arXiv cs.AI· Tingxin Yang, Zefeng Wang, Mengyue Wang, Jinhe Bi, Xingcheng Zhou, Volker Tresp, Yunpu Ma·· 13 小时前AI 评分31
DeepLook：用不确定性引导分支与前瞻置信度实现高效推理
DeepLook: Uncertainty-Guided Branching with Lookahead Confidence for Efficient Reasoning
AI 导读
DeepLook 是一个免训练解码框架，通过监控片段级置信度、仅在不确定性瓶颈处分支，并用 Average Lookahead Confidence（ALC）对候选续写排序后再投票。
正文 · 原文
Abstract:Test-time scaling improves large language model reasoning, but spends compute uniformly along a reasoning trace, even in stretches that are already stable. Motivated by the observation that reasoning failures show an early onset of uncertainty, we introduce DeepLook, a training-free decoding framework that monitors segment-level confidence, branches only at uncertainty bottlenecks, and ranks candidate continuations by Average Lookahead Confidence (ALC)---their mean confidence over fixed-horizon rollouts---before completing and voting over a pruned subset. On four competition mathematics benchmarks across four open models, DeepLook matches or exceeds the accuracy of DeepConf-low in 12 of 16 settings while generating 87.3% fewer tokens, showing that a model's own confidence is informative enough to decide where extra decoding compute is worth spending. On knowledge-intensive GPQA-Diamond the accuracy advantage disappears even though the token saving does not: coherent hallucinations also receive high ALC, so internal confidence localizes instability, not falsehood. Code is available at this https URL.
| Comments: | EMNLP 2026 Workshop UncertaiNLP |
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2607.22602 [cs.AI] |
| (or arXiv:2607.22602v2 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2607.22602 arXiv-issued DOI via DataCite |
Submission history
From: Zefeng Wang [view email]
[v1]
Fri, 12 Jun 2026 16:51:29 UTC (451 KB)
[v2]
Wed, 21 Oct 2026 06:46:53 UTC (461 KB)
来源：arXiv cs.AI · arxiv.org