arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2305.00633cs.CLcs.AIcs.LG
收藏

自评估引导的束搜索用于推理

Self-Evaluation Guided Beam Search for Reasoning

National University of Singapore(新加坡国立大学);The Hong Kong University of Science and Technology(香港科技大学)

Yuxi Xie, Kenji Kawaguchi, Yiran Zhao, Xu Zhao, Min-Yen Kan, Junxian He, Qizhe Xie

AI总结:

本文提出一种自评估引导的随机束搜索解码方法,通过逐步自评估校准LLM多步推理过程,在多个推理基准上显著超越基线并提升一致性与鲁棒性。

摘要:

将问题分解为中间步骤已在大语言模型(LLM)推理中展现出显著性能。然而,推理链的增长会引入不确定性并导致误差累积,使得获取准确的最终结果变得困难。为应对多步推理中的这一不确定性挑战,我们引入了一种逐步自评估机制来引导和校准LLM的推理过程。我们提出了一种解码算法,通过随机束搜索整合自评估引导。自评估引导作为一种校准更优的自动准则,有助于在推理空间中进行高效搜索,并带来更优的预测质量。随机束搜索通过温度控制的随机性平衡搜索空间的利用与探索。我们的方法在GSM8K、AQuA和StrategyQA基准上的少样本准确率分别比相应的Codex骨干基线高出6.34%、9.56%和5.46%。在算术推理任务上使用Llama-2的实验结果表明,我们的方法在计算预算相当的情况下优于基线方法,体现了其高效性。对多步推理的进一步分析发现,我们的自评估引导能够准确定位逻辑失败,并带来更高的一致性和鲁棒性。我们的代码已在 https://guideddecoding.github.io/ 公开。

英文摘要:

Breaking down a problem into intermediate steps has demonstrated impressive performance in Large Language Model (LLM) reasoning. However, the growth of the reasoning chain introduces uncertainty and error accumulation, making it challenging to elicit accurate final results. To tackle this challenge of uncertainty in multi-step reasoning, we introduce a stepwise self-evaluation mechanism to guide and calibrate the reasoning process of LLMs. We propose a decoding algorithm integrating the self-evaluation guidance via stochastic beam search. The self-evaluation guidance serves as a better-calibrated automatic criterion, facilitating an efficient search in the reasoning space and resulting in superior prediction quality. Stochastic beam search balances exploitation and exploration of the search space with temperature-controlled randomness. Our approach surpasses the corresponding Codex-backboned baselines in few-shot accuracy by $6.34\%$, $9.56\%$, and $5.46\%$ on the GSM8K, AQuA, and StrategyQA benchmarks, respectively. Experiment results with Llama-2 on arithmetic reasoning demonstrate the efficiency of our method in outperforming the baseline methods with comparable computational budgets. Further analysis in multi-step reasoning finds our self-evaluation guidance pinpoints logic failures and leads to higher consistency and robustness. Our code is publicly available at https://guideddecoding.github.io/.

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。

↑