1 paper · 1 filter
Fan He, Yan Li, Xiaoyang Zeng
Large language models are increasingly composed into agent loops that plan, call tools, and resume the same task after each action. These loops press a shared memory hierarchy hard…