Reasoning Evaluation on DeepfakeJudge Reason 1.0 (test)
9BLEU-1Qwen-3-VL-30B-Instruct
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Qwen-3-VL-30B-InstructType=Open2026.02 | 9 | 3 | 3 | 36 | 6 | 20 | 18 | 62 | 3.31 | |
| InternVL3.5-GPT-OSS-20B-A4BType=Open2026.02 | 8 | 3 | 3 | 34 | 6 | 20 | 17 | 60 | 2.79 | |
| Qwen-3-VL-2B-InstructType=Open2026.02 | 7 | 2 | 2 | 31 | 4 | 18 | 15 | 59 | 2.36 | |
| Microsoft-Phi-4-Multimodal-InstructType=Open2026.02 | 6 | 2 | 2 | 30 | 6 | 18 | 12 | 60 | 2.82 | |
| Gemini-2.5-FlashType=Closed2026.02 | 5 | 2 | 2 | 30 | 5 | 17 | 17 | 60 | 3.17 | |
| InternVL3.5-1B-HFType=Open2026.02 | 5 | 2 | 2 | 27 | 5 | 17 | 15 | 56 | 2.44 | |
| Google-Gemma-3-12BType=Open2026.02 | 5 | 2 | 2 | 29 | 5 | 18 | 12 | 60 | 2.7 | |
| Qwen2.5-VL-Gen-BusterType=DF2026.02 | 5 | 2 | 2 | 26 | 3 | 15 | 15 | 57 | 2.33 | |
| Qwen-3-VL-235B-InstructType=Open2026.02 | 4 | 1 | 1 | 30 | 4 | 17 | 16 | 60 | 3.59 | |
| Qwen-3-VL-30B-ThinkingType=Thinking2026.02 | 3 | 1 | 1 | 26 | 3 | 15 | 15 | 59 | 3.21 | |
| Qwen-3-VL-235B-ThinkingType=Thinking2026.02 | 3 | 1 | 1 | 27 | 4 | 16 | 15 | 60 | 3.43 | |
| Qwen-3-VL-8B-ThinkingType=Thinking2026.02 | 2 | 1 | 1 | 25 | 3 | 14 | 13 | 58 | 2.81 | |
| ChatGPT-4o-miniType=Closed2026.02 | 1 | 1 | 1 | 15 | 1 | 8 | 5 | 35 | 2.83 | |
| Qwen-3-VL-4B-InstructType=Open2026.02 | 1 | 1 | 1 | 20 | 1 | 11 | 12 | 56 | 2.93 | |
| Qwen-3-VL-8B-InstructType=Open2026.02 | 1 | 1 | 1 | 16 | 1 | 10 | 9 | 53 | 2.51 | |
| SIDA-13B-DescriptionType=DF2026.02 | 1 | 1 | 1 | 24 | 3 | 16 | 15 | 58 | 2.32 |