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 2412.01113cs.CL
收藏

LLMs忠实且迭代地计算答案期间进行Co-T:多步算术中的系统分析

LLMs Faithfully and Iteratively Compute Answers During CoT: A Systematic Analysis With Multi-step Arithmetics

Tohoku University(东北大学);RIKEN(理化学研究所);MBZUAI

Keito Kudo, Yoichi Aoki, Tatsuki Kuribayashi, Shusaku Sone, Masaya Taniguchi, Ana Brassard, Keisuke Sakaguchi, Kentaro Inui

AI总结:

研究LLMs在Co-T推理中的内部信息流,发现其在生成推理链时逐步计算子答案,而非预先确定最终答案,从而验证了推理链的忠实性。

摘要:

本研究探讨了大型语言模型(LLMs)在进行链式思维(CoT)推理时的内部信息流。特别关注CoT解释对LLMs最终答案的忠实性,探索(i)LLMs的答案在CoT开始前或后是否已确定,以及(ii)CoT信息对最终答案的因果影响强度。通过受控的算术任务实验,揭示了LLMs存在系统性的内部推理机制。它们在输入进入模型时并未得出答案,而是边生成推理链边计算(子)答案。因此,生成的推理链可被视为模型内部计算的忠实反映。

英文摘要:

This study investigates the internal information flow of large language models (LLMs) while performing chain-of-thought (CoT) style reasoning. Specifically, with a particular interest in the faithfulness of the CoT explanation to LLMs' final answer, we explore (i) when the LLMs' answer is (pre)determined, especially before the CoT begins or after, and (ii) how strongly the information from CoT specifically has a causal effect on the final answer. Our experiments with controlled arithmetic tasks reveal a systematic internal reasoning mechanism of LLMs. They have not derived an answer at the moment when input was fed into the model. Instead, they compute (sub-)answers while generating the reasoning chain on the fly. Therefore, the generated reasoning chains can be regarded as faithful reflections of the model's internal computation.

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。

↑