Metraipolis › archive › dc_a7ef70ec367e

Expanding LLM Reasoning — arXiv:2610.05584

Source: arXiv
Original: https://arxiv.org/abs/2610.05584
License (as archived): CC0-1.0
Archived: 2026-10-06T05:27:01.662Z
SHA-256 of the archived text: 348c2832feefb7eae158f429c13ef20a060ad6ade5e5dfd46062c2770badc5ec
Stable URL of this page: https://metraipolis.com/documento/dc_a7ef70ec367e

How to cite

Expanding LLM Reasoning — arXiv:2610.05584 — arXiv, https://arxiv.org/abs/2610.05584 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_a7ef70ec367e (Metraipolis archive dc_a7ef70ec367e, sha256 348c2832feefb7eae158f429c13ef20a060ad6ade5e5dfd46062c2770badc5ec)

BibTeX

@misc{dca7ef70ec367e,
  title = {Expanding LLM Reasoning — arXiv:2610.05584},
  howpublished = {arXiv},
  url = {https://arxiv.org/abs/2610.05584},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_a7ef70ec367e; sha256 348c2832feefb7eae158f429c13ef20a060ad6ade5e5dfd46062c2770badc5ec},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv, https://arxiv.org/abs/2610.05584. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Extra inference compute is usually spent on sampling more reasoning chains. We study where inside an existing chain an additional continuation should begin. We define expansion utility, the change in correctness from restarting a chain at a stored step, and measure it at every eligible step for nine models on six benchmarks (41 model and benchmark cells). Restart position matters: steps selected on one set of continuations beat uniform placement when scored on disjoint ones, in held-out audits on 5, 16, and 38 cells (+4.25 points [+2.51, +6.63] in a fresh five-cell audit). A fixed rule that restarts from the last eligible steps, always-last, is a strong baseline: our learned router beats uniform placement but shows no detected gain over it, and on DeepSeek-R1-Distill-Qwen-14B/MATH-500 always-last exceeds the exact self-consistency frontier at matched aggregate generated output by +0.052 [+0.008, +0.098], using 0.774x the aggregate generated output of four-sample self-consistency. Cross-fitted oracle selection still finds held-out headroom beyond declared positional classes, a target for future selectors. Finally, breaking step-label ties by earliest index flips the sign of a pointwise selector's gain over uniform placement in every seed of a five-seed diagnostic with four rollouts per step; randomized ties remove the bias.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI); Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2610.05584 — 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 2026-10-06T05:24:06.048Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).