Tag: Tool-Integrated Reasoning
All the articles with the tag "Tool-Integrated Reasoning".
-
Agent RL Scaling Law: Agent RL with Spontaneous Code Execution for Mathematical Problem Solving
本文通过ZeroTIR框架利用强化学习训练基础大型语言模型自发执行Python代码解决数学问题,揭示了训练步数与代码使用频率、响应长度及任务准确率的正相关规律(Agent RL Scaling Law),并在数学基准上显著优于无工具基线。