Metraipolis › archive › dc_4bf3932fd90a
Beneath the Surface of Chains-of-Thought: A Mechanistic Interpretation of Reasoning Operations in LLMs — arXiv:2609.04753
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.04753
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-16T22:35:10.687Z
- SHA-256 of the archived text
dd01159fbb1624f43912d9819d190e0660c4c0e3b32762aa11d2e79d15188cb6- Stable URL of this page
https://metraipolis.com/documento/dc_4bf3932fd90a
How to cite
Beneath the Surface of Chains-of-Thought: A Mechanistic Interpretation of Reasoning Operations in LLMs — arXiv:2609.04753 — arXiv (en), https://arxiv.org/abs/2609.04753 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_4bf3932fd90a (Metraipolis archive dc_4bf3932fd90a, sha256 dd01159fbb1624f43912d9819d190e0660c4c0e3b32762aa11d2e79d15188cb6)
BibTeX
@misc{dc4bf3932fd90a,
title = {Beneath the Surface of Chains-of-Thought: A Mechanistic Interpretation of Reasoning Operations in LLMs — arXiv:2609.04753},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.04753},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_4bf3932fd90a; sha256 dd01159fbb1624f43912d9819d190e0660c4c0e3b32762aa11d2e79d15188cb6},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.04753. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
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 or positional confounds. Across layers, token-wise operation-alignment becomes more distributed over spans, while identical surface tokens are represented differently depending on the operation of its surrounding chunk. Attention-masking interventions further show that operation-aligned representations at chunk onset depend on preceding reasoning context. Consequently, our work demonstrates that language models maintain representational correspondence between linguistic reasoning expressions and their internal geometric structures. Code and project materials are available at this https URL . Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-07 in cs.CL). Categorie arXiv: Computation and Language (cs.CL). Pagina del lavoro: https://arxiv.org/abs/2609.04753 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-16T22:21:16.280Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-07?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).