Gen-Verse/Open-AgentRL-SFT-3K
收藏官方服务:
资源简介:
这是一个包含3K个代理推理场景的自制、真实、端到端多轮SFT(情景代理任务)数据集,它由s1k、ReTool-SFT和自制的LeetCode问题组成,用于研究和提高代理推理的强化学习算法。
This is a self-curated, real, end-to-end multi-turn SFT (Situation-Agent Task) dataset consisting of 3K agentic reasoning scenarios, composed of s1k, ReTool-SFT, and self-curated LeetCode problems, designed for researching and improving reinforcement learning algorithms for agentic reasoning.
提供机构:
Gen-Verse


