Counting on CoMET-Bench
2.7MAECoMET-Agent (Gemini 3 Flash)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| CoMET-Agent (Gemini 3 Flash)#Size=-2026.06 | 2.7 | 62 | 12 | |
| CoMET-Agent (GPT-5)#Size=-2026.06 | 2.8 | 61.4 | 13.7 | |
| GPT-5#Size=-2026.06 | 3.2 | 56.8 | 9 | |
| TRACE#Size=7B2026.06 | 3.2 | 60.9 | 1 | |
| Gemini 2.5 Pro#Size=-2026.06 | 3.3 | 53.6 | 4.8 | |
| DisTime#Size=7B2026.06 | 3.3 | 60 | -0.6 | |
| TimeLens#Size=7B2026.06 | 3.3 | 60.7 | 0 | |
| Gemini 3 Flash#Size=-2026.06 | 3.4 | 54.2 | 6.6 | |
| LLaVA-Video#Size=7B2026.06 | 3.5 | 55.3 | 2.4 | |
| InternVL3.5#Size=8B2026.06 | 3.6 | 52.3 | 3.2 | |
| Qwen3-VL#Size=30B2026.06 | 3.6 | 51.5 | 3.4 | |
| VideoARM#Size=-2026.06 | 3.6 | 52.1 | 14.5 | |
| LLaVA-OV1.5#Size=7B2026.06 | 3.7 | 47.7 | 3.9 | |
| LITA#Size=13B2026.06 | 3.7 | 58.4 | 0.2 | |
| MiMO-VL#Size=7B2026.06 | 3.8 | 48.5 | -1.3 | |
| Eagle2.5#Size=8B2026.06 | 3.9 | 48.8 | 1 | |
| CoMET-Agent (Qwen3-VL)#Size=8.6B2026.06 | 3.9 | 54.5 | 11 | |
| InternVL3.5#Size=38B2026.06 | 4 | 48.3 | 13.4 | |
| VideoMind#Size=7B2026.06 | 4.4 | 14.8 | -0.5 | |
| Qwen3-VL#Size=8.3B2026.06 | 4.8 | 51.7 | 9.8 | |
| LongVILA-R1#Size=7B2026.06 | 5.7 | 38.2 | 11.1 |