Search papers, labs, and topics across Lattice.
1
0
2
0
LongRCA Bench reveals that diagnosing failures in long-horizon agents requires distinct metrics for responsible roles and root causes, with a new method achieving over 50% accuracy in role attribution.
