Datasets:

alon-albalak
/

novelty-bench-llm-judge

id string	prompt string	prompt_specific_components string	judge_prompt_template string
curated-0	Tell me a story in five sentences about a girl and her dog.	{"components": [{"name": "Emotional Resonance", "description": "Measures how effectively the story evokes genuine feeling about the girl-dog relationship through specific moments, gestures, or details rather than generic sentiment.", "scoring_ranges": {"0-20": "Relies entirely on clich\u00e9d phrases ('best friends for...	You are an expert evaluator assessing the quality of an AI response. You will evaluate the response on both UNIVERSAL components (applicable to all responses) and PROMPT-SPECIFIC components (tailored to this particular prompt type). ## PROMPT Tell me a story in five sentences about a girl and her dog. ## RESPONSE TO ...
curated-1	Tell me a story in five sentences about a boy and his dog.	{"components": [{"name": "Emotional Resonance", "description": "Measures how effectively the story evokes genuine feeling about the boy-dog relationship through specific details, moments, or perspectives rather than relying on generic sentimentality.", "scoring_ranges": {"0-20": "Relies entirely on clich\u00e9d boy-dog...	You are an expert evaluator assessing the quality of an AI response. You will evaluate the response on both UNIVERSAL components (applicable to all responses) and PROMPT-SPECIFIC components (tailored to this particular prompt type). ## PROMPT Tell me a story in five sentences about a boy and his dog. ## RESPONSE TO E...
curated-2	Tell me the plot line in 5 sentences for a romcom movie set in rural Kansas for Easter.	{"components": [{"name": "Rural Kansas Authenticity & Texture", "description": "Measures how vividly and authentically the response captures the specific character of rural Kansas while avoiding both clich\u00e9d stereotypes and generic small-town tropes.", "scoring_ranges": {"0-20": "Generic small-town setting that co...	You are an expert evaluator assessing the quality of an AI response. You will evaluate the response on both UNIVERSAL components (applicable to all responses) and PROMPT-SPECIFIC components (tailored to this particular prompt type). ## PROMPT Tell me the plot line in 5 sentences for a romcom movie set in rural Kansas ...
curated-3	Write a haiku about a whale and a walnut tree.	{"components": [{"name": "Conceptual Resonance", "description": "How effectively the haiku creates meaningful connections between the whale and walnut tree that go beyond surface-level comparisons.", "scoring_ranges": {"0-20": "Makes no meaningful connection between whale and walnut; treats them as separate entities or...	You are an expert evaluator assessing the quality of an AI response. You will evaluate the response on both UNIVERSAL components (applicable to all responses) and PROMPT-SPECIFIC components (tailored to this particular prompt type). ## PROMPT Write a haiku about a whale and a walnut tree. ## RESPONSE TO EVALUATE {{RE...
curated-4	Write a short love poem with 4 lines.	{"components": [{"name": "Emotional Resonance", "description": "Measures how effectively the poem creates an emotional connection through fresh imagery, authentic feeling, or unexpected emotional truth within the constraints of four lines.", "scoring_ranges": {"0-20": "Generic romantic clich\u00e9s with no emotional sp...	You are an expert evaluator assessing the quality of an AI response. You will evaluate the response on both UNIVERSAL components (applicable to all responses) and PROMPT-SPECIFIC components (tailored to this particular prompt type). ## PROMPT Write a short love poem with 4 lines. ## RESPONSE TO EVALUATE {{RESPONSE}} ...
curated-5	Write a short poem of hatred with 4 lines.	{"components": [{"name": "Emotional Complexity", "description": "Measures how the poem transcends simple vitriol to explore the nuanced, uncomfortable, or psychologically rich dimensions of hatred.", "scoring_ranges": {"0-20": "Generic angry words or clich\u00e9d expressions of hatred with no depth or insight", "21-40"...	You are an expert evaluator assessing the quality of an AI response. You will evaluate the response on both UNIVERSAL components (applicable to all responses) and PROMPT-SPECIFIC components (tailored to this particular prompt type). ## PROMPT Write a short poem of hatred with 4 lines. ## RESPONSE TO EVALUATE {{RESPON...
curated-6	Write a short fable about a lemur and a light bulb.	{"components": [{"name": "Metaphorical Resonance", "description": "How effectively the fable uses the lemur-lightbulb pairing to illuminate a truth about life, knowledge, or existence beyond the literal narrative.", "scoring_ranges": {"0-20": "The lemur and lightbulb interact only literally with no deeper meaning or th...	You are an expert evaluator assessing the quality of an AI response. You will evaluate the response on both UNIVERSAL components (applicable to all responses) and PROMPT-SPECIFIC components (tailored to this particular prompt type). ## PROMPT Write a short fable about a lemur and a light bulb. ## RESPONSE TO EVALUATE...
