用闭环控制动态调节大模型推理，减少冗余思考并提升准确率。
Adaptive Activation Steering for Efficient LLM Reasoning via Closed-Loop PID Control
- 基于轻量级冗余检测器的PID控制器动态调整抑制强度。
- 在GSM8K上准确率提升3.9个百分点，输出长度减少23%。
- 无需训练，适合追求高效推理的部署场景。
通过长链思维训练的推理型大模型常出现过度思考：耗费大量令牌在冗余的反思与过渡上，增加成本却未提升准确率。静态激活调节（如SEAL）使用固定向量抑制此类内容，但无论当前片段冗余程度如何，均采用相同强度。本文提出PID-steering，一种无需训练、解码时使用的动态调节方法，通过轻量级片段级冗余分类器驱动的PID控制器调节抑制强度。在深度求解-1.5B小模型的GSM8K子集上，准确率从85.7%提升至89.6%（+3.9个百分点），平均输出长度由1026降至790令牌（-23%）。该结果为小规模概念验证，非基准性能。
原文摘要 · Abstract (English)
Reasoning LLMs trained with long chain-of-thought often overthink: they spend tokens on redundant reflection and transitions that inflate cost without improving accuracy. Static activation steering (e.g.\ SEAL) suppresses such content with a fixed vector, but applies the same strength regardless of how redundant the current chunk actually is. We describe PID-steering, a training-free, decoding-time method that modulates the steering strength with a PID controller driven by a lightweight chunk-level redundancy classifier. On a subset of GSM8K with DeepSeek-R1-Distill-Qwen-1.5B, the method improves accuracy from 85.7\% to 89.6\% (+3.9 pp) while cutting average output length from 1026 to 790 tokens ($-$23\%). We report it as a small-scale proof of concept rather than a benchmark result.
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。