Back to AI News
arxiv

Adaptive Semantic Branching: Disagreement-Localized Cross-Trajectory Regeneration for Reasoning Candidate Expansion

Adaptive Semantic Branching (ASB) is a training-free inference procedure that leverages disagreements among reasoning trajectories to generate new candidate solutions. It groups trajectories by final answer, aligns steps via dynamic programming, and identifies divergence points to splice prefixes from one trajectory with reasoning windows from another. On a 500-problem MATH evaluation with Qwen2.5-7B-Instruct, ASB increased oracle coverage from 89.0% to 91.0%, recovering correct candidates for 10 previously uncovered problems using 576 additional generation calls.

AI Brief

Introduces a practical, training-free method to improve LLM reasoning coverage by exploiting trajectory disagreements, with clear experimental validation.

arXiv:2611.07918v1 Announce Type: new 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.
Be the first to rate this article.

Comments & Questions (0)

Captcha
Please be respectful — let's keep the conversation friendly.

No comments yet

Be the first to comment!