[论文] DoubtLess: Training-Free Contrastive Decoding for Efficient Reasoning

论文概要

研究领域: NLP 作者: Minghao Luo, Liang Chen 发布时间: 2026-10-13 arXiv: 2610.18862

中文摘要

推理模型通过长思维链（CoT）取得强大性能，但往往花费比解题实际所需更多的 token。现有的 CoT 缩短方法需要为每个模型额外训练、向模型隐藏状态添加固定转向向量、或使用提前停止或禁用「wait」等词解码启发式方法，这些都可能降低准确率。我们提出 DoubtLess——一种无需训练的解码方法，用一个更小的推理模型引导大模型走向简洁。小模型对大模型目前的 CoT 阅读两遍，一遍以简洁提示（简洁思考者），一遍以冗长提示（冗长思考者），其两个 logit 向量的差值（乘以强度系数）在 CoT 的每一步被加到大模型的 logits 上。在竞赛数学、研究生水平科学和竞赛编程共七个基准上，DoubtLess 将 Qwen3-32B 的 CoT 缩短 31% 且准确率不变，优于五个推理时基线。我们还证明 DoubtLess 在自调试编码智能体中每轮减少 20% 思考 token 且任务性能不变。它还可泛化到另外两个模型族而不影响准确率。与基于训练的压缩方法结合时，它还能进一步缩短已针对简洁性训练过的模型的 CoT。

原文摘要

Reasoning models achieve strong performance through long chains of thought (CoT), but often spend more tokens than needed to solve a problem. Existing ways to shorten the CoT require additional training for each model, add a fixed steering vector to the model's hidden states, or use decoding heuristics such as stopping early or banning words like "wait", which can reduce accuracy. We propose DoubtLess, a training-free decoding method in which a much smaller reasoning model steers the large one towards brevity. The small model reads the large model's CoT so far twice, once prompted to be brief (the brief thinker) and once prompted to be verbose (the verbose thinker), and the difference between its two logit vectors, scaled by a strength, is added to the large model's logits at every step of...

#论文 #arXiv #NLP #小凯