LMUnit: Fine-grained Evaluation with Natural Language Unit Tests

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Saad-Falcon, Jon, Vivek, Rajan, Berrios, William, Naik, Nandita Shankar, Franklin, Matija, Vidgen, Bertie, Singh, Amanpreet, Kiela, Douwe, Mehri, Shikib
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866915833376145408
author Saad-Falcon, Jon
Vivek, Rajan
Berrios, William
Naik, Nandita Shankar
Franklin, Matija
Vidgen, Bertie
Singh, Amanpreet
Kiela, Douwe
Mehri, Shikib
author_facet Saad-Falcon, Jon
Vivek, Rajan
Berrios, William
Naik, Nandita Shankar
Franklin, Matija
Vidgen, Bertie
Singh, Amanpreet
Kiela, Douwe
Mehri, Shikib
contents As language models become integral to critical workflows, assessing their behavior remains a fundamental challenge -- human evaluation is costly and noisy, while automated metrics provide only coarse, difficult-to-interpret signals. We introduce natural language unit tests, a paradigm that decomposes response quality into explicit, testable criteria, along with a unified scoring model, LMUnit, which combines multi-objective training across preferences, direct ratings, and natural language rationales. Through controlled human studies, we show this paradigm significantly improves inter-annotator agreement and enables more effective LLM development workflows. LMUnit achieves state-of-the-art performance on evaluation benchmarks (FLASK, BigGenBench) and competitive results on RewardBench. These results validate both our proposed paradigm and scoring model, suggesting a promising path forward for language model evaluation and development.
format Preprint
id arxiv_https___arxiv_org_abs_2412_13091
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle LMUnit: Fine-grained Evaluation with Natural Language Unit Tests
Saad-Falcon, Jon
Vivek, Rajan
Berrios, William
Naik, Nandita Shankar
Franklin, Matija
Vidgen, Bertie
Singh, Amanpreet
Kiela, Douwe
Mehri, Shikib
Computation and Language
Artificial Intelligence
As language models become integral to critical workflows, assessing their behavior remains a fundamental challenge -- human evaluation is costly and noisy, while automated metrics provide only coarse, difficult-to-interpret signals. We introduce natural language unit tests, a paradigm that decomposes response quality into explicit, testable criteria, along with a unified scoring model, LMUnit, which combines multi-objective training across preferences, direct ratings, and natural language rationales. Through controlled human studies, we show this paradigm significantly improves inter-annotator agreement and enables more effective LLM development workflows. LMUnit achieves state-of-the-art performance on evaluation benchmarks (FLASK, BigGenBench) and competitive results on RewardBench. These results validate both our proposed paradigm and scoring model, suggesting a promising path forward for language model evaluation and development.
title LMUnit: Fine-grained Evaluation with Natural Language Unit Tests
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2412.13091