arXiv cs.AI· Zach Studdiford, Gary Lupyan·· 10 天前AI 评分36
推理即模式匹配：人类与 LLM 日常推理的共享机制
Reasoning as Pattern Matching: Shared Mechanisms in Human and LLM Everyday Reasoning
AI 导读
一项 arXiv 研究对比 46 个 LLM 与两组人类参与者的常识推理表现，发现两者在多种日常情境中呈现趋同的推理模式。通过分析内容不变与内容敏感的模型神经元，研究发现真正让模型与人类对齐的是内容敏感机制，而非内容不变的世界模型。这表明人类与 LLM 的日常因果推理都大量依赖模式匹配。
正文 · 原文
Abstract:When large language models (LLMs) fail to generalize or make content-sensitive errors in reasoning, it is often taken as evidence that LLMs are not truly reasoning, but rather performing a kind of pattern matching. The implication is that human behavior does not exhibit the same types of failures because human reasoning relies on principled and content-invariant world models. We test this assumption by first evaluating humans and LLMs on their ability to engage in common-sense reasoning about a variety of everyday situations. Our results reveal convergent patterns of reasoning across 46 LLMs and two cohorts of human participants. We then ask whether this behavioral convergence is due to LLMs having acquired content-invariant world models or a set of pattern-matching heuristics by characterizing the roles of content-invariant and content-sensitive model neurons in producing human-like responses. We find that while LLMs encode both content-invariant and content-sensitive representations, it is content-sensitive mechanisms which are causally responsible for aligning models with humans. Taken together, our results suggest that everyday causal reasoning in people and LLMs makes heavy use of pattern-matching.
| Comments: | 13 pages main text, 59 pages supplementary text |
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2606.13607 [cs.AI] |
| (or arXiv:2606.13607v4 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2606.13607 arXiv-issued DOI via DataCite |
Submission history
From: Zach Studdiford [view email]
[v1]
Thu, 11 Jun 2026 17:23:10 UTC (5,525 KB)
[v2]
Sat, 13 Jun 2026 20:58:19 UTC (5,526 KB)
[v3]
Thu, 1 Oct 2026 02:06:49 UTC (3,870 KB)
[v4]
Mon, 5 Oct 2026 23:56:28 UTC (3,858 KB)
来源：arXiv cs.AI · arxiv.org