List your company

Safety Evaluation on PKU-SafeRLHF-V

77.8AccuracyDebate about images
67.60870.25472.975.546Nov 29, 2025

Evaluation results

MethodLinks
Debate about imagesModel=GPT-4o2025.1177.80.568
Debate with imagesModel=Claude-Sonnet-42025.1176.80.5312
Debate about imagesModel=Qwen2.5-VL-72B2025.1176.30.5313
Debate with imagesModel=GPT-4o2025.1176.20.5514
Debate with imagesModel=Gemini-2.5-Pro2025.1174.90.525
Debate with imagesModel=Qwen2.5-VL-72B2025.1174.80.522
CoT promptModel=Qwen2.5-VL-72B2025.1174.70.4915
Debate about imagesModel=Claude-Sonnet-42025.1174.50.5813
Direct promptModel=GPT-4o2025.1174.40.511
Majority voteModel=Qwen2.5-VL-72B2025.1173.90.4813
Majority voteModel=Claude-Sonnet-42025.1173.30.4712
CoT promptModel=Claude-Sonnet-42025.1172.90.4612
Majority voteModel=GPT-4o2025.1172.30.4910
Debate about imagesModel=Gemini-2.5-Pro2025.1172.20.4526
Direct promptModel=Claude-Sonnet-42025.1172.10.4413
Direct promptModel=Qwen2.5-VL-72B2025.11720.4415
CoT promptModel=GPT-4o2025.1170.10.4513
Majority voteModel=Gemini-2.5-Pro2025.1169.90.4413
CoT promptModel=Gemini-2.5-Pro2025.1168.40.4116
Direct promptModel=Gemini-2.5-Pro2025.11680.3725