Search papers, labs, and topics across Lattice.
1
0
3
Exploiting the mismatch between implicit human context and LLM safety alignment can lead to unprecedented attack success rates, revealing a critical vulnerability in current AI systems.
