Temporal video understanding on TOMATO
53AccuracyGPT-4o
Evaluation Results
| Method | Links | |
|---|---|---|
| GPT-4o2026.06 | 53 | |
| Gemini 1.5 ProSource=[60]2026.06 | 48.6 | |
| Gemini 1.5 ProSource=[59]2026.06 | 48.3 | |
| GPT-4o mini2026.06 | 44.1 | |
| AdaCodecToken budget=comparable, Backbone=Qwen2-VL-7B2026.06 | 40 | |
| AdaCodecToken budget=1/7, Backbone=Qwen2-VL-7B2026.06 | 39.8 | |
| Claude 3.5 Sonnet2026.06 | 39.6 | |
| Molmo2-8B2026.06 | 39.6 | |
| Gemini 1.5 Flash2026.06 | 39.1 | |
| Molmo2-O-7B2026.06 | 36.2 | |
| Qwen2-VL-7B2026.06 | 35.7 | |
| PLM-7B2026.06 | 33.2 | |
| Key-VL-1.5-8B2026.06 | 33 | |
| VideoChat-Flash-7B2026.06 | 32.5 | |
| Eagle-X5-7B2026.06 | 31 | |
| GLM-4V-9B2026.06 | 30 | |
| MiniCPM-V-2.6-8B2026.06 | 29.8 | |
| CoPE-VideoLM-7B2026.06 | 28.3 | |
| LLaVA-Video-7B2026.06 | 24.9 | |
| InternVL2-8B2026.06 | 24.6 |