Spatial Understanding on MME Position
141MME_Pos ScoreLLaVA_5K_SiTe-ratio
Evaluation results
| Method | Links | |
|---|---|---|
| LLaVA_5K_SiTe-ratioTraining Stage=Supervised Fine-tuning, Backbone Architecture=LLaVA-v1.5-7B, Data Augmentation Method=SiTe-ratio, SFT Sample Count=5K2025.12 | 141 | |
| LLaVASiTe-randTraining Stage=Pretraining, Backbone Architecture=LLaVA-v1.5-7B, Data Augmentation Method=SiTe-rand, Pretraining Dataset=558K2025.12 | 133.33 |