StepBack：步骤级错误定位重采样方法实现高效测试时推理
StepBack: Step-Level Error-Localized Resampling for Efficient Test-Time Reasoning
AI 导读
论文提出 StepBack，一种步骤级错误定位重采样方法，用于高效测试时推理。该方法从少量草稿轨迹中定位首个可能出错步骤，仅重采样该步骤及其后缀，再进行多数投票。在 7 个数学推理与多跳 QA 基准上，StepBack 平均精度与 self-consistency 相当，同时 token 用量最多减少 41.26%，并可与 Adaptive Consistency 等自适应推理方法结合。
正文 · 原文
Abstract:Self-consistency (SC) is a widely used decoding strategy for improving chain-of-thought reasoning, but it requires sampling many complete reasoning trajectories with a preset budget. Its efficient variants reduce the cost mainly by stopping full-trajectory sampling early or distributing the number of samples across questions. In this paper, we propose StepBack, a step-level error-localized resampling method for efficient test-time reasoning. Given a small set of draft trajectories, StepBack localizes likely first error positions using lightweight signals and then resamples the detected error step and its suffix, followed by majority answer voting. Empirically, StepBack matches the accuracy of SC on average across seven mathematical reasoning and multi-hop QA benchmarks, while reducing token usage by up to 41.26%. In addition, by nature, StepBack can be combined with adaptive inference methods such as Adaptive Consistency, further improving their inference-time efficiency.
| Comments: | NeurIPS 2026 |
| Subjects: | Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG) |
| Cite as: | arXiv:2611.08403 [cs.AI] |
| (or arXiv:2611.08403v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2611.08403 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Jung-Eun Kim [view email]
[v1]
Thu, 5 Nov 2026 18:12:49 UTC (141 KB)
来源：arXiv cs.AI · arxiv.org