NoveltyBench

Tests whether a model can produce several genuinely different, still-good answers to the same prompt across repeated samples, rather than near-duplicate outputs; current models fall well short of human writers.

Also known as: novelty-bench, NoveltyBench: Evaluating Language Models for Humanlike Diversity

unassessed

This page is a discovery lead. Nobody has yet assessed it against the catalogue contract, so it carries no disposition. Absence of evidence here is not evidence of staleness.
Categorygeneration
Subcategoryoutput diversity across repeated generations (mode collapse)
Page statusactive
Metricdistinct_k (count of functionally distinct outputs among k generations) and utility_k (patience-weighted cumulative utility of novel outputs)
Directionhigher_is_better
Unitcount (distinct_k, out of k); utility points (utility_k, reward-model scale 1-10)
Dataset size1100
Dataset licenceMIT (code repository); the Hugging Face dataset card itself carries no separate data licence tag
PublisherCarnegie Mellon University

What it measures

NoveltyBench targets mode collapse: the tendency of a language model to give the same or barely- reworded answer every time it is sampled on a prompt that legitimately has many good answers (write a riddle, name a capital city in Africa, pick a favourite car). It generates k independent completions (10 by default) per prompt across 1,100 prompts designed or filtered to admit multiple valid, meaningfully different responses -- spanning randomness, underspecified factual recall, creative writing, and subjective opinion -- and asks whether the k outputs are actually distinct, useful alternatives rather than trivial rewordings of one underlying answer.

Task format

For each prompt, a model produces k independent samples. A fine-tuned classifier groups the k outputs into equivalence classes based on whether a user who saw one would gain anything from seeing another; a separate reward model scores each output's quality. Two metrics are reported per prompt and averaged across the dataset: distinct_k, the count of equivalence classes among the k samples, and utility_k, a patience-weighted cumulative score that credits only the first (highest-quality, earliest-generated) output in each class, discounted by a patience factor for later, still-novel classes.

Models reporting this benchmark

No model card in ModelSpec reports this benchmark yet.

Data

This page as JSON · Edit on GitHub