Search papers, labs, and topics across Lattice.
1
0
2
Harmful prompts can slip through safety nets when embedded in benign batches, revealing a critical vulnerability in current AI safety measures.
