Cognition 模型 / Devin 博客·精选AI 评分61
Cognition 发布 SWE-grep 与 SWE-grep-mini,用 RL 训练快速并行上下文检索模型
Introducing SWE-grep and SWE-grep-mini: RL for Multi-Turn, Fast Context Retrieval2025.10.16We trained SWE-grep and SWE-grep-mini, fast agentic models specialized in highly parallel context retrieval. They match the retrieval capabilities of frontier coding models, while taking an order of magnitude less time. Available now in Windsurf’s new Fast Context subagent, and our new SWE-grep demo playground!
AI 导读
Cognition 发布 SWE-grep 和 SWE-grep-mini,用多轮强化学习训练的智能体检索模型,检索能力比肩前沿编码模型但耗时少一个数量级。模型每轮最多发出 8 个并行工具调用、最多 4 轮,借助 Cerebras 推理分别达到超过 2,800 tokens/秒和超过 650 tokens/秒。
推荐理由
官方披露了多轮 RL 训练并行检索模型的方法与训练细节,读者可以据此理解快慢模型分工的实现路径。
来源:Cognition 模型 / Devin 博客 · cognition.com