Search papers, labs, and topics across Lattice.
1
0
4
Current AI safety filters can't tell a joke from a threat, especially when humor relies on cultural context – this new benchmark exposes that blind spot.
