ResearchBenchmarksAudio-Visual Speech Recognition on AVSpeech (1,000 manually filtered samples)Follow25WERUSR 2.024.21629.50834.840.092Feb 22, 2026Evaluation ResultsMethodMethodLinksWERUSR 2.0Decoding=greedyDecoding=greedy2026.0225AV-HuBERTDecoding=greedyDecoding=greedy2026.0226USRDecoding=greedyDecoding=greedy2026.0234.7BRAVEnDecoding=greedyDecoding=greedy2026.0244.6