Video Question Answering on WorldSense
54AccuracyQwen3-Omni-30B-A3B-Instruct
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen3-Omni-30B-A3B-InstructEvaluation Setting=Direct QA2026.06 | 54 | |
| Gemini-2.5-FlashEvaluation Setting=Direct QA2026.06 | 50.9 | |
| OmniCaptioner-IF-7BEvaluation Setting=Direct QA2026.06 | 49.4 | |
| R-MSD (4B)Frames=642026.03 | 49.2 | |
| video-SALMONN-2-7BEvaluation Setting=Direct QA2026.06 | 48.6 | |
| Qwen3-VL-4B-InstructFrames=642026.03 | 46.7 | |
| Original SFT+RL (4B)Frames=642026.03 | 46.3 | |
| R-MSD (2B)Frames=642026.03 | 45.7 | |
| Qwen2.5-Omni-7BEvaluation Setting=Direct QA2026.06 | 45.4 | |
| OmniCaptioner-IF-7BEvaluation Protocol=Caption-to-QA, LLM-as-a-judge=Gemini-2.5-Pro2026.06 | 43.2 | |
| Qwen3-VL-2B-InstructFrames=642026.03 | 42.6 | |
| GPT-4oEvaluation Setting=Direct QA2026.06 | 42.6 | |
| ASID-Captioner-7BEvaluation Protocol=Caption-to-QA, LLM-as-a-judge=Gemini-2.5-Pro2026.06 | 34 | |
| Gemini-2.5-ProEvaluation Protocol=Caption-to-QA, LLM-as-a-judge=Gemini-2.5-Pro2026.06 | 33.8 | |
| Gemini-2.5-FlashEvaluation Protocol=Caption-to-QA, LLM-as-a-judge=Gemini-2.5-Pro2026.06 | 31 | |
| AVoCaDO-7BEvaluation Protocol=Caption-to-QA, LLM-as-a-judge=Gemini-2.5-Pro2026.06 | 25.7 | |
| VideoLLaMA2-7BEvaluation Setting=Direct QA2026.06 | 25.4 | |
| video-SALMONN-2-7BEvaluation Protocol=Caption-to-QA, LLM-as-a-judge=Gemini-2.5-Pro2026.06 | 18.2 | |
| Qwen3-Omni-Instruct-30B-A3BEvaluation Protocol=Caption-to-QA, LLM-as-a-judge=Gemini-2.5-Pro2026.06 | 12.7 | |
| UGC-VideoCaptioner-3BEvaluation Protocol=Caption-to-QA, LLM-as-a-judge=Gemini-2.5-Pro2026.06 | 11.2 | |
| ARC-Hunyuan-Video-7BEvaluation Protocol=Caption-to-QA, LLM-as-a-judge=Gemini-2.5-Pro2026.06 | 8.7 | |
| Qwen2.5-Omni-7BEvaluation Protocol=Caption-to-QA, LLM-as-a-judge=Gemini-2.5-Pro2026.06 | 8.6 | |
| MiniCPM-o-2.6-8BEvaluation Protocol=Caption-to-QA, LLM-as-a-judge=Gemini-2.5-Pro2026.06 | 7.2 | |
| HumanOmniV2-7BEvaluation Protocol=Caption-to-QA, LLM-as-a-judge=Gemini-2.5-Pro2026.06 | 6.6 |