Visual Reasoning on CLEVR VL-ICL Bench
38.9Average AccuracyLLaVA-OneVision-7B
Evaluation results
| Method | Links | |
|---|---|---|
| LLaVA-OneVision-7BVL-Data=10.4M, Params trained=8B, TTA=FT≤302025.06 | 38.9 | |
| LLaVA-OneVision-72BVL-Data=10.4M, Params trained=73.2B, TTA=ICL2025.06 | 37.2 | |
| LLaVA-OneVision-7BVL-Data=10.4M, Params trained=8B, TTA=ICL2025.06 | 34.9 | |
| LAQ-7B + MAPDVL-Data=1.3M, Params trained=24M, TTA=FT≤302025.06 | 31.4 | |
| LAQ-7B + In-ContextPDVL-Data=1.3M, Params trained=24M, TTA=FT≤302025.06 | 30.9 | |
| Qwen2.5-VL-7B-InstructVL-Data=-NA-, Params trained=8B, TTA=FT≤302025.06 | 29.1 | |
| LLaVA-Next-7BVL-Data=1.3M, Params trained=7.06B, TTA=FT≤302025.06 | 28.6 | |
| Qwen2-VL-7B-InstructVL-Data=-NA-, Params trained=8B, TTA=ICL2025.06 | 27.9 | |
| LAQ-7B + In-ContextPDVL-Data=1.3M, Params trained=24M, TTA=ICL2025.06 | 24.1 | |
| Qwen2.5-VL-7B-InstructVL-Data=-NA-, Params trained=8B, TTA=ICL2025.06 | 22 | |
| LLaVA-Next-7BVL-Data=1.3M, Params trained=7.06B, TTA=ICL2025.06 | 21.1 | |
| LLaVA v1.5-7BVL-Data=1.2M, Params trained=7B, TTA=FT≤302025.06 | 16.9 | |
| LAQ-7B + MAPDVL-Data=1.3M, Params trained=24M, TTA=ICL2025.06 | 12.3 | |
| LLaVA v1.5-7BVL-Data=1.2M, Params trained=7B, TTA=ICL2025.06 | 10.9 |