Search papers, labs, and topics across Lattice.
1
0
3
4
Audio-language models struggle significantly with hidden evaluation tasks, with accuracy plummeting by nearly 12 percentage points on average, highlighting the challenge of true audio comprehension.
