让AI多“想”几遍再回答，白捡的准确率
大模型回答前多“想”几遍，通常要花更多算力。但有一类模型天生就会反复思考——它把同一个计算块循环执行好几轮，每轮都产出一个中间结果。过去，这些中间结果被直接扔掉，只留最后一轮。这篇论文发现，中间结果其实是有用的“草稿”:前几轮想得浅、容易错，最后一轮想得深、更准，两者一对比，就能像考试时排除明显错误选项一样，帮模型选出更靠谱的答案。研究者做了个无需额外训练的框架，在多个模型上把数学竞赛题准确率从61.88%提到73.33%，代码题从22.56%提到31.71%。更妙的是，因为答案更准了，模型可以少循环几轮，算力反而省了22%到48%。这不是你明天能直接用的工具，但它指向一个趋势：AI的“思考过程”本身正在变成可挖掘的资产，而不是被丢弃的废料。
📄 原文摘要(英文)
Looped Transformers achieve parameter efficiency by repeatedly executing a shared block across recurrent loops. Each loop yields an intermediate representation decodable for the same next token, yet standard decoding discards earlier states. Because earlier loops embody less computation, recurrence inherently supplies aligned weak-and-strong prediction pairs without auxiliary models or external training. We introduce LoopCD, a training-free contrastive decoding framework that guides token selection by contrasting the final prediction with an earlier recurrent pass, operating either in logit space with one extra output pass (LoopCD-Logits) or in hidden-state space with zero output overhead (LoopCD-Hidden). Across four looped Transformer families, LoopCD delivers substantial, consistent gains at full recurrent depth: LoopCD-Logits raises Ouro-2.6B-Thinking's AIME 2024 pass@1 from 61.88% to 73.33%, while LoopCD-Hidden lifts Huginn's HumanEval pass@1 from 22.56% to 31.71%. Crucially, these performance gains enable halving the number of recurrent loops while still matching or exceeding full-depth unguided baselines, reducing forward FLOPs by 22.5% to 48.2%. By transforming intermediate recurrent states into effective guidance signals, LoopCD achieves superior decoding quality while substantially reducing inference compute.