用深层思考标记量化大模型推理努力，提升准确率并降低成本
Think Deep, Not Just Long: Measuring LLM Reasoning Effort via Deep-Thinking Tokens
- 识别深层思考令牌：模型深层迭代修正的预测标记
- 深层思考比例越高，推理准确率越强，超越长度与置信度指标
- 提出Think@n策略，早期淘汰低潜力生成，降低推理开销
大语言模型通过延长思维链（CoT）提升推理能力，但单纯依赖生成长度不可靠：长度增长不必然提升准确率，反而可能引发‘过度思考’导致性能下降。本文提出通过识别深层思考令牌——即在深层模型层中预测发生显著修正的令牌——来量化推理过程中的实际努力。在四个高难度数学与科学基准（AIME 24/25、HMMT 25、GPQA-diamond）及多个推理型模型（GPT-OSS、DeepSeek-R1、Qwen3）上验证，深层思考比例与准确率呈现稳定正相关，显著优于基于长度或置信度的基线方法。基于此，提出Think@n测试时扩展策略，优先保留高深层思考比例样本。实验表明，Think@n可达到或超过标准自一致性性能，同时通过短前缀提前剔除低潜力生成，大幅降低推理成本。
原文摘要 · Abstract (English)
Large language models (LLMs) have demonstrated impressive reasoning capabilities by scaling test-time compute via long Chain-of-Thought (CoT). However, recent findings suggest that raw token counts are unreliable proxies for reasoning quality: increased generation length does not consistently correlate with accuracy and may instead signal "overthinking," leading to performance degradation. In this work, we quantify inference-time effort by identifying deep-thinking tokens -- tokens where internal predictions undergo significant revisions in deeper model layers prior to convergence. Across four challenging mathematical and scientific benchmarks (AIME 24/25, HMMT 25, and GPQA-diamond) and a diverse set of reasoning-focused models (GPT-OSS, DeepSeek-R1, and Qwen3), we show that deep-thinking ratio (the proportion of deep-thinking tokens in a generated sequence) exhibits a robust and consistently positive correlation with accuracy, substantially outperforming both length-based and confidence-based baselines. Leveraging this insight, we introduce Think@n, a test-time scaling strategy that prioritizes samples with high deep-thinking ratios. We demonstrate that Think@n matches or exceeds standard self-consistency performance while significantly reducing inference costs by enabling the early rejection of unpromising generations based on short prefixes.
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。