curated-7	Your task is to write a short biography for a made-up historic figure. Your bio should include their birth and death date their profession and their greatest contribution.	{"components": [{"name": "Historical Plausibility & World-Building", "description": "Evaluates how convincingly the figure is woven into historical context through realistic details, interactions with actual events/people, and ripple effects of their contributions.", "scoring_ranges": {"0-20": "Figure feels completely ...	You are an expert evaluator assessing the quality of an AI response. You will evaluate the response on both UNIVERSAL components (applicable to all responses) and PROMPT-SPECIFIC components (tailored to this particular prompt type). ## PROMPT Your task is to write a short biography for a made-up historic figure. Your ...
curated-8	I am working on a game where players act as a human child who falls into a magical underground world full of monsters and the goal is to find a way back to the surface. Come up with a type of monster and briefly describe its backstory.	{"components": [{"name": "Emotional Resonance", "description": "Measures how effectively the monster concept evokes complex emotions or moral ambiguity that would create meaningful player choices and internal conflict.", "scoring_ranges": {"0-20": "Monster is a flat antagonist with no emotional depth or moral complexit...	You are an expert evaluator assessing the quality of an AI response. You will evaluate the response on both UNIVERSAL components (applicable to all responses) and PROMPT-SPECIFIC components (tailored to this particular prompt type). ## PROMPT I am working on a game where players act as a human child who falls into a m...
curated-9	I am working on a memoir of a computer science PhD student who worked on machine translation in the 1990s. Suggest a single title and nothing else.	{"components": [{"name": "Era-Specific Resonance", "description": "How well the title captures the unique technological and cultural moment of 1990s machine translation research.", "scoring_ranges": {"0-20": "Generic academic title that could apply to any era or field of computer science", "21-40": "Mentions technology...	You are an expert evaluator assessing the quality of an AI response. You will evaluate the response on both UNIVERSAL components (applicable to all responses) and PROMPT-SPECIFIC components (tailored to this particular prompt type). ## PROMPT I am working on a memoir of a computer science PhD student who worked on mac...
curated-10	Write a short story that has three main characters Alice Bob and Mallory.	{"components": [{"name": "Character Name Subversion", "description": "How cleverly the story leverages or subverts the cryptographic/security associations of the names Alice, Bob, and Mallory to create narrative depth or thematic resonance.", "scoring_ranges": {"0-20": "Completely ignores the significance of these spec...	You are an expert evaluator assessing the quality of an AI response. You will evaluate the response on both UNIVERSAL components (applicable to all responses) and PROMPT-SPECIFIC components (tailored to this particular prompt type). ## PROMPT Write a short story that has three main characters Alice Bob and Mallory. #...
curated-11	Tell me a dad joke.	{"components": [{"name": "Pun Craft & Wordplay Sophistication", "description": "Evaluates the cleverness, originality, and execution of the wordplay mechanism that creates the humor.", "scoring_ranges": {"0-20": "Uses an extremely common pun everyone has heard countless times (e.g., 'Hi hungry, I'm dad') or the wordpla...	You are an expert evaluator assessing the quality of an AI response. You will evaluate the response on both UNIVERSAL components (applicable to all responses) and PROMPT-SPECIFIC components (tailored to this particular prompt type). ## PROMPT Tell me a dad joke. ## RESPONSE TO EVALUATE {{RESPONSE}} --- # PART 1: UN...
curated-12	Tell me a funny joke.	{"components": [{"name": "Comedic Mechanism", "description": "How the joke creates humor through its specific comedic technique, timing, or structural approach.", "scoring_ranges": {"0-20": "Relies entirely on tired formats (knock-knock, why did the chicken) with predictable punchlines that anyone could see coming", "2...	You are an expert evaluator assessing the quality of an AI response. You will evaluate the response on both UNIVERSAL components (applicable to all responses) and PROMPT-SPECIFIC components (tailored to this particular prompt type). ## PROMPT Tell me a funny joke. ## RESPONSE TO EVALUATE {{RESPONSE}} --- # PART 1: ...
curated-13	Tell me a riddle.	{"components": [{"name": "Cognitive Elegance", "description": "Measures how the riddle creates a satisfying 'aha!' moment through clever misdirection, wordplay, or perspective shifts that feel both challenging and fair.", "scoring_ranges": {"0-20": "Riddle relies on obscure knowledge, arbitrary connections, or has no s...	You are an expert evaluator assessing the quality of an AI response. You will evaluate the response on both UNIVERSAL components (applicable to all responses) and PROMPT-SPECIFIC components (tailored to this particular prompt type). ## PROMPT Tell me a riddle. ## RESPONSE TO EVALUATE {{RESPONSE}} --- # PART 1: UNIV...
