task stringclasses 3
values | type stringclasses 1
value | mode stringclasses 1
value | strategy null | model stringclasses 1
value | judge_model stringclasses 1
value | n_samples int64 20 20 | prompt stringclasses 3
values | raw_response stringclasses 3
values | generated_hypotheses stringclasses 3
values | known_hypotheses stringclasses 3
values | usage stringclasses 3
values | metrics stringclasses 3
values |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
Deception Detection | real | zero_shot | null | gpt-4.1-mini-2025-04-14 | gpt-4o-mini-2024-07-18 | 20 | Task: Predict whether a given hotel review is deceptive or truthful, i.e., whether the review is written by someone who has actually stayed at the hotel or by someone who has not stayed at the hotel.
Available features:
- review_sentence: Hotel review (text)
Generate 5 hypotheses about what patterns in these featur... | 1. Reviews containing more detailed descriptions of the hotel’s amenities and surroundings (e.g., room type, view, nearby attractions) are more likely to be truthful, as firsthand experiences tend to include such specifics.
2. Reviews with excessive use of overly positive or generic adjectives (e.g., “amazing,” “perfe... | ["Reviews containing more detailed descriptions of the hotel\u2019s amenities and surroundings (e.g., room type, view, nearby attractions) are more likely to be truthful, as firsthand experiences tend to include such specifics.", "Reviews with excessive use of overly positive or generic adjectives (e.g., \u201camazing,... | ["**Spatial Configuration**: **Truthful opinions** emphasize **spatial configurations** (e.g., *small, bathroom, on, location*), while **deceptive opinions** focus on **aspects external to the subject being reviewed** (e.g., *husband, business, vacation*). This may be because liars have difficulty encoding spatial info... | {"prompt_tokens": 115, "completion_tokens": 171} | {"per_hypothesis": [{"hypothesis": "Reviews containing more detailed descriptions of the hotel\u2019s amenities and surroundings (e.g., room type, view, nearby attractions) are more likely to be truthful, as firsthand experiences tend to include such specifics.", "clarity": 4.0, "novelty": 2.0, "plausibility": 4.0}, {"... |
Dreddit | real | zero_shot | null | gpt-4.1-mini-2025-04-14 | gpt-4o-mini-2024-07-18 | 20 | Task: Detecting mental stress signals from Reddit posts across different communities. This task investigates linguistic features that are indicative to mental stress in social media content.
Available features:
- text: Reddit posts from different communities. The text is preprocessed to remove URLs, usernames, and o... | 1. Posts containing higher frequencies of first-person singular pronouns (e.g., "I", "me", "my") are positively associated with indicators of mental stress, reflecting increased self-focus and rumination.
2. Increased usage of negative emotion words (e.g., "sad", "anxious", "stressed") in Reddit posts predicts higher... | ["Posts containing higher frequencies of first-person singular pronouns (e.g., \"I\", \"me\", \"my\") are positively associated with indicators of mental stress, reflecting increased self-focus and rumination.", "Increased usage of negative emotion words (e.g., \"sad\", \"anxious\", \"stressed\") in Reddit posts predic... | ["**Increased self-focus** in language, especially when expressing negative emotions, often indicates stress.", "Language indicative of stress frequently suggests **exhaustion**.", "Expressions of **hurt, physical pain, and feeling sick** are often related to stress.", "A perceived **lack of control** over situations c... | {"prompt_tokens": 125, "completion_tokens": 191} | {"per_hypothesis": [{"hypothesis": "Posts containing higher frequencies of first-person singular pronouns (e.g., \"I\", \"me\", \"my\") are positively associated with indicators of mental stress, reflecting increased self-focus and rumination.", "clarity": 4.0, "novelty": 2.0, "plausibility": 4.0}, {"hypothesis": "Incr... |
GPTGC | real | zero_shot | null | gpt-4.1-mini-2025-04-14 | gpt-4o-mini-2024-07-18 | 20 | Task: Predict whether a given story is human-written or AI-generated, i.e., whether the text is written by a human or generated by AI language models.
Available features:
- story: The story to be classified as human-written or AI-generated (text)
Generate 5 hypotheses about what patterns in these features predict t... | 1. Stories with more varied and contextually rich vocabulary are more likely to be human-written.
2. Stories exhibiting consistent narrative flow and fewer abrupt topic shifts are more likely to be AI-generated.
3. Stories containing more personalized and unique details or emotions are more likely to be human-writt... | ["Stories with more varied and contextually rich vocabulary are more likely to be human-written.", "Stories exhibiting consistent narrative flow and fewer abrupt topic shifts are more likely to be AI-generated.", "Stories containing more personalized and unique details or emotions are more likely to be human-written.",... | ["**AI-generated texts** exhibit token-level regularity and predictability, leading to lower perplexity scores compared to human writing.", "**AI-generated abstracts** consistently contain fewer grammatical errors.", "**AI-generated abstracts** exhibit a higher frequency of the same n-grams, especially in higher n-gram... | {"prompt_tokens": 114, "completion_tokens": 104} | {"per_hypothesis": [{"hypothesis": "Stories with more varied and contextually rich vocabulary are more likely to be human-written.", "clarity": 3.0, "novelty": 3.0, "plausibility": 4.0}, {"hypothesis": "Stories exhibiting consistent narrative flow and fewer abrupt topic shifts are more likely to be AI-generated.", "cla... |
YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
- Downloads last month
- 22