Search papers, labs, and topics across Lattice.
1
0
2
1
Single-score evaluations hide critical differences in multi-party conversational AI, so MPCEval breaks down generation quality into speaker modeling, content, and consistency to reveal nuanced model behaviors.
