动态分配推理计算，让模型自适应优化生成方式。
Adaptive Test-Time Compute Allocation with Evolving In-Context Demonstrations
- 测试时动态分配计算资源，结合上下文示范逐步优化生成
- 在数学、编程和推理任务上表现优于基线，且推理耗能更低
- 适合需要高效推理的部署场景，尤其关注计算成本
尽管增加测试时计算量可显著提升模型性能，但现有方法或依赖静态计算分配，或从固定生成分布中采样。本文提出一种测试时计算分配框架，同时自适应决定计算投入位置与生成方式。方法首先通过预热阶段识别简单问题，并从测试集中构建初始问答对池。随后进入自适应阶段，将更多计算集中在未解决的问题上，并通过不断演化的上下文示范重设生成分布——即基于语义相关问题的成功响应来条件化每次生成，而非从固定分布重采样。在数学、编程和推理基准上的实验表明，该方法持续优于现有基线，且显著降低推理时计算消耗。
原文摘要 · Abstract (English)
While scaling test-time compute can substantially improve model performance, existing approaches either rely on static compute allocation or sample from fixed generation distributions. In this work, we introduce a test-time compute allocation framework that jointly adapts where computation is spent and how generation is performed. Our method begins with a warm-up phase that identifies easy queries and assembles an initial pool of question-response pairs from the test set itself. An adaptive phase then concentrates further computation on unresolved queries while reshaping their generation distributions through evolving in-context demonstrations -- conditioning each generation on successful responses from semantically related queries rather than resampling from a fixed distribution. Experiments across math, coding, and reasoning benchmarks demonstrate that our approach consistently outperforms existing baselines while consuming substantially less inference-time compute.
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。