Multi-modal Understanding on MMBench (test) (Language Scores)
65.3MMBench Accuracy (En)ScalSelect
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ScalSelectSelection Budget=400K2026.02 | 65.3 | 58.97 | |
| Full-FinetuneSelection Budget=625K (Full)2026.02 | 63.96 | 56.73 | |
| ScalSelectSelection Budget=300K2026.02 | 63.05 | 55.09 | |
| COINCIDESelection Budget=100K2026.02 | 62.16 | 55.44 | |
| PRISMSelection Budget=100K2026.02 | 61.49 | 52.96 | |
| ScalSelectSelection Budget=200K2026.02 | 61.1 | 54.15 | |
| ScalSelectSelection Budget=100K2026.02 | 59.19 | 52.8 | |
| RandomSelection Budget=100K2026.02 | 57.96 | 51.74 | |
| ScalSelectSelection Budget=50K2026.02 | 57.79 | 51.68 | |
| RDS+Selection Budget=100K2026.02 | 56.05 | 50.16 | |
| LengthSelection Budget=100K2026.02 | 49.1 | 36.43 | |
| PerplexitySelection Budget=100K2026.02 | 31.74 | 32.47 |