Search papers, labs, and topics across Lattice.
1
0
2
1
Triage quality in health AI agents varies dramatically, with top models achieving 88% success while the weakest falter at just 32%, underscoring the urgent need for better evaluation frameworks.