通过实时检测熵突增，动态修复大模型推理错误。
SPREG: Structured Plan Repair with Entropy-Guided Test-Time Intervention for Large Language Model Reasoning
- 用熵突增识别逻辑错误，触发动态修复机制。
- 在AIME25上准确率提升20.0%，有效抑制熵漂移。
- 适合需要高可靠推理的复杂任务场景。
大语言模型在长链推理中易出现逻辑幻觉和随机漂移。尽管无分类器引导（CFG）能增强指令遵循性，但传统静态实现常导致语义稀释和语言质量下降。我们提出SPREG（结构化计划引导的实时熵门控），一种轻量级推理时框架，用于精准纠错。SPREG采用自适应双阈值机制实时监控熵变化，将突发的“熵峰”作为逻辑失败的可靠指标。检测到异常后，通过用历史高置信状态合成的参考分布替换无信息的空先验，动态修正推理路径。根据结构化推理阶段（如动作、观察）调节引导强度，使模型重回稳定流形而不影响流畅性。实验表明，SPREG在AIME25上取得20.0%的绝对准确率提升，有效控制复杂任务中的不可控熵漂移。
原文摘要 · Abstract (English)
Large Language Models (LLMs) are prone to logical hallucinations and stochastic drifts during long-chain reasoning. While Classifier-Free Guidance (CFG) can improve instruction adherence, standard static implementations often cause semantic dilution and linguistic degradation. We propose SPREG (Structured Plan-guided Real-time Entropy Gating), a lightweight inference-time framework for surgical error rectification. SPREG employs an adaptive dual-threshold mechanism to monitor real-time entropy, identifying sudden ``entropy spikes'' as reliable indicators of logical failure. Upon detection, it triggers a dynamic repair by replacing uninformative null-priors with reference distributions synthesized from historical high-confidence states. By modulating guidance intensity according to structured reasoning stages (e.g., Action, Observation), SPREG steers the model back to a stable manifold without compromising fluency. Our experiments demonstrate significant gains, notably a 20.0% absolute accuracy improvement on AIME25, while effectively suppressing uncontrolled entropy drift in complex tasks.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。