Share your thoughts, 1 month free Claude Pro on usSee more

Pairwise Comparison on DeepfakeJudge Meta

96.2Pairwise Accuracy

DeepfakeJudge-7B

Updated 1mo ago

Evaluation Results

Method	Links
DeepfakeJudge-7B 2026.02		96.2
DeepfakeJudge-3B 2026.02		94.4
Qwen-3-VL-235B-Instruct 2026.02		93.2
Qwen-3-VL-30B-Thinking 2026.02		92.5
Gemini-Flash-2.5 2026.02		91.7
Qwen-3-VL-30B-Instruct 2026.02		91.3
Qwen-3-VL-235B-Thinking 2026.02		90.8
GPT-4o-Mini 2026.02		90.3
Qwen-3-VL-8B-Thinking 2026.02		89.2
Qwen-3-VL-8B-Instruct 2026.02		86
Qwen-3-VL-4B-Instruct 2026.02		75.8
Qwen-3-VL-2B-Instruct 2026.02		74.8