Metraipolis › archive › dc_73e7fc52163f

Depth as a Candidate Source: Portfolio Allocation for Recurrent Test-Time Scaling in Large Language Models — arXiv:2611.12293

Source
arXiv (en)
Original
https://arxiv.org/abs/2611.12293
License (as archived)
CC0-1.0
Archived
2026-11-10T18:41:49.012Z
SHA-256 of the archived text
252849549f037f8d8333101f36b543b11871e7837cb0f3b63797fac8b37ebe40
Stable URL of this page
https://metraipolis.com/documento/dc_73e7fc52163f

How to cite

Depth as a Candidate Source: Portfolio Allocation for Recurrent Test-Time Scaling in Large Language Models — arXiv:2611.12293 — arXiv (en), https://arxiv.org/abs/2611.12293 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_73e7fc52163f (Metraipolis archive dc_73e7fc52163f, sha256 252849549f037f8d8333101f36b543b11871e7837cb0f3b63797fac8b37ebe40)
BibTeX
@misc{dc73e7fc52163f,
  title = {Depth as a Candidate Source: Portfolio Allocation for Recurrent Test-Time Scaling in Large Language Models — arXiv:2611.12293},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.12293},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_73e7fc52163f; sha256 252849549f037f8d8333101f36b543b11871e7837cb0f3b63797fac8b37ebe40},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.12293. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Recurrent latent reasoning in large lanugage models (LLMs) scales test-time computation by increasing recurrent depth, naturally framing inference as a problem of deciding when to stop. We identify a previously overlooked limitation of this single-depth view through systematic preliminary analysis: deeper recurrence can lose correct solutions found at shallower depths, while depth-specific success sets remain broadly non-nested even after controlling for marginal accuracy. This suggests that depth determines not only computation cost, but also access to distinct candidate solutions. Motivated by this finding, we propose Depth Portfolio Allocation (DPA), which reformulates recurrent test-time scaling as budgeted allocation across depth-specific candidate sources. DPA supports both fixed portfolios and feedback-guided sequential allocation under a shared inference budget. Experiments on GSM8K, SVAMP, MultiArith, and MBPP show that mixed-depth portfolios expand oracle coverage under matched candidate counts and nominal costs, while feedback-guided allocation generally matches or improves uniform allocation and can reduce computation in high-budget settings. Code and data are avaible at : this https URL .

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/2611.12293 — 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 2026-11-10T18:37:37.545Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).