General Visual Question Answering on GQA
61.94AccuracyVanilla
Evaluation Results
| Method | Links | |
|---|---|---|
| VanillaBackbone=LLaVA-1.5-7B, TFLOPs=100%2026.06 | 61.94 | |
| VSkip+Backbone=LLaVA-1.5-7B, TFLOPs=76%, N=202026.06 | 61.63 | |
| VanillaBackbone=Qwen3-VL-8B, TFLOPs=100%2026.06 | 61.6 | |
| APETBackbone=Qwen3-VL-8B, TFLOPs=66%2026.06 | 60.96 | |
| Op-SkipBackbone=LLaVA-1.5-7B, TFLOPs=66%, N=202026.06 | 60.84 | |
| VSkipBackbone=LLaVA-1.5-7B, TFLOPs=76%, N=202026.06 | 60.77 | |
| Op-SkipBackbone=Qwen3-VL-8B, TFLOPs=66%, N=202026.06 | 60.7 | |
| VSkipBackbone=Qwen3-VL-8B, TFLOPs=76%, N=202026.06 | 60.67 | |
| ShortVBackbone=Qwen3-VL-8B, TFLOPs=58%, N=202026.06 | 60.59 | |
| V2DropBackbone=LLaVA-1.5-7B, TFLOPs=66%2026.06 | 60.51 | |
| ShortVBackbone=LLaVA-1.5-7B, TFLOPs=55%, N=202026.06 | 60.49 | |
| APETBackbone=LLaVA-1.5-7B, TFLOPs=66%2026.06 | 60.45 | |
| VanillaBackbone=Qwen2.5-VL-7B, TFLOPs=100%2026.06 | 60.4 | |
| V2DropBackbone=Qwen3-VL-8B, TFLOPs=66%2026.06 | 60.06 | |
| APETBackbone=Qwen2.5-VL-7B, TFLOPs=55%2026.06 | 59.38 | |
| Op-SkipBackbone=Qwen2.5-VL-7B, TFLOPs=55%, N=202026.06 | 59.29 | |
| V2DropBackbone=Qwen2.5-VL-7B, TFLOPs=55%2026.06 | 58.87 | |
| AVLM-2BModel Scale=2B2026.06 | 58.4 | |
| VSkipBackbone=Qwen2.5-VL-7B, TFLOPs=67%, N=162026.06 | 57.26 | |
| VTWBackbone=LLaVA-1.5-7B, TFLOPs=55%, K=162026.06 | 54.77 | |
| ShortVBackbone=Qwen2.5-VL-7B, TFLOPs=50%, N=202026.06 | 49.47 | |
| VTWBackbone=Qwen2.5-VL-7B, TFLOPs=50%, K=162026.06 | 46.01 | |
| GenLIPArch=g/16, Data=8.0B2026.05 | 45.5 | |
| SigLIP2Arch=g/16, Data=40.0B2026.05 | 45.2 | |
| GenLIPArch=So/16, Data=8.0B2026.05 | 44 | |
| AIMv2Arch=L/14, Data=12.0B2026.05 | 43.9 | |
| VTWBackbone=Qwen3-VL-8B, TFLOPs=58%, K=162026.06 | 43.72 | |
| SigLIP2Arch=So/16, Data=40.0B2026.05 | 43.5 | |
| OVision2Arch=L/16, Data=12.8B2026.05 | 42.7 | |
| SigLIP2Arch=L/16, Data=40.0B2026.05 | 42.6 | |
| SigLIPArch=L/16, Data=40.0B2026.05 | 41.7 | |
| GenLIPArch=L/16, Data=8.0B2026.05 | 41.5 | |
| CLIPArch=L/14, Data=12.8B2026.05 | 39.8 | |
| Gemma-4-E2B-itModel Scale=4B2026.06 | 39.51 | |
| Qwen2.5-Omni-3BModel Scale=3B2026.06 | 26.46 |