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 2402.02648cs.CLcs.AI
收藏

递归反馈链防止冗余提示导致的性能下降

Recursive Chain-of-Feedback Prevents Performance Degradation from Redundant Prompting

University of California, Berkeley（加州大学伯克利分校）

Jinwoo Ahn, Kyuseung Shin

备注：
Still Ongoing Work; 8 Pages; 2 Figures
阅读 PDF
arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文研究重复无意义提示导致 LLM 推理质量下降的问题，提出 R-CoF，通过递归分解错误步骤并修正初始回答，使多数失败问题无需逻辑示例即可被正确解答。

摘要：

大型语言模型（LLM）在复杂推理任务中经常遇到困难，无法构建通向解决方案的逻辑合理步骤。针对这种行为，用户往往会反复提示 LLM，希望获得更好的回答。本文通过定义一种新颖设定——反馈链（Chain-of-Feedback，CoF），研究这种重复行为及其影响。该设定将需要多步推理的问题作为输入。在得到回答后，我们反复给出无意义反馈（例如“再试一次”），要求进行额外尝试。令人惊讶的是，我们的初步结果表明，重复的无意义反馈会逐渐降低回答质量，最终导致与预期结果的偏差更大。为缓解这些问题，我们提出一种新方法——递归反馈链（Recursive Chain-of-Feedback，R-CoF）。遵循计算机科学中的递归逻辑，R-CoF 通过将每个错误推理步骤分解为更小的独立问题，递归修正最初错误的回答。初步结果表明，在没有任何概述逻辑过程的样本数据的情况下，使用 R-CoF 可以回答大多数 LLM 未能正确回答的问题。

英文摘要：

Large Language Models (LLMs) frequently struggle with complex reasoning tasks, failing to construct logically sound steps towards the solution. In response to this behavior, users often try prompting the LLMs repeatedly in hopes of reaching a better response. This paper studies such repetitive behavior and its effect by defining a novel setting, Chain-of-Feedback (CoF). The setting takes questions that require multi-step reasoning as an input. Upon response, we repetitively prompt meaningless feedback (e.g. 'make another attempt') requesting additional trials. Surprisingly, our preliminary results show that repeated meaningless feedback gradually decreases the quality of the responses, eventually leading to a larger deviation from the intended outcome. To alleviate these troubles, we propose a novel method, Recursive Chain-of-Feedback (R-CoF). Following the logic of recursion in computer science, R-CoF recursively revises the initially incorrect response by breaking down each incorrect reasoning step into smaller individual problems. Our preliminary results show that majority of questions that LLMs fail to respond correctly can be answered using R-CoF without any sample data outlining the logical process.

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。

↑