arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2410.04838cs.CL
收藏

通过成对自评估实现具备理由感知的答案验证

Rationale-Aware Answer Verification by Pairwise Self-Evaluation

The Asahi Shimbun Company(朝日新闻社);National Institute of Informatics(信息学研究所)

Akira Kawabata, Saku Sugawara

AI总结:

针对答案正确但理由错误导致验证器不可靠的问题,提出 REPS 成对自评估方法筛选有效理由,并在三个推理基准上提升验证器性能。

摘要:

答案验证旨在从大型语言模型(LLM)生成的候选解答中识别正确解答。当前方法通常仅根据最终答案是否与标准答案一致,将解答标记为正确或错误,以此训练验证器模型。然而,这种做法忽视了那些得出正确答案的解答中可能存在的有缺陷理由,从而削弱验证器区分合理理由与有缺陷理由的能力。我们通过实验表明,在 StrategyQA 中,答案正确的 LLM 生成解答里只有 19% 具有有效理由,因此会导致验证器不可靠。此外,我们证明,基于有效理由训练验证器可显著提升其区分有效理由与有缺陷理由的能力。为了在无需额外人工监督的情况下构建更好的验证器,我们提出 REPS(Rationale Enhancement through Pairwise Selection,即通过成对选择增强理由),该方法通过迭代使用生成解答的同一 LLM 进行成对自评估,从候选解答中选择有效理由。在 ARC-Challenge、DROP 和 StrategyQA 三个推理基准上,使用 REPS 选出的解答训练出的验证器优于采用传统训练方法训练的验证器。我们的结果表明,训练可靠验证器除了需要保证最终答案正确之外,还必须确保理由有效,这对于辅助人类完成复杂推理任务的模型至关重要。

英文摘要:

Answer verification identifies correct solutions among candidates generated by large language models (LLMs). Current approaches typically train verifier models by labeling solutions as correct or incorrect based solely on whether the final answer matches the gold answer. However, this approach neglects any flawed rationale in the solution yielding the correct answer, undermining the verifier's ability to distinguish between sound and flawed rationales. We empirically show that in StrategyQA, only 19% of LLM-generated solutions with correct answers have valid rationales, thus leading to an unreliable verifier. Furthermore, we demonstrate that training a verifier on valid rationales significantly improves its ability to distinguish valid and flawed rationale. To make a better verifier without extra human supervision, we introduce REPS (Rationale Enhancement through Pairwise Selection), a method for selecting valid rationales from candidates by iteratively applying pairwise self-evaluation using the same LLM that generates the solutions. Verifiers trained on solutions selected by REPS outperform those trained using conventional training methods on three reasoning benchmarks (ARC-Challenge, DROP, and StrategyQA). Our results suggest that training reliable verifiers requires ensuring the validity of rationales in addition to the correctness of the final answers, which would be critical for models assisting humans in solving complex reasoning tasks.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