Wound Infection Classification on wound infection (test)
86.8Accuracy (ACC)Infection-Reasoner-4B [DFT + GRPO]
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Infection-Reasoner-4B [DFT + GRPO]Model category=Fine-tuned Models, Parameter scale=4B, Evaluation protocol=DFT + GRPO2026.04 | 86.8 | 86.4 | 87.1 | 87.1 | 86.6 | 88.1 | |
| GPT-5.1 (Teacher)Model category=Proprietary Models2026.04 | 81.8 | 80 | 83.6 | 83.1 | 80.7 | 81.5 | |
| Infection-Reasoner-4B [DFT Only]Model category=Fine-tuned Models, Parameter scale=4B, Evaluation protocol=DFT Only2026.04 | 77.1 | 73.6 | 80.7 | 79.3 | 75.3 | 76.3 | |
| SCARWID (Image-Text input)Model category=Traditional Deep Learning Classifiers, Input modality=Image-Text input (GPT-4o captions)2026.04 | 76.8 | 75 | 78.6 | 77.8 | 75.9 | 76.3 | |
| Gemini-3.1-ProModel category=Proprietary Models2026.04 | 75.4 | 86.4 | 64.3 | 70.7 | 82.7 | 77.8 | |
| Infection-Reasoner-4B [GRPO Only]Model category=Fine-tuned Models, Parameter scale=4B, Evaluation protocol=GRPO Only2026.04 | 74.3 | 88.6 | 60 | 69 | 84 | 77.5 | |
| Gemini-3-FlashModel category=Proprietary Models2026.04 | 72.1 | 86.4 | 57.9 | 67.3 | 81.1 | 75.6 | |
| GPT-5.2Model category=Proprietary Models2026.04 | 69.6 | 59.3 | 80 | 74.8 | 66.3 | 66.1 | |
| Qwen3.5-27BModel category=Open-source Models, Parameter scale=> 10B2026.04 | 67.1 | 85.7 | 48.6 | 62.5 | 77.3 | 72.3 | |
| EfficientNet-B0Model category=Traditional Deep Learning Classifiers2026.04 | 62.5 | 46.4 | 78.6 | 68.4 | 59.5 | 55.3 | |
| DeiT-BaseModel category=Traditional Deep Learning Classifiers2026.04 | 62.5 | 53.6 | 71.4 | 65.2 | 60.6 | 58.8 | |
| Llama-3.2-90B-Vision-InstructModel category=Open-source Models, Parameter scale=> 10B2026.04 | 61.1 | 96.4 | 25.7 | 56.5 | 87.7 | 71.3 | |
| Qwen3.5-4BModel category=Open-source Models, Parameter scale=< 10B2026.04 | 60.7 | 100 | 21.4 | 56 | 100 | 71.8 | |
| Qwen3-VL-32B-ThinkingModel category=Open-source Models, Parameter scale=> 10B2026.04 | 60.7 | 91.4 | 30 | 56.6 | 77.9 | 69.9 | |
| Claude-Sonnet-4.6Model category=Proprietary Models2026.04 | 56.4 | 100 | 12.9 | 53.5 | 100 | 69.7 | |
| Lingshu-32B [Medical]Model category=Open-source Models, Parameter scale=> 10B2026.04 | 56.4 | 95.7 | 17.1 | 53.6 | 79.8 | 68.7 | |
| HuatuoGPT-V-34B [Medical]Model category=Open-source Models, Parameter scale=> 10B2026.04 | 55.7 | 100 | 11.4 | 53 | 100 | 69.3 | |
| MedVLM-R1-2B [Medical]Model category=Open-source Models, Parameter scale=< 10B2026.04 | 53.6 | 96.4 | 10.7 | 51.9 | 75 | 67.5 | |
| Qwen3-VL-4B-ThinkingModel category=Open-source Models, Parameter scale=< 10B2026.04 | 52.5 | 96.4 | 8.6 | 51.4 | 67.7 | 67 | |
| Lingshu-7B [Medical]Model category=Open-source Models, Parameter scale=< 10B2026.04 | 52.1 | 97.9 | 6.4 | 51.1 | 76.7 | 67.2 | |
| MedGemma-4B [Medical]Model category=Open-source Models, Parameter scale=< 10B2026.04 | 50 | 92.9 | 7.1 | 50 | 50 | 65 |