Beneath the Surface of Chains-of-Thought: A Mechanistic Interpretation of Reasoning Operations in LLMs
- 类型:arxiv
- 标识:2609.04753
- 链接:https://arxiv.org/abs/2609.04753
- 主分类:engineering
- 形态:position
- TLDR:Reasoning in large language models unfolds through diverse functional operations, such as problem formulation, goal decomposition, and deduction. Although these operations are explicitly distinguished in text, little is known about how they are geometrically organized in representation spaces. To this end, we investigate whether distinct reasoning operations exhibit corresponding geometric structure in hidden representations. We find that operations are separable in held-out representations, with separability peaking in middle layers, and verify that this structure is not explained by lexical
- 待LLM分类:是
- 来源文件:
- /inbox/tom/_candidates/2026-09-07-agent-rag-longcontext-candidates.json