Content Moderation on Dynamic Evaluation Sets Adversarial 1.0 benchmark leaderboard · SOTA2 Research

Content Moderation on Dynamic Evaluation Sets Adversarial 1.0

86.72Overall Composite ScoreYuvion-32B (Agent)

Evaluation Results

Method	Links
Yuvion-32B (Agent)2026.06		86.72	86.1	87.34
Yuvion-32B2026.06		86.06	85.78	86.34
Yuvion-8B2026.06		82.55	82.38	82.72
Qwen3-Max2026.06		82.21	81.41	83
GPT-5.42026.06		80.56	80.73	80.4
Qwen3.6-Plus2026.06		80.14	81.83	78.44
KIMI-K2.52026.06		77.23	74.61	79.84
Qwen3.5-122B-A10B2026.06		76.96	77.3	76.61
GLM-52026.06		76.9	79.84	73.96
Qwen3.5-Plus2026.06		75.65	77.92	73.38
Qwen3.5-27B2026.06		75.55	77.17	73.92
DeepSeek-R12026.06		75.2	80.54	69.85
Qwen3.5-397B-A17B2026.06		75.04	77.45	72.62
Qwen3.5-35B-A3B2026.06		74.11	75.46	72.75
Qwen3-32B2026.06		73.95	79.79	68.1
DeepSeek-V3.22026.06		73.08	76.27	69.88
Qwen3.5-9B2026.06		71.28	72.68	69.89
Qwen3-8B2026.06		70.64	73.71	67.57
Qwen3-30B-A3B-25072026.06		66.64	77.14	56.14
Minimax-M2.52026.06		64.46	72.6	56.32
Qwen3Guard-8B2026.06		0.01	0	0.01
Llama-Guard4-12B2026.06		0	0	0