veggiebird/MATPO-data

MATPO是一个多智能体工具集成策略优化数据集，它允许在单个大型语言模型内训练多个专业智能体角色（规划器和工作者智能体）。数据集通过强化学习训练，旨在解决当前单智能体方法在多轮工具集成规划中的关键限制，如上下文长度瓶颈和噪声工具响应问题。MATPO is a Multi-Agent Tool-Integrated Policy Optimization dataset designed for training multiple specialized agent roles (planner and worker agents) within a single large language model using reinforcement learning. It addresses the limitations of current single-agent approaches for multi-turn tool-integrated planning, such as context length bottleneck and noisy tool responses.