AI-generated Content Detection on FakeClue in-distribution
98.6AccuracyIFM-AIGCSPOTTER-3B
Evaluation results
| Method | Links | ||
|---|---|---|---|
| IFM-AIGCSPOTTER-3BTraining Protocol=Detection-only training2026.04 | 98.6 | 98.9 | |
| FakeVLM (11b)Training Protocol=Detection + Explanation training2026.04 | 98.6 | 98.1 | |
| IFM-AIGCSPOTTER-3BTraining Protocol=Detection + Explanation training2026.04 | 97.1 | 97.7 | |
| Llama3.2-11bTraining Protocol=Detection-only training2026.04 | 95.6 | 96.6 | |
| Llama3.2-11bTraining Protocol=Detection + Explanation training2026.04 | 95.1 | 96.2 | |
| Qwen2.5-VL-3bTraining Protocol=Detection-only training2026.04 | 95 | 96.2 | |
| Gemma3-4b-itTraining Protocol=Detection-only training2026.04 | 94.6 | 95.8 | |
| Gemma3-4bTraining Protocol=Detection + Explanation training2026.04 | 93.7 | 95.1 | |
| Qwen2.5-VL-3bTraining Protocol=Detection + Explanation training2026.04 | 93.2 | 94.8 | |
| Qwen2-VL-72BTraining Protocol=Prompting only2026.04 | 57.8 | 56.5 | |
| InternVL2-40BTraining Protocol=Prompting only2026.04 | 50.7 | 46.3 | |
| InternVL2-8BTraining Protocol=Prompting only2026.04 | 50.6 | 49 | |
| Deepseek-VL2Training Protocol=Prompting only2026.04 | 47.5 | 54.1 | |
| GPT-4oTraining Protocol=Prompting only2026.04 | 47.4 | 42 | |
| Qwen2-VL-7BTraining Protocol=Prompting only2026.04 | 45.7 | 59.2 | |
| Deepseek-VL2-smallTraining Protocol=Prompting only2026.04 | 40.4 | 54.2 |