Multi-task Knowledge on MMSU
67.1AccuracyASR + Qwen2.5-7B
Evaluation results
| Method | Links | ||
|---|---|---|---|
| ASR + Qwen2.5-7BSystem Type=Cascaded topline, ASR=Whisper-v3-Large2025.10 | 67.1 | 3.7 | |
| Qwen2.5-Omni-7BSystem Type=End-to-end system2025.10 | 61 | -9.8 | |
| ASR + Qwen2.5-3BSystem Type=Cascaded topline, ASR=Whisper-v3-Large2025.10 | 58.1 | 3.8 | |
| SALAD-7BTraining Stage=Stage II2025.10 | 57.5 | 13.3 | |
| SALAD-7BTraining Stage=Stage I2025.10 | 55.3 | 15.5 | |
| SALAD-3BTraining Stage=Stage II2025.10 | 52.5 | 9.4 | |
| SALAD-3BTraining Stage=Stage I2025.10 | 47.3 | 14.6 | |
| GLM-4-Voice-9BSystem Type=End-to-end system2025.10 | 38.6 | 27.6 | |
| DiVA-Llama3.1-8BSystem Type=End-to-end system2025.10 | 36.1 | 25.8 | |
| Qwen2-Audio-7BSystem Type=End-to-end system2025.10 | 29.5 | 18.7 | |
| Random2025.10 | 25 | — |