Metraipolis › archive › dc_ea65c13255a5

Test-Time Scaling for Open-Ended Reasoning at Small Model Scale — arXiv:2611.02208

Source
arXiv (en)
Original
https://arxiv.org/abs/2611.02208
License (as archived)
CC0-1.0
Archived
2026-11-03T18:19:44.009Z
SHA-256 of the archived text
cff030c4733eef5962f57cb1b1a91cd90013db1c7c74906c3b4f6bdf1a13822c
Stable URL of this page
https://metraipolis.com/documento/dc_ea65c13255a5

How to cite

Test-Time Scaling for Open-Ended Reasoning at Small Model Scale — arXiv:2611.02208 — arXiv (en), https://arxiv.org/abs/2611.02208 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ea65c13255a5 (Metraipolis archive dc_ea65c13255a5, sha256 cff030c4733eef5962f57cb1b1a91cd90013db1c7c74906c3b4f6bdf1a13822c)
BibTeX
@misc{dcea65c13255a5,
  title = {Test-Time Scaling for Open-Ended Reasoning at Small Model Scale — arXiv:2611.02208},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.02208},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ea65c13255a5; sha256 cff030c4733eef5962f57cb1b1a91cd90013db1c7c74906c3b4f6bdf1a13822c},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.02208. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

A vote over a model's samples can only help when the samples disagree. For open-ended generation the binding question is therefore not how to combine samples but whether a model produces samples worth combining, a question test-time scaling has answered for math but not for free-form text. We study it at small scale with semantic cluster voting, a judge-free aggregator that groups a small instruct model's samples by meaning and returns the largest cluster's representative, with no reward model or self-judging. On a 250-prompt benchmark with three tiers of answer checkability, the same pattern holds across three model families (Qwen2.5-3B, Llama-3.2-3B, Gemma-2-2B) and two generation seeds: voting raises accuracy on the math control, where samples vary but stay checkable (a gain of +10 to +29 points across the three models), yet adds at most +0.30 judge points on open-ended text. On a larger 200-problem MATH-500 run with the primary model, the math gain is +8.5 points (95% CI [+4.0, +13.0]). Almost every open-ended sample falls into one meaning cluster, so there is little to select over. The collapse is consistent with instruction-tuning rather than model size, and raising decoding temperature does not recover usable diversity, because it costs more quality than it adds (net -0.64 judge points, 95% CI [-0.93, -0.37]).

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/2611.02208 — 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 2026-11-03T18:09:53.664Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).