Metraipolis › archive › dc_cfec3546f96f

Beyond Repeated Sampling: Learning Search Policies for LLM Reasoning — arXiv:2609.26704

Source: arXiv (en)
Original: https://arxiv.org/abs/2609.26704
License (as archived): CC0-1.0
Archived: 2026-10-08T13:47:26.752Z
SHA-256 of the archived text: 135cb455a2edaf6518085e3c54e17c9e440bb705cb7790657dd0691d1e13cae7
Stable URL of this page: https://metraipolis.com/documento/dc_cfec3546f96f

How to cite

Beyond Repeated Sampling: Learning Search Policies for LLM Reasoning — arXiv:2609.26704 — arXiv (en), https://arxiv.org/abs/2609.26704 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_cfec3546f96f (Metraipolis archive dc_cfec3546f96f, sha256 135cb455a2edaf6518085e3c54e17c9e440bb705cb7790657dd0691d1e13cae7)

BibTeX

@misc{dccfec3546f96f,
  title = {Beyond Repeated Sampling: Learning Search Policies for LLM Reasoning — arXiv:2609.26704},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.26704},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_cfec3546f96f; sha256 135cb455a2edaf6518085e3c54e17c9e440bb705cb7790657dd0691d1e13cae7},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.26704. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Large language models increasingly tackle hard reasoning problems by spending more test-time compute, yet the dominant strategy remains naive repeated sampling: draw many independent solutions and hope one is correct. Because such sampling explores only through local decoding noise, it tends to produce many near duplicate attempts rather than genuinely different ideas. We ask whether exploration can instead be steered at a semantic level, by first sampling problem specific concepts, hints, or strategies and then conditioning answer generation on them. We refine this into a simple, more exploratory procedure that emits many diverse concepts in a single trajectory, and evaluate it on hard problems where repeated sampling struggles. We then go a step further and make concept generation trainable: a small concept generator is optimized with reinforcement learning so that its concepts maximize the downstream success of a larger, frozen answer generator. On hard mathematical reasoning problems, the trained concept generator substantially improves the answer generator's pass@k over naive repeated sampling at the same answer generation allocation, surpasses concepts drawn from much larger untuned models, and transfers to answer generators it was never trained against, including a model from a different family. A small model can thus be trained into an effective, reusable search policy for a much larger one.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-23 in cs.AI).
Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.26704 — 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 2026-10-08T13:18:38.337Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-23?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).