Metraipolis › archive › dc_d0de73cd7b47

A Four-Stage Decomposition of Word-Problem Solving and Mechanistic Fragility in LLM Math Reasoning — arXiv:2609.17804

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.17804
License (as archived)
CC0-1.0
Archived
2026-10-15T15:05:10.951Z
SHA-256 of the archived text
953a2e0330cbb62a2ada662400556ccfbfcc0d8821a906b51584644a2a7e32d0
Stable URL of this page
https://metraipolis.com/documento/dc_d0de73cd7b47

How to cite

A Four-Stage Decomposition of Word-Problem Solving and Mechanistic Fragility in LLM Math Reasoning — arXiv:2609.17804 — arXiv (en), https://arxiv.org/abs/2609.17804 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d0de73cd7b47 (Metraipolis archive dc_d0de73cd7b47, sha256 953a2e0330cbb62a2ada662400556ccfbfcc0d8821a906b51584644a2a7e32d0)
BibTeX
@misc{dcd0de73cd7b47,
  title = {A Four-Stage Decomposition of Word-Problem Solving and Mechanistic Fragility in LLM Math Reasoning — arXiv:2609.17804},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.17804},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d0de73cd7b47; sha256 953a2e0330cbb62a2ada662400556ccfbfcc0d8821a906b51584644a2a7e32d0},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.17804. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Large language models solve grade-school math word problems with high accuracy, yet a single irrelevant clause inserted into the problem can collapse it. We reconcile these observations with a mechanistic account. We show that the model's internal computation decomposes into a four-stage sequential pipeline, Schema Abstraction, Operation Planning, Operand Binding, and Computation, each stage producing a distinct intermediate representation in an identifiable band of layers. Using the same scaffold to diagnose distractor-induced failure, we localize the corruption to a single stage, Operation Planning, implemented by a set of attention heads whose causal role we validate bidirectionally. In short, we provide a mechanistic interpretation of math word problem reasoning in LLMs, and their failure when distracted.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-17 in cs.LG).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.17804 — 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 2026-10-15T14:08:19.208Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-17?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).