Negative Query Recognition on CoMET-Bench
68.1Rej.-F1CoMET-Agent (Gemini 3 Flash)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CoMET-Agent (Gemini 3 Flash)#Size=-2026.06 | 68.1 | 5.8 | |
| CoMET-Agent (GPT-5)#Size=-2026.06 | 66.5 | 6.4 | |
| InternVL3.5#Size=8B2026.06 | 64.7 | 16.8 | |
| Gemini 3 Flash#Size=-2026.06 | 61.5 | 51.7 | |
| VideoARM#Size=-2026.06 | 61 | 9.5 | |
| Gemini 2.5 Pro#Size=-2026.06 | 60.8 | 37.3 | |
| GPT-5#Size=-2026.06 | 60.6 | 5.3 | |
| CoMET-Agent (Qwen3-VL)#Size=8.6B2026.06 | 57.4 | 8 | |
| Qwen3-VL#Size=8.3B2026.06 | 52.1 | 5.3 | |
| LLaVA-Video#Size=7B2026.06 | 48.9 | 58.5 | |
| Qwen3-VL#Size=30B2026.06 | 47.2 | 5.9 | |
| InternVL3.5#Size=38B2026.06 | 40.3 | 9.1 | |
| Eagle2.5#Size=8B2026.06 | 39.9 | 14.6 | |
| DisTime#Size=7B2026.06 | 12.1 | 93.5 | |
| LLaVA-OV1.5#Size=7B2026.06 | 9.3 | 2.6 | |
| MiMO-VL#Size=7B2026.06 | 8.8 | 3.8 | |
| LongVILA-R1#Size=7B2026.06 | 5.6 | 97.1 | |
| TRACE#Size=7B2026.06 | 2.5 | 98.8 | |
| LITA#Size=13B2026.06 | 1.7 | 99.1 | |
| VideoMind#Size=7B2026.06 | 1.2 | 99.4 | |
| TimeLens#Size=7B2026.06 | 0 | 100 |