Metraipolis › archive › dc_42566b73237f

"Close enough?" An investigation into multiplication errors in LLMs — arXiv:2610.38634

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.38634
License (as archived)
CC0-1.0
Archived
2026-10-31T06:09:01.760Z
SHA-256 of the archived text
02414fc5262705f638d19867d0c972a166cd6fa5a8948347ccfe65e20a9cab46
Stable URL of this page
https://metraipolis.com/documento/dc_42566b73237f

How to cite

"Close enough?" An investigation into multiplication errors in LLMs — arXiv:2610.38634 — arXiv (en), https://arxiv.org/abs/2610.38634 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_42566b73237f (Metraipolis archive dc_42566b73237f, sha256 02414fc5262705f638d19867d0c972a166cd6fa5a8948347ccfe65e20a9cab46)
BibTeX
@misc{dc42566b73237f,
  title = {"Close enough?" An investigation into multiplication errors in LLMs — arXiv:2610.38634},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.38634},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_42566b73237f; sha256 02414fc5262705f638d19867d0c972a166cd6fa5a8948347ccfe65e20a9cab46},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.38634. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Large language models (LLMs) frequently fail at multi-digit arithmetic, but the underlying mechanisms governing these failures remain poorly understood. While recent work demonstrates that addition errors follow a "noisy quantisation" model localised around carry boundaries, whether this framework generalises to multiplication remains an open question. In this work, we evaluate multiplication capabilities across several small base LLMs on a dataset of 16,991 problems spanning varied operand lengths. We show that multiplication errors do not conform to the addition-style noisy quantisation model. Instead, we uncover a novel error regime in small Qwen models: errors in lower-order columns predominantly preserve parity. Conversely, higher-order digit errors shift toward off-by-one carry failures. Our findings suggest that arithmetic computation in LLMs relies on distinct, multi-periodic internal representations where parity is tracked more robustly than exact digit identity, providing new mechanistic insights into numerical reasoning in transformers.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2610.38634 — 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 2026-10-31T06:07:08.487Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).