Small reward models trained on LitBench reach 78% agreement with upvote-derived human preferences in creative writing, beating all zero-shot LLM judges tested.
Michele Elam
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
LitBench: A Benchmark and Dataset for Reliable Evaluation of Creative Writing
Small reward models trained on LitBench reach 78% agreement with upvote-derived human preferences in creative writing, beating all zero-shot LLM judges tested.