Reward Modeling on ExeVR-Bench Overall
84.7AccuracyExeVRM
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| ExeVRMParameters=8B2026.03 | 84.7 | 82.9 | 87.7 | |
| Seed-2.0 ProModel Category=Proprietary2026.03 | 80.3 | 83.9 | 74.7 | |
| ExeVRMParameters=4B2026.03 | 80.1 | 79.2 | 82.5 | |
| Gemini 3 ProModel Category=Proprietary2026.03 | 75.1 | 74.2 | 76.7 | |
| GPT-5.2Model Category=Proprietary2026.03 | 75 | 82.7 | 66.5 | |
| Qwen3-VLParameters=4B2026.03 | 72.9 | 73.7 | 71.3 | |
| Gemini 2.5 ProModel Category=Proprietary2026.03 | 72.8 | 83.5 | 56.7 | |
| Qwen3-VLParameters=8B2026.03 | 67.7 | 77.6 | 49.9 | |
| Qwen2.5-VLParameters=7B2026.03 | 64.5 | 66.6 | 59.2 | |
| InternVL-3.5Parameters=8B2026.03 | 56.6 | 58.9 | 55.9 | |
| LLaVA-Next-VideoParameters=7B2026.03 | 13.6 | 48.4 | 19.4 |