Metraipolis › archive › dc_15cd50536e06

AutoCRAT: Within-trajectory Joint Control of Stochasticity and Compute for LLM Reasoning — arXiv:2608.29988

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.29988
License (as archived)
CC0-1.0
Archived
2026-10-17T07:44:27.488Z
SHA-256 of the archived text
a42a067b8dea086348a9107f8d8db41c8b6e657c1beac46f98168f3dd08de746
Stable URL of this page
https://metraipolis.com/documento/dc_15cd50536e06

How to cite

AutoCRAT: Within-trajectory Joint Control of Stochasticity and Compute for LLM Reasoning — arXiv:2608.29988 — arXiv (en), https://arxiv.org/abs/2608.29988 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_15cd50536e06 (Metraipolis archive dc_15cd50536e06, sha256 a42a067b8dea086348a9107f8d8db41c8b6e657c1beac46f98168f3dd08de746)
BibTeX
@misc{dc15cd50536e06,
  title = {AutoCRAT: Within-trajectory Joint Control of Stochasticity and Compute for LLM Reasoning — arXiv:2608.29988},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.29988},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_15cd50536e06; sha256 a42a067b8dea086348a9107f8d8db41c8b6e657c1beac46f98168f3dd08de746},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.29988. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Large language models (LLMs) achieve strong reasoning performance, which depends critically on inference-time decisions. Yet these decisions are commonly handled by static, one-size-fits-all policies, limiting adaptation to diverse tasks and reasoning stages. Recent adaptive methods partially address this limitation, but they primarily adapt either decoding stochasticity (how the model explores) or reasoning compute (how long the model reasons) in isolation, leaving their interaction within a single reasoning trajectory unmodeled. To address this challenge, we shift toward a within-trajectory joint control view, and instantiate it in AutoCRAT, a decoder-side controller for frozen backbones. Using only signals available during decoding, AutoCRAT jointly adjusts sampling stochasticity and reasoning budget during generation. AutoCRAT operates over a discrete action space and updates control decisions only at semantic boundaries, improving stability while remaining responsive to the evolving reasoning process. Comprehensive evaluation across 6 benchmarks demonstrates that AutoCRAT (I) uses 13.8-52.7% fewer inference tokens on average than recommended static configurations, (II) surpasses recommended static and adaptive baselines by 1.5-4.5% in relative accuracy, and (III) enjoys strong cross-backbone transferability.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-01 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2608.29988 — 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 2026-10-17T07:28:26.573Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).