arXiv cs.AI· Jeong-Yoon Kim, Tae-Yoon Lee·· 12 小时前AI 评分20
Adaptive Semantic Branching（ASB）：利用跨轨迹分歧再生成扩展推理候选
Adaptive Semantic Branching: Disagreement-Localized Cross-Trajectory Regeneration for Reasoning Candidate Expansion
AI 导读
研究者提出免训练的推理流程 Adaptive Semantic Branching（ASB），将同一问题的多条推理轨迹按最终答案分组，用单调动态规划对齐推理步骤，并借持续分歧目标定位路径分叉点，再以竞争轨迹的推理窗口为条件重新生成候选。
正文 · 原文
Abstract:Reasoning trajectories generated for the same problem often arrive at conflicting answers. These disagreements reveal alternative intermediate decisions that can be reused to explore new solutions. We introduce Adaptive Semantic Branching (ASB), a training-free inference procedure that turns cross-trajectory disagreement into targeted candidate generation. ASB first groups independently sampled trajectories by extracted final answer, selects a representative from each group, and aligns their reasoning steps with a monotonic dynamic program. A persistent-divergence objective identifies the point at which initially similar reasoning paths develop different continuations. For each selected directed pair, ASB retains the source trajectory's textual prefix, extracts a short reasoning window from a competing donor trajectory, and generates a fresh solution conditioned on both. A diversity-aware policy allocates up to three additional generations according to the answer-cluster structure. In a 500-problem MATH evaluation with Qwen2.5-7B-Instruct, ASB increases oracle candidate coverage from 445/500 (89.0%) with seven initial samples to 455/500 (91.0%), recovering correct candidates for 10 of the 55 initially uncovered problems. The method uses 576 additional generation calls across the evaluation set, averaging 8.152 calls per problem. We provide a precise specification of the branching operator, alignment objective, generation policy, and experimental configuration.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.07918 [cs.AI] |
| (or arXiv:2611.07918v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2611.07918 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Jeong-Yoon Kim [view email]
[v1]
Thu, 5 Nov 2026 12:29:57 UTC (13 KB)
来源：arXiv cs.AI · arxiv.org