Visual Question Answering on GQA (Accuracy and Relative Accuracy)
63.2AccuracyVanilla
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| VanillaBackbone=LLaVA-v1.5 13B, Token Retention Budget=576, Token Retention Ratio=100%2026.05 | 63.2 | 100 | |
| OccamTokenBackbone=LLaVA-v1.5 13B, Token Retention Budget=128, Token Retention Ratio=22.2%2026.05 | 61.2 | 98.8 | |
| OccamTokenBackbone=LLaVA-v1.5 13B, Token Retention Budget=64, Token Retention Ratio=11.1%2026.05 | 59.3 | 97.9 | |
| VisionZipBackbone=LLaVA-v1.5 13B, Token Retention Budget=128, Token Retention Ratio=22.2%2026.05 | 57.9 | 97.1 | |
| VisionZipBackbone=LLaVA-v1.5 13B, Token Retention Budget=64, Token Retention Ratio=11.1%2026.05 | 56.2 | 93.6 |