Image-to-text subject generation on VLM
99.5AUCFor-Value
Evaluation results
| Method | Links | ||
|---|---|---|---|
| For-ValueVLM Model=Llama-3.2-11B-vision2025.08 | 99.5 | 98.5 | |
| For-ValueVLM Model=Qwen2.5-VL-3B-Instruct2025.08 | 99.4 | 89.7 | |
| HyperINFVLM Model=Llama-3.2-11B-vision2025.08 | 99.3 | 91.9 | |
| DataInfVLM Model=Qwen2.5-VL-3B-Instruct2025.08 | 98.9 | 83.6 | |
| HyperINFVLM Model=Qwen2.5-VL-3B-Instruct2025.08 | 98.8 | 90.2 | |
| Hessian-freeVLM Model=Qwen2.5-VL-3B-Instruct2025.08 | 97.9 | 73.8 | |
| Hessian-freeVLM Model=Llama-3.2-11B-vision2025.08 | 96.1 | 76.5 | |
| DataInfVLM Model=Llama-3.2-11B-vision2025.08 | 95.8 | 79.7 | |
| EmbVLM Model=Qwen2.5-VL-3B-Instruct2025.08 | 84.1 | 20.6 | |
| EmbVLM Model=Llama-3.2-11B-vision2025.08 | 84.1 | 20.6 |