Multimodal Understanding on ScienceQA SQA^I
90.5AccuracyQwen3-VL + DiG
Evaluation results
| Method | Links | |
|---|---|---|
| Qwen3-VL + DiGLLM=Qwen3-8B2025.12 | 90.5 | |
| Qwen3-VLLLM=Qwen3-8B2025.12 | 90.3 | |
| Qwen3-VL + DiGLLM=Qwen3-4B2025.12 | 89.2 | |
| Qwen3-VLLLM=Qwen3-4B2025.12 | 86.9 | |
| LLaVA-NeXTLLM=Vicuna1.5-7B2025.12 | 70.1 |