List your company

Multi-image Content Safety Classification on MIIT-dataset (OOD)

87.7F1 (Unsafe)Claude Opus 4.6
49.94859.74969.5579.351Jul 1, 2026

Evaluation results

MethodLinks
Claude Opus 4.62026.0787.776.380.4
MiShield-8B2026.0787.485.986.7
Gemini 3.1 Pro2026.0783.980.782.4
GPT-5.42026.0783.585.884.7
InternVL3-8B2026.0772.469.671.1
InternVL3.5-8B2026.0754.773.966.9
Qwen3-VL-8B-Instruct2026.0751.47663.5
Qwen3.5-9B2026.0751.47663.5