Search papers, labs, and topics across Lattice.
1
0
3
2
State-of-the-art Large Audio Language Models are surprisingly vulnerable to hallucination attacks, with success rates as high as 95%, revealing a critical reliability gap masked by standard benchmarks.
