Role-Playing Evaluation (Visual-Element-Groundedness) on VEG
65Win RateEBM
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| EBMOpponent=haruhi2026.05 | 65 | 18.4 | 16.6 | 46.6 | 77.9 | 1.42 | |
| EBMOpponent=qwen2.5-vl-7b-sft2026.05 | 53.6 | 34.2 | 12.2 | 19.4 | 61 | 4.23 | |
| EBMOpponent=intervl8b2026.05 | 53.2 | 31 | 15.8 | 22.2 | 63.2 | 7 | |
| EBMOpponent=RoleMRC2026.05 | 53.2 | 29.2 | 17.6 | 24 | 64.6 | 3.56 | |
| EBMOpponent=crab2026.05 | 53 | 28.2 | 18.8 | 24.8 | 65.3 | 7.68 | |
| EBMOpponent=qwen2.5-vl-7b2026.05 | 51.4 | 26.8 | 21.8 | 24.6 | 65.7 | 4.96 | |
| EBMOpponent=Internvl14b2026.05 | 46 | 33.4 | 20.6 | 12.6 | 57.9 | 0.0018 | |
| EBMOpponent=Qwen32b2026.05 | 45.6 | 36.6 | 17.8 | 9 | 55.5 | 0.0299 | |
| EBMOpponent=Internvl38b2026.05 | 40.8 | 40.6 | 18.6 | 0.2 | 50.1 | 1 |