Audio-Video Understanding on AV-Odyssey
32.1ScoreQwen2.5-Omni + AVATAR
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen2.5-Omni + AVATARReinforcement Learning Strategy=AVATAR2025.08 | 32.1 | |
| Qwen2.5-Omni + GRPOReinforcement Learning Strategy=GRPO2025.08 | 31.3 | |
| Omni-R1Reinforcement Learning Strategy=N/A2025.08 | 31.2 | |
| EcholnkReinforcement Learning Strategy=N/A2025.08 | 31.1 | |
| HumanOmniReinforcement Learning Strategy=N/A2025.08 | 30.3 | |
| Qwen2.5-OmniReinforcement Learning Strategy=Baseline2025.08 | 29.8 | |
| Ola-7B + AVATARReinforcement Learning Strategy=AVATAR2025.08 | 28.8 | |
| Ola-7B + GRPOReinforcement Learning Strategy=GRPO2025.08 | 27 | |
| AV-ReasonerReinforcement Learning Strategy=N/A2025.08 | 25.6 | |
| Ola-7BReinforcement Learning Strategy=Baseline2025.08 | 25.6 |