Metraipolis › archive › dc_ac97bbd18dfd

To Think or Not to Think: Allocating Reasoning Where It Helps — arXiv:2609.29664

Source: arXiv (en)
Original: https://arxiv.org/abs/2609.29664
License (as archived): CC0-1.0
Archived: 2026-10-08T04:26:18.462Z
SHA-256 of the archived text: 4bfb7c7397094500c9d4ddf71682ec5645da61bf68d5b70bb569b97169c544a9
Stable URL of this page: https://metraipolis.com/documento/dc_ac97bbd18dfd

How to cite

To Think or Not to Think: Allocating Reasoning Where It Helps — arXiv:2609.29664 — arXiv (en), https://arxiv.org/abs/2609.29664 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ac97bbd18dfd (Metraipolis archive dc_ac97bbd18dfd, sha256 4bfb7c7397094500c9d4ddf71682ec5645da61bf68d5b70bb569b97169c544a9)

BibTeX

@misc{dcac97bbd18dfd,
  title = {To Think or Not to Think: Allocating Reasoning Where It Helps — arXiv:2609.29664},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.29664},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ac97bbd18dfd; sha256 4bfb7c7397094500c9d4ddf71682ec5645da61bf68d5b70bb569b97169c544a9},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.29664. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Reinforcement learning (RL) has proven effective in enhancing the reasoning performance of large language models (LLMs), particularly in complex mathematical and programming tasks. However, this capability comes with systematic \textit{length misallocation}, in which models devote excessive reasoning to simple questions while terminating prematurely on harder ones, degrading inference efficiency with negligible accuracy improvement. Many length-adaptive methods mitigate this issue by allocating token budgets according to question difficulty, under the implicit assumption that harder questions benefit monotonically from extended reasoning. In contrast, we find that the effect of reasoning length on accuracy is concentrated on \textit{partially solvable} questions. Our further analysis reveals that explicit length rewards can produce unintended training dynamics. Motivated by these findings, we propose \textbf{CARE}---\textbf{C}ontrastive \textbf{A}ccuracy \textbf{R}eward \textbf{E}stimation---which compares the beneficial length adjustment per question from online sampled responses and applies adaptive length rewards within Group Relative Policy Optimization, with no extra hyperparameters or additional inference cost. Experiments across multiple reasoning benchmarks demonstrate that our method improves Pass@1 by up to \(4\%\) while simultaneously reducing reasoning length by \(37\%\), achieving higher token efficiency. Code will be available upon the acceptance of this paper.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-25 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.29664 — 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 2026-10-08T03:50:57.817Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-25?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).