1 paper · 1 filter
Tianle Chen, Pushkar Singh, Ketan Todi +8
Language-model agents that run long tasks often redo work they have already done: they re-issue a tool call they made earlier and get back exactly the same observation, a step that…