提示策略在扩展推理大语言模型中的温度依赖性能
Temperature-Dependent Performance of Prompting Strategies in Extended Reasoning Large Language Models
AI总结：
研究评估了不同温度下链式思维和零样本提示策略在扩展推理中的表现，发现零样本提示在中等温度下性能最佳，链式思维在极端温度下表现最佳，扩展推理的收益随温度增加而提升。
摘要：
扩展推理模型通过在测试时进行显式计算，显著提升了大语言模型（LLM）在复杂问题解决中的能力。然而，这些系统中采样温度和提示策略的最优配置仍缺乏研究。本文系统评估了链式思维和零样本提示策略，在四个温度设置（0.0、0.4、0.7和1.0）下使用Grok-4.1在AMO-Bench上的39道数学问题上进行扩展推理。研究发现，零样本提示在中等温度下达到峰值性能，59%的准确率在T=0.4和T=0.7时取得，而链式思维提示在温度极值处表现最佳。最值得注意的是，扩展推理的收益从T=0.0时的6倍增加到T=1.0时的14.3倍。这些结果表明，温度应与提示策略联合优化，挑战了在推理任务中常用T=0的做法。
英文摘要：
Extended reasoning models represent a transformative shift in Large Language Model (LLM) capabilities by enabling explicit test-time computation for complex problem solving. However, the optimal configuration of sampling temperature and prompting strategy for these systems remains largely underexplored. We systematically evaluate chain-of-thought and zero-shot prompting across four temperature settings (0.0, 0.4, 0.7, and 1.0) using Grok-4.1 with extended reasoning on 39 mathematical problems from AMO-Bench, a challenging International Mathematical Olympiad-level benchmark. We find that zero-shot prompting achieves peak performance at moderate temperatures, reaching 59% accuracy at T=0.4 and T=0.7, while chain-of-thought prompting performs best at the temperature extremes. Most notably, the benefit of extended reasoning increases from 6x at T=0.0 to 14.3x at T=1.0. These results suggest that temperature should be optimized jointly with prompting strategy, challenging the common practice of using T=0 for reasoning tasks.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
