Linguistic Complexity Evaluation on LIN-Bench
12.49ARILLaVA1.6+Ours
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| LLaVA1.6+OursBackbone=LLaVA1.6, Method=Patch Matters / divide-then-aggregate2025.04 | 12.49 | 11.37 | 12.98 | 12.28 | |
| Mini-Gemini+OursBackbone=Mini-Gemini, Method=Patch Matters / divide-then-aggregate2025.04 | 12.11 | 10.77 | 12.31 | 11.73 | |
| LLaVA1.5+OursBackbone=LLaVA1.5, Method=Patch Matters / divide-then-aggregate2025.04 | 11.5 | 10.42 | 12.24 | 11.38 | |
| LLaVA1.6+SynBackbone=LLaVA1.6, Training Strategy=Synthetic Data2025.04 | 10.75 | 9.85 | 11.79 | 10.8 | |
| Mini-Gemini+SynBackbone=Mini-Gemini, Training Strategy=Synthetic Data2025.04 | 10.69 | 9.57 | 11.35 | 10.53 | |
| LLaVA1.6+ITBackbone=LLaVA1.6, Training Strategy=Instruction Tuning2025.04 | 10.05 | 9.48 | 11.58 | 10.37 | |
| LLaVA1.5+SynBackbone=LLaVA1.5, Training Strategy=Synthetic Data2025.04 | 9.74 | 8.93 | 10.97 | 9.88 | |
| LLaVA1.6Backbone=LLaVA1.62025.04 | 9.71 | 9.19 | 11.41 | 10.1 | |
| Mini-Gemini+ITBackbone=Mini-Gemini, Training Strategy=Instruction Tuning2025.04 | 9.52 | 8.75 | 10.97 | 9.74 | |
| Mini-GeminiBackbone=Mini-Gemini2025.04 | 9.31 | 8.55 | 10.87 | 9.57 | |
| LLaVA1.5+ITBackbone=LLaVA1.5, Training Strategy=Instruction Tuning2025.04 | 8.85 | 8.35 | 10.86 | 9.35 | |
| LLaVA1.5Backbone=LLaVA1.52025.04 | 8.69 | 8.18 | 10.8 | 9.22 |