General Visual Question Answering on MMBench En (dev)
91.5Overall ScoreKeye-VL
Evaluation results
| Method | Links | |
|---|---|---|
| Keye-VLOpenness=semi-open, Model Size=8B2025.10 | 91.5 | |
| Bee-8BOpenness=fully open, Model Size=8B, Training Strategy=RL2025.10 | 85.5 | |
| Bee-8BOpenness=fully open, Model Size=8B, Training Strategy=SFT2025.10 | 83 | |
| Qwen2.5-VLOpenness=semi-open, Model Size=7B2025.10 | 82.1 | |
| LLaVA OneVisionOpenness=fully open, Model Size=7B2025.10 | 80.8 | |
| Bunny-v1.1-4BBackbone=Phi-3-3.8B, Param.=4B, Train Data=2.7M2026.02 | 74.2 | |
| Imp-4BBackbone=Phi-3-3.8B, Param.=4B, Train Data=1.6M2026.02 | 73.3 | |
| HIVEBackbone=Huginn, Param.=4B, Train Data=6.5M, Recurrence Level (r)=322026.02 | 69.6 | |
| MobileVLM V2 7BBackbone=Vicuna-7B-v1.5, Param.=7B, Train Data=3.6M2026.02 | 69.2 | |
| Emu3Param.=8B2026.02 | 58.5 |