Search papers, labs, and topics across Lattice.
1
0
3
Reward hacking isn't a bug to fix, but an inevitable consequence of how we evaluate AI, and it gets exponentially worse as agents gain more tools.
