Research map / Vision-language and multimodal / Cross-modal learning
When AVSR Meets Video Conferencing: Dataset, Degradation, and the Hidden Mechanism Behind Performance Collapse
Yihuan Huang, Jun Xue, Liu Jiajun, Daixian Li, Tong Zhang, Zhuolin Yi, Yanzhen Ren, Kai Li
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| fine-tune on MLD-VC | LiPS-AVSR · Chinese-Lips (Tencent Meeting) | CER: 10.97 -> 9.65 | Improved |
| fine-tune on MLD-VC | LiPS-AVSR · Chinese-Lips (Lark) | CER: 18.53 -> 13.64 | Improved |
| fine-tune on MLD-VC | LiPS-AVSR · Chinese-Lips (Zoom) | CER: 9.22 -> 7.93 | Improved |
| fine-tune on MLD-VC | LiPS-AVSR · MLD-VC test | CER: 42.37 -> 13.91 | Improved |
| add online (transmitted) data to fine-tuning (ablation) | LiPS-AVSR · Chinese-Lips (Zoom) | CER: 10.53 -> 7.93 | Improved |
| add online (transmitted) data to fine-tuning (ablation) | LiPS-AVSR · Chinese-Lips (Lark) | CER: 15.52 -> 13.64 | Improved |
| add online (transmitted) data to fine-tuning (ablation) | LiPS-AVSR · Chinese-Lips (Tencent Meeting) | CER: 10.15 -> 9.65 | Improved |
| add hyper-expression (Lombard) data to fine-tuning (ablation) | LiPS-AVSR · Chinese-Lips (Zoom) | CER: 9.61 -> 7.93 | Improved |
| add hyper-expression (Lombard) data to fine-tuning (ablation) | LiPS-AVSR · Chinese-Lips (Lark) | CER: 14.48 -> 13.64 | Improved |
| add hyper-expression (Lombard) data to fine-tuning (ablation) | LiPS-AVSR · Chinese-Lips (Tencent Meeting) | CER: 10.01 -> 9.65 | Improved |
| evaluate on Lombard-Grid instead of LRS3 (Lombard-containing data) under Zoom | Auto-AVSR · audio-visual Zoom | WER: 33.09 -> 12.36 | Improved |
| add visual modality to audio-only under Zoom | mWhisper-Flamingo · LRS3 (Zoom) | WER: 10.38 -> 9.22 | Improved |
Closest papers by meaning
- $\texttt{AVROBUSTBENCH}$: Benchmarking the Robustness of Audio-Visual Recognition Models at Test-Time (NeurIPS 2025 · similarity 0.87)
- Multi-Task Corrupted Prediction for Learning Robust Audio-Visual Speech Representation (ICLR 2025 · similarity 0.86)
- VSRo-200: A Romanian Visual Speech Recognition Dataset for Studying Supervision and Multimodal Robustness (NeurIPS 2026 · similarity 0.86)
- Enhancing Audiovisual Speech Recognition Through Bifocal Preference Optimization (AAAI 2025 · similarity 0.86)
- Unified Speech Recognition: A Single Model for Auditory, Visual, and Audiovisual Inputs (NeurIPS 2024 · similarity 0.86)
- MoHAVE: Mixture of Hierarchical Audio-Visual Experts for Robust Speech Recognition (ICML 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
