Search papers, labs, and topics across Lattice.
1
0
3
5
ResponseGuard outperforms traditional reasoning-based guards in harmfulness detection while operating at a fraction of the time cost, reshaping our understanding of safety in AI responses.
