Metraipolis › archive › dc_bad40f39bbb2
Test-Time Scaling in Reasoning LLMs: Inference Regimes, Evaluation, and Reproducibility — arXiv:2608.04001
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.04001
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-31T12:19:44.106Z
- SHA-256 of the archived text
7c5548e20310a2cbe11c7aa50a749d469cb2491ff46654df9323aebfada11aa5- Stable URL of this page
https://metraipolis.com/documento/dc_bad40f39bbb2
How to cite
Test-Time Scaling in Reasoning LLMs: Inference Regimes, Evaluation, and Reproducibility — arXiv:2608.04001 — arXiv (en), https://arxiv.org/abs/2608.04001 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_bad40f39bbb2 (Metraipolis archive dc_bad40f39bbb2, sha256 7c5548e20310a2cbe11c7aa50a749d469cb2491ff46654df9323aebfada11aa5)
BibTeX
@misc{dcbad40f39bbb2,
title = {Test-Time Scaling in Reasoning LLMs: Inference Regimes, Evaluation, and Reproducibility — arXiv:2608.04001},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.04001},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_bad40f39bbb2; sha256 7c5548e20310a2cbe11c7aa50a749d469cb2491ff46654df9323aebfada11aa5},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.04001. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Large language models can solve harder reasoning problems with more inference-time compute. The term "test-time scaling," however, covers several inference algorithms: extending deliberation along one trajectory, sampling completed candidates and aggregating them by voting or verification, and searching over partial states. These algorithms differ in statistical structure, compute requirements, and failure modes. Treating them as interchangeable under a scalar "budget," or reporting accuracy without specifying the inference protocol, makes results difficult to compare across studies. We study test-time scaling along three axes. First, we formalize it as budgeted inference over the implicit prefix tree of an autoregressive model and distinguish single-trajectory sequential scaling, leaf-level scaling with terminal reduction, and prefix-level scaling. Second, we treat the full inference system as the evaluated object and separate end-to-end performance from candidate-bank diagnostics. We introduce an evaluation profile whose coordinates and simple functionals recover or bound common repeated-sampling metrics, and require compute accounting and uncertainty estimates that match the protocol. Third, we distinguish exact replay from distributional reproducibility and state the requirements for each. We also organize open-weight reasoning models by model-side and interface mechanisms. Our empirical study covers broad knowledge, symbolic reasoning, and competition mathematics, and we publicly release 1,403,520 sampled model attempts. The project website is available at this https URL . The released datasets are Trace ( this https URL ), Lite ( this https URL ), Math ( this https URL ), and SuperGPQA ( this https URL ). Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-05 in cs.AI). Categorie arXiv: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2608.04001 — 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 2026-10-31T12:05:53.108Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-08-05?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).