Video-grounded Role-playing on Video-grounded Role-playing Dataset Movie Scripts
74.95VEGInternVL3-38B
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| InternVL3-38BModel Scale=38B, Model Type=General Visual-Language Model (VLM)2026.05 | 74.95 | 70.82 | 74.43 | 73.4 | |
| Qwen2.5-VL-32BModel Scale=32B, Model Type=General Visual-Language Model (VLM)2026.05 | 74.61 | 71.01 | 73.32 | 72.98 | |
| Char-EBM-CLIP-MaxReward Aggregation Method=Max, Backbone=Qwen2.5-VL-7B2026.05 | 74.25 | 70.37 | 74.78 | 73.13 | |
| InternVL3-14BModel Scale=14B, Model Type=General Visual-Language Model (VLM)2026.05 | 73.87 | 67.75 | 72.78 | 71.47 | |
| Char-EBM-CLIP-TopKReward Aggregation Method=TopK, Backbone=Qwen2.5-VL-7B2026.05 | 73.47 | 69.94 | 73.69 | 72.37 | |
| InternVL3-8BModel Scale=8B, Model Type=General Visual-Language Model (VLM)2026.05 | 71.85 | 65.53 | 72.76 | 70.05 | |
| Qwen2.5-VL-7B-SFTModel Type=General Visual-Language Model (VLM), Training=SFT on CoT data2026.05 | 70.82 | 66.46 | 72.08 | 69.79 | |
| Qwen2.5-VL-7BModel Scale=7B, Model Type=General Visual-Language Model (VLM)2026.05 | 70.75 | 65.05 | 71.71 | 69.17 | |
| RoleMRCModel Type=Strong Text-Only Role-Playing Model2026.05 | 70.51 | 66.77 | 73.14 | 70.14 | |
| CrabModel Type=Strong Text-Only Role-Playing Model2026.05 | 67.86 | 66.87 | 75.23 | 69.99 | |
| HaruhiModel Type=Strong Text-Only Role-Playing Model2026.05 | 67.09 | 62.09 | 72.38 | 67.19 |