Building AI Research Teams for Autonomous Discovery of Linear Attention Architectures

One-line takeaway

UCSB builds teams of AI researcher agents with varied expertise to autonomously discover linear-attention architectures, using diversity for orthogonal exploration and balancing exploration and exploitation when choosing which researchers propose and run the next costly experiments.

UCSB 用多个专长不同的 AI 研究员智能体组队，自动探索线性注意力架构：靠不同背景促成正交探索、跳出局部最优，并在实验昂贵、只能测少量想法时平衡探索与利用，挑选由谁提出和执行下一轮实验。