Search papers, labs, and topics across Lattice.
1
0
3
2
Unsupervised detection of adversarial attacks in RAG systems is possible using generator activations and uncertainty measures, even without knowing the target prompt.
