Website Accessibility Correction on WebAccessVL (test)
0.439ViolationsLlama 3.2 Vision
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Llama 3.2 VisionVLM=true, Fine-tuned=true, Training=Violation-conditioned2025.12 | 0.439 | 0.26 | 33.1 | 3.021 | |
| LLaVA 1.6VLM=true, Fine-tuned=true, Training=Violation-conditioned2025.12 | 0.509 | 0.242 | 30 | 3.053 | |
| GPT-5VLM=false, Fine-tuned=false2025.12 | 1.681 | 0.144 | 0.5 | 15.573 | |
| Claude3.5 SonnetVLM=false, Fine-tuned=false2025.12 | 2.31 | 0.223 | 18.3 | 7.603 | |
| GPT-4o VisionVLM=true, Fine-tuned=false2025.12 | 2.918 | 0.25 | 40.8 | 4.939 | |
| GPT-4oVLM=false, Fine-tuned=false2025.12 | 2.929 | 0.227 | 38.9 | 5.081 | |
| Qwen 2.5 VL 7BVLM=true, Fine-tuned=false2025.12 | 3.77 | 0.217 | 1.9 | 3.983 | |
| Llama 3.2 3BVLM=false, Fine-tuned=false2025.12 | 4.382 | 0.173 | 2 | 14.097 | |
| Qwen 2.5 Coder 7BVLM=false, Fine-tuned=false2025.12 | 4.466 | 0.167 | 2.2 | 4.119 | |
| Llama 3.2 Vision 11BVLM=true, Fine-tuned=false2025.12 | 4.765 | 0.165 | 1.9 | 15.474 | |
| LLaVA 1.5VLM=true, Fine-tuned=false2025.12 | 4.944 | 0.156 | 2.1 | 4.443 | |
| Qwen 2 7BVLM=false, Fine-tuned=false2025.12 | 5.173 | 0.178 | 2.2 | 1.235 | |
| LLaVA 1.6VLM=true, Fine-tuned=false2025.12 | 5.203 | 0.169 | 2.2 | 1.095 | |
| Qwen 2 VL 7BVLM=true, Fine-tuned=false2025.12 | 5.214 | 0.153 | 2.2 | 0.125 | |
| Qwen 2.5 7BVLM=false, Fine-tuned=false2025.12 | 5.303 | 0.134 | 1.9 | 0.314 | |
| Janus Pro 7BVLM=true, Fine-tuned=false2025.12 | 5.335 | 0.125 | 2.2 | 12.616 | |
| DeepSeek VL 2 smallVLM=true, Fine-tuned=false2025.12 | 5.505 | 0.149 | 2.2 | 2.09 | |
| Gemini 1.5 FlashVLM=false, Fine-tuned=false2025.12 | 6.213 | 0.229 | 28.6 | 5.762 |