List your company

Safety Classification on UnsafeBench (ECE, NLL)

0.061ECELlama
0.04780.13690.2260.3151May 1, 2026

Evaluation results

MethodLinks
LlamaPrompt Selection Protocol=Mean2026.050.0610.607
InternVLPrompt Selection Protocol=Mean2026.050.0860.566
MiniCPMPrompt Selection Protocol=Selected2026.050.090.641
MiniCPMPrompt Selection Protocol