General Visual Question Answering on VQA v2 (Accuracy)
83.5AccuracyFP16 (Baseline)
Evaluation Results
| Method | Links | |
|---|---|---|
| FP16 (Baseline)Backbone=Qwen 2.5 VL 7B, Avg. Bit width=162025.09 | 83.5 | |
| OmniQuantBackbone=Qwen 2.5 VL 7B, Avg. Bit width=42025.09 | 82.9 | |
| AWQBackbone=Qwen 2.5 VL 7B, Avg. Bit width=42025.09 | 82.7 | |
| GPTQBackbone=Qwen 2.5 VL 7B, Avg. Bit width=42025.09 | 82.6 | |
| OmniQuantBackbone=Qwen 2.5 VL 7B, Avg. Bit width=32025.09 | 81.6 | |
| GPTQBackbone=Qwen 2.5 VL 7B, Avg. Bit width=32025.09 | 81.5 | |
| LUQBackbone=Qwen 2.5 VL 7B, Avg. Bit width=2.75, Quantized Layers=12 layer2025.09 | 79.7 | |
| FP16 (Baseline)Backbone=LLaVA-1.5 7B, Avg. Bit width=162025.09 | 78.5 | |
| SlimLLMBackbone=LLaVA-1.5 7B, Avg. Bit width=42025.09 | 76.8 | |
| AWQBackbone=LLaVA-1.5 7B, Avg. Bit width=42025.09 | 76.6 | |
| OmniQuantBackbone=LLaVA-1.5 7B, Avg. Bit width=42025.09 | 76.6 | |
| GPTQBackbone=LLaVA-1.5 7B, Avg. Bit width=42025.09 | 76.3 | |
| LUQBackbone=LLaVA-1.5 7B, Avg. Bit width=2.54, Quantized Layers=16 layer2025.09 | 74.9 | |
| SlimLLMBackbone=LLaVA-1.5 7B, Avg. Bit width=32025.09 | 74.2 | |
| GPTQBackbone=LLaVA-1.5 7B, Avg. Bit width=32025.09 | 73.5 | |
| OmniQuantBackbone=LLaVA-1.5 7B, Avg. Bit width=32025.09 | 72.9 | |
| SigLIP2Arch=g/16, Data=40.0B2026.05 | 50.1 | |
| GenLIPArch=g/16, Data=8.0B2026.05 | 49.1 | |
| AIMv2Arch=L/14, Data=12.0B2026.05 | 48.1 | |
| SigLIP2Arch=So/16, Data=40.0B2026.05 | 46.5 | |
| GenLIPArch=So/16, Data=8.0B2026.05 | 46.5 | |
| CLIPArch=L/14, Data=12.8B2026.05 | 46.1 | |
| GenLIPArch=L/16, Data=8.0B2026.05 | 44.4 | |
| OVision2Arch=L/16, Data=12.8B2026.05 | 44 | |
| SigLIP2Arch=L/16, Data=40.0B2026.05 | 43.1 | |
| SigLIPArch=L/16, Data=40.0B2026.05 | 41.3 | |
| BiLLMBackbone=Qwen 2.5 VL 7B, Avg. Bit width=1.082025.09 | 39.5 | |
| BiLLMBackbone=LLaVA-1.5 7B, Avg. Bit width=1.082025.09 | 37.2 |