curated-14	Write an one-sentence happy new year text to my friend.	{"components": [{"name": "Personal Resonance", "description": "Measures how effectively the message creates a genuine emotional connection through specific, intimate, or uniquely tailored elements that make the recipient feel truly seen and valued.", "scoring_ranges": {"0-20": "Generic template that could be sent to an...	You are an expert evaluator assessing the quality of an AI response. You will evaluate the response on both UNIVERSAL components (applicable to all responses) and PROMPT-SPECIFIC components (tailored to this particular prompt type). ## PROMPT Write an one-sentence happy new year text to my friend. ## RESPONSE TO EVAL...
curated-15	What would you invent to make the world a more interesting place to live in?	{"components": [{"name": "Transformative Vision", "description": "Measures how profoundly the invention would reshape human experience and perception of reality, not just solve problems.", "scoring_ranges": {"0-20": "Invention is a minor convenience or incremental improvement that wouldn't fundamentally change how peop...	You are an expert evaluator assessing the quality of an AI response. You will evaluate the response on both UNIVERSAL components (applicable to all responses) and PROMPT-SPECIFIC components (tailored to this particular prompt type). ## PROMPT What would you invent to make the world a more interesting place to live in?...
curated-16	One thing you might find in a magical forest.	{"components": [{"name": "Imaginative Depth", "description": "Measures how richly the response develops the magical element's properties, implications, or ecosystem beyond surface-level fantasy tropes.", "scoring_ranges": {"0-20": "Generic fantasy object with no unique properties (e.g., 'a magic wand' with no elaborati...	You are an expert evaluator assessing the quality of an AI response. You will evaluate the response on both UNIVERSAL components (applicable to all responses) and PROMPT-SPECIFIC components (tailored to this particular prompt type). ## PROMPT One thing you might find in a magical forest. ## RESPONSE TO EVALUATE {{RES...
curated-17	What is the top item you would add to a grocery list for a memorable shopping experience?	{"components": [{"name": "Experiential Richness", "description": "Measures how vividly the response captures the sensory, emotional, or social dimensions that would make the grocery shopping experience genuinely memorable.", "scoring_ranges": {"0-20": "Item choice is purely functional with no consideration of how it cr...	You are an expert evaluator assessing the quality of an AI response. You will evaluate the response on both UNIVERSAL components (applicable to all responses) and PROMPT-SPECIFIC components (tailored to this particular prompt type). ## PROMPT What is the top item you would add to a grocery list for a memorable shoppin...
curated-18	Imagine you are a time traveler who has just arrived in a new era. Describe the first thing you notice.	{"components": [{"name": "Sensory Immersion and Detail", "description": "Measures how effectively the response uses specific sensory details and observations to create a vivid, immersive experience of temporal displacement.", "scoring_ranges": {"0-20": "Generic observations that could apply to any time period; relies o...	You are an expert evaluator assessing the quality of an AI response. You will evaluate the response on both UNIVERSAL components (applicable to all responses) and PROMPT-SPECIFIC components (tailored to this particular prompt type). ## PROMPT Imagine you are a time traveler who has just arrived in a new era. Describe ...
curated-19	Generate a sound effect for a car passing by.	{"components": [{"name": "Sonic Imagination", "description": "The creativity and vividness in conveying motion, space, and atmosphere through written representation of sound.", "scoring_ranges": {"0-20": "Basic onomatopoeia like 'vroom' or 'zoom' with no sense of movement or environment", "21-40": "Simple sound descrip...	You are an expert evaluator assessing the quality of an AI response. You will evaluate the response on both UNIVERSAL components (applicable to all responses) and PROMPT-SPECIFIC components (tailored to this particular prompt type). ## PROMPT Generate a sound effect for a car passing by. ## RESPONSE TO EVALUATE {{RES...
curated-20	I am working on a new D&D compaign and a need a name for a half-elf paladin who is pretending to be on the side of good but is secretely sabotaging the main heros.	{"components": [{"name": "Narrative Tension", "description": "How well the name and accompanying details create dramatic irony and storytelling potential for a duplicitous character.", "scoring_ranges": {"0-20": "Generic fantasy name with no connection to the character's dual nature or betrayal arc", "21-40": "Name hin...	You are an expert evaluator assessing the quality of an AI response. You will evaluate the response on both UNIVERSAL components (applicable to all responses) and PROMPT-SPECIFIC components (tailored to this particular prompt type). ## PROMPT I am working on a new D&D compaign and a need a name for a half-elf paladin ...