivate their favor') common in AI outputs. Second, it lacks personal detail, concrete examples, or idiosyncratic language and instead presents a neat triadic structure typical of... | {
"rationale": "First, the text is very brief and uses generic, widely applicable moral statements that AI often generate. Second, its polished grammar, lack of personal detail or anecdote, and balanced, neutral tone suggest algorithmic output rather than a lived voice. Moreover, there are no idiosyncratic phrasings ... | {
"rationale": "First, the text is very polished and generic, offering broad moral statements rather than personal detail, which is common in AI outputs. Second, its balanced, evenly structured sentences and lack of colloquialism or sensory grounding suggest algorithmic generation. Moreover, there are no typos or idi... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 187 |
deepseek-v3.2_3327 | ai | I visited this establishment after my husband brought home their lemon pepper wings, which are excellent. They offer a wide variety of flavors, from familiar options like lemon pepper and mild to distinctive New Orleans-style selections such as yaka mein and noonie bird. Their beverage selection is also noteworthy; the... | null | ProductReview | {
"rationale": "First, the review uses polished, generic positive phrases and balanced sentence patterns common in AI-generated text. Second, it includes odd or incongruous details (e.g., 'yaka mein' and 'noonie bird') and a sudden shift to recommending the proprietor, suggesting algorithmic blending of facts. Moreov... | {
"rationale": "First, the review uses polished, generic positive language and smooth transitions (e.g., 'The business location...') that are common in AI-generated text. Second, it lists specific but incongruous details ('lemon pepper wings' paired with yaka mein and noonie bird) suggesting associative blending rath... | {
"rationale": "First, the review uses polished, generic positive adjectives and balanced sentence patterns common in AI-generated text. Second, there are odd phrasings (e.g., 'not the most premium') and a sudden shift to a specific personal detail ('visited this establishment...') suggesting algorithmic assembly rat... | {
"rationale": "First, the review uses very polished, generic positive language and lacks vivid sensory detail beyond listing items, a common trait of AI-generated text. Second, there are odd or incongruous specifics (e.g., 'yaka mein' and 'noonie bird') suggesting algorithmic assembly rather than lived experience. M... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 291 |
deepseek-v3.2_1006 | ai | I'm a huge fan of Arnold Vosloo, so I was really happy to see him finally starring in a recent movie instead of just playing a small part.
My college film appreciation class makes it hard to ignore certain flaws, though. The movie was filmed in a deserted part of town with no background traffic or curious onlookers, w... | null | Entertainment | {
"rationale": "First, the prose is very even, generic, and uses balanced transitional phrases (e.g., 'in the end') in a way common to AI-generated text. Second, it lacks vivid sensory detail or idiosyncratic personal specifics and instead offers structured, neutral critiques, suggesting algorithmic summarization rat... | {
"rationale": "First, the review uses generic, evenly balanced phrasing and lacks sensory specifics (no dialogue, minimal setting), a common trait of AI-generated text. Second, it shows repetitive transitional patterns ('Overall', 'In the end') and hedging language ('some continuity glitches... they're easy to overl... | {
"rationale": "First, the writing uses generic, evenly balanced phrasing and frequent transitional phrases (e.g., 'overall', 'in the end') common in AI-generated text. Second, it lacks vivid sensory detail or personal idiosyncrasies and shows mild repetition and smoothing that AI often produces. Moreover, there are ... | {
"rationale": "First, the writing uses very even, generic review phrases (\"I was really happy...\", \"in the end...\") and lacks sensory detail, suggesting algorithmic generation. Second, there are mechanical transitions and balanced sentence patterns common in AI outputs rather than a distinct personal voice. More... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 345 |
deepseek-v3.2_2290 | ai | God presents three doors. The first promises eternal happiness; the second, infinite knowledge. Then there is Door Three—the one He has never opened.
Eternal happiness… what does that even mean? A place where you never feel sorrow? I remember sitting on the examination table, waiting. The big wooden door stayed shut. ... | null | LiteratureCreativeWriting | {
"rationale": "First, the prose is polished and evenly balanced with generic transitions and abstract philosophical statements rather than vivid sensory detail, which often indicates AI generation. Second, there are mechanical repetitions (e.g., 'Eternal happiness...') and tidy rhetorical questions that AI commonly ... | {
"rationale": "First, the prose is polished and evenly paced with balanced transitions and generic metaphors (eternal happiness, door representing choice), which is common in AI-generated text. Second, there are subtle inconsistencies—a neat three-door scheme with an abrupt closure and some clichéd phrasings like 'G... | {
"rationale": "First, the prose is polished and evenly paced with balanced transitions and generic metaphors (eternal happiness, knowing God), which is common in AI-generated text. Second, there are mechanical repetitions (the 'But... Or...' pattern) and lack of vivid sensory specificity, suggesting algorithmic smoo... | {
"rationale": "First, the prose is polished and evenly paced with balanced clauses and generic transitions, which often indicates algorithmic generation. Second, it relies on explicit philosophical exposition rather than showing through concrete sensory detail and repeats rhetorical questions, suggesting template-dr... | [
{
"content": "You are an expert in textual forensics. Your job is to decide whether the following text is AI-generated or human-written. Please reason step by step, considering the style, structure, and content of the text. Justify your decision by pointing out specific features or characteristics of the text t... | AI | 533 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.