Search papers, labs, and topics across Lattice.
1
0
3
37
Despite the intuition that noisy environments should make models rely more on visual cues, AVSR models stubbornly cling to audio, even when it's heavily degraded.
