Search papers, labs, and topics across Lattice.
Carnegie Mellon University
1
0
3
Continual learning for LLM agents hits a wall: scaling models doesn't reliably improve skill generation, and self-feedback can lead to recursive drift.
