Search papers, labs, and topics across Lattice.
Artificial Intelligence Laboratory, Fudan University
1
0
2
A unified framework reveals that existing LLM policy optimization methods often overlook compound failures that require simultaneous adjustments to both trajectory and reward components.
