Fine-Grained Visual Categorization on Flowers-102
89.9AccuracyFinedefics-8B
Evaluation results
| Method | Links | |
|---|---|---|
| Finedefics-8BFT Methods=SFT, FT Types=Fully-FT2026.01 | 89.9 | |
| ReFine-RFT-AOFT Methods=RFT, FT Types=Lora2026.01 | 81.4 | |
| ReFine-RFT-CoTFT Methods=RFT, FT Types=Lora2026.01 | 81 | |
| SFT-AOFT Methods=SFT, FT Types=Lora2026.01 | 74.8 | |
| SFT-CoTFT Methods=SFT, FT Types=Lora2026.01 | 74.4 | |
| Visual-RFTFT Methods=RFT, FT Types=Lora2026.01 | 74.1 | |
| Claude 3.5 sonnetOptimization strategy=false2025.12 | 73.53 | |
| Visual-RFTFT Methods=RFT, FT Types=Fully-FT2026.01 | 71.4 | |
| No-Thinking-RFTFT Methods=RFT, FT Types=Fully-FT2026.01 | 71.2 | |
| GPT-4oOptimization strategy=false2025.12 | 70.98 | |
| Gemini 1.5 proOptimization strategy=false2025.12 | 65.1 | |
| Doubao 1.5 vision proOptimization strategy=false2025.12 | 64.51 | |
| Step 1vOptimization strategy=false2025.12 | 59.22 | |
| SFT-AOFT Methods=SFT, FT Types=Fully-FT2026.01 | 58.5 | |
| InternVL 2.5-8BOptimization strategy=true2025.12 | 58.43 | |
| GLM v plusOptimization strategy=false2025.12 | 56.67 | |
| Qwen2-VL-2BFT Methods=Zero-shot, FT Types=-2026.01 | 54.8 | |
| Qwen-VL-chat-78BOptimization strategy=false2025.12 | 53.73 | |
| Hunyuan visionOptimization strategy=false2025.12 | 46.67 | |
| InternVL 2.5-8BOptimization strategy=false2025.12 | 30.98 | |
| LLaVA 1.5-7BOptimization strategy=true2025.12 | 29.8 | |
| LLaVA 1.5-7BOptimization strategy=false2025.12 | 19.02 |