Complex Video Reasoning on VideoMathQA
30.5AccuracyVideoLatent-7B (Ours)
Evaluation results
| Method | Links | |
|---|---|---|
| VideoLatent-7B (Ours)Frames=16, Architecture Category=Latent MLLM2026.06 | 30.5 | |
| VideoLatent-7B (Ours)Frames=32, Architecture Category=Latent MLLM2026.06 | 30 | |
| Qwen2.5-VL-7BFrames=64, Architecture Category=Standard MLLM2026.06 | 29.8 | |
| VideoLatent-7B (Ours)Frames=64, Architecture Category=Latent MLLM2026.06 | 29.5 | |
| Qwen2.5-VL-7BFrames=16, Architecture Category=Standard MLLM2026.06 | 29 | |
| LVR-7BFrames=64, Architecture Category=Latent MLLM2026.06 | 28.8 | |
| Qwen2.5-VL-7BFrames=32, Architecture Category=Standard MLLM2026.06 | 27.9 | |
| LVR-7BFrames=16, Architecture Category=Latent MLLM2026.06 | 26.9 | |
| Video-R1-7BFrames=16, Architecture Category=Standard MLLM2026.06 | 26.4 | |
| Mull-Token-7BFrames=16, Architecture Category=Latent MLLM2026.06 | 26.4 | |
| Open-o3-Video-7BFrames=32, Architecture Category=Standard MLLM2026.06 | 25.7 | |
| Video-R1-7BFrames=64, Architecture Category=Standard MLLM2026.06 | 25.2 | |
| LVR-7BFrames=32, Architecture Category=Latent MLLM2026.06 | 25 | |
| Open-o3-Video-7BFrames=64, Architecture Category=Standard MLLM2026.06 | 25 | |
| Mull-Token-7BFrames=64, Architecture Category=Latent MLLM2026.06 | 25 | |
| Mull-Token-7BFrames=32, Architecture Category=Latent MLLM2026.06 | 24.8 | |
| Video-R1-7BFrames=32, Architecture Category=Standard MLLM2026.06 | 24.3 | |
| Monet-7BFrames=32, Architecture Category=Latent MLLM2026.06 | 14 | |
| Monet-7BFrames=64, Architecture Category=Latent MLLM2026.06 | 12.9 | |
| Monet-7BFrames=16, Architecture Category=Latent MLLM2026.06 | 11.7 |