arXiv：cs.CL· Chao Wan, Albert Gong, Mihir Mishra, Carl-Leander Henneking, Claas Beger, Kilian Q. Weinberger·· 16 小时前
π-CoT：用 Prolog 初始化链式推理，提升多跳问答表现
$\pi$-CoT: Prolog-Initialized Chain-of-Thought Prompting for Multi-Hop Question-Answering
AI 导读
研究提出 Prolog-Initialized Chain-of-Thought（π-CoT）提示策略，将多跳问题改写为 Prolog 查询并拆解为单跳子查询，依次求解后用中间结果初始化后续 CoT 推理。在多跳问答基准上，π-CoT 持平或优于标准 RAG 与上下文 CoT，在检索能提供必要证据的问题上增益最大。
正文
Abstract:Chain-of-Thought (CoT) prompting significantly enhances large language models' (LLMs) problem-solving capabilities, but still struggles with complex multi-hop questions, often falling into circular reasoning patterns or deviating from the logical path entirely. This limitation is particularly acute in retrieval-augmented generation (RAG) settings, where obtaining the right context is critical. We introduce Prolog-Initialized Chain-of-Thought ($\pi$-CoT), a novel prompting strategy that combines logic programming's structural rigor with language models' flexibility. $\pi$-CoT reformulates multi-hop questions into Prolog queries decomposed as single-hop sub-queries. These are resolved sequentially, producing intermediate artifacts, with which we initialize the subsequent CoT reasoning procedure. Extensive experiments demonstrate that $\pi$-CoT matches or outperforms standard RAG and in-context CoT on multi-hop question-answering benchmarks, with the largest gains on questions for which retrieval surfaces the necessary evidence.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2506.20642 [cs.CL] |
| (or arXiv:2506.20642v3 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2506.20642 arXiv-issued DOI via DataCite |
Submission history
From: Albert Gong [view email]
[v1]
Wed, 25 Jun 2025 17:37:59 UTC (1,749 KB)
[v2]
Sun, 15 Feb 2026 19:49:34 UTC (600 KB)
[v3]
Wed, 14 Oct 2026 02:34:19 UTC (604 KB)
来源：arXiv：cs.CL · arxiv.org