SlideASR on ChineseLips
1.298CERVAPO-7B
Evaluation Results
| Method | Links | |
|---|---|---|
| VAPO-7BContext setting=Slide image as context (End-to-End)2025.10 | 1.298 | |
| VAPO-3BContext setting=Slide image as context (End-to-End)2025.10 | 1.548 | |
| Qwen2.5-Omni-3BContext setting=Contextless2025.10 | 1.937 | |
| Qwen3-Omni-30B-A3BContext setting=Contextless2025.10 | 2.202 | |
| Qwen2.5-Omni-7BContext setting=Contextless2025.10 | 2.243 | |
| MiniCPM-o-2.6Context setting=Contextless2025.10 | 2.252 | |
| Mi-DashengContext setting=Contextless2025.10 | 3.311 | |
| Qwen2-AudioContext setting=Contextless2025.10 | 12.536 | |
| Qwen2.5-Omni-7BContext setting=Slide image as context (End-to-End)2025.10 | 14.34 | |
| Qwen2.5-Omni-3BContext setting=Slide image as context (End-to-End)2025.10 | 24.847 | |
| Qwen3-Omni-30B-A3BContext setting=Slide image as context (End-to-End)2025.10 | 41.93 | |
| Mi-DashengContext setting=Slide text as context (Pipeline)2025.10 | 65.505 | |
| Qwen3-Omni-30B-A3BContext setting=Slide text as context (Pipeline)2025.10 | 69.172 | |
| MiniCPM-o-2.6Context setting=Slide image as context (End-to-End)2025.10 | 76.203 | |
| Qwen2-AudioContext setting=Slide text as context (Pipeline)2025.10 | 84.291 |