Stanford 与 Together AI 论文提出 Self-Organizing Agent Teams，3 模型协作团队准确率达 66.7%
中文摘要
Stanford与Together AI提出SAT框架，通过学习过往协作分工与流程实现自我组织，3模型协作准确率达66.7%。
English Summary
Stanford and Together AI introduced SAT, letting agent teams learn reusable roles and workflows from past collaborations instead of debating, reaching 66.7% accuracy with three models.
原文节选
Stanford 与 Together AI 的论文提出 Self-Organizing Agent Teams（SAT），让固定智能体团队从过往协作中学习可复用的角色分工与流程，而非辩论后投票。