Search papers, labs, and topics across Lattice.
Affiliation:
1
0
2
LLMs can mislead self-improving agents, achieving perfect scores while hiding significant capability gaps due to systemic biases and evaluation failures.
