模型推理初期就已内含正确答案预测信号，仅需数步即可判断结果。
Temporal Predictors of Outcome in Reasoning Language Models
- 用前t步隐状态训练线性分类器，探测模型何时锁定答案
- 仅需几步就能高精度预测最终正确性，即使题目需长推理链
- 难题常出现在长推理链中，导致预测准确率下降的假象
思维链（CoT）范式通过逐步推理过程作为推理的代理，逐步优化模型对解的潜在表示。然而，大语言模型在何时内部确定最终结果仍不明确。我们通过在前t个推理标记后的隐藏状态上训练线性分类器来探究这一问题，发现即使需要较长输出才能得出明确答案，最终正确性在仅几轮推理后就已高度可预测。对于更难的问题，预测准确率下降揭示了选择偏差：困难样本在长思维链中占比较高。总体而言，结果表明推理模型的内部自我评估在数步内即已出现，这对可解释性和推理时控制具有重要意义。
原文摘要 · Abstract (English)
The chain-of-thought (CoT) paradigm uses the elicitation of step-by-step rationales as a proxy for reasoning, gradually refining the model's latent representation of a solution. However, it remains unclear just how early a Large Language Model (LLM) internally commits to an eventual outcome. We probe this by training linear classifiers on hidden states after the first t reasoning tokens, showing that eventual correctness is highly predictable after only a few tokens, even when longer outputs are needed to reach a definite answer. We show that, for harder questions, a drop in predictive accuracy highlights a selection artifact: hard items are disproportionately represented in long CoTs. Overall, our results imply that for reasoning models, internal self-assessment of success tends to emerge after only a few tokens, with implications for interpretability and for inference-time control.
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。