只干预少数不确定的词，就能显著提升大模型推理准确率。
Less is More: Improving LLM Reasoning with Minimal Test-Time Intervention
- 仅在高不确定性位置施加轻量引导，避免全序列计算
- 在六项任务上提升9.28%，AIME2024达11.25%准确率
- 无需训练，兼容现有模型，适合高效推理场景
近期大语言模型的进展多依赖测试时扩展以提升推理能力，但常以效率为代价。我们重新审视测试时行为，发现推理不确定性高度集中——仅有少量高熵令牌显著影响输出正确性。基于此，提出无需训练的最小化测试时干预（MTI）框架：(i) 仅在不确定位置施加选择性分类器自由引导；(ii) 利用主模型的键值缓存高效近似无条件解码。MTI在通用、编程与STEM任务中均表现优异，例如对DeepSeek-R1-7B在六项基准上平均提升9.28%，在AIME2024上使用Ling-mini-2.0-模型达11.25%提升，且开销极低。
原文摘要 · Abstract (English)
Recent progress in large language models (LLMs) has focused on test-time scaling to improve reasoning via increased inference computation, but often at the cost of efficiency. We revisit test-time behavior and uncover a simple yet underexplored phenomenon: reasoning uncertainty is highly localized-only a small subset of high-entropy tokens dominantly affects output correctness. Motivated by this, we propose Minimal Test-Time Intervention (MTI), a training-free framework that enhances reasoning accuracy and stability with minimal overhead. MTI includes: (i) Selective CFG intervention, applying classifier-free guidance only at uncertain positions; and (ii) Lightweight negative-prompt guidance, reusing the main model's KV cache to approximate unconditional decoding efficiently. MTI yields consistent gains across general, coding, and STEM tasks-e.g., +9.28% average improvement on six benchmarks for DeepSeek-R1-7B and +11.25% on AIME2024 using Ling-mini-2.0-while remaining highly efficient.
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。