Vision-Language Understanding (Image) on RefCOCO Resolution-Sensitive and General
70.4Absolute Macro ScorePARCEL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PARCELVisual Budget=256 Tokens2026.05 | 70.4 | 95.1 | |
| PARCELVisual Budget=64 Tokens2026.05 | 70.1 | 94.7 | |
| MQTVisual Budget=256 Tokens2026.05 | 69.1 | 93.3 | |
| MQTVisual Budget=64 Tokens2026.05 | 68.1 | 91.6 | |
| M3Visual Budget=256 Tokens2026.05 | 67.6 | 91.1 | |
| M3Visual Budget=64 Tokens2026.05 | 66.5 | 89.2 | |
| PARCELVisual Budget=16 Tokens2026.05 | 64.8 | 86.8 | |
| MQTVisual Budget=16 Tokens2026.05 | 64.3 | 85.8 | |
| M3Visual Budget=16 Tokens2026.05 | 63.9 | 85.2 |