Search papers, labs, and topics across Lattice.
1
0
2
A novel hacker-fixer loop can eliminate reward hacking vulnerabilities in agent benchmarks, transforming how we secure AI evaluation metrics.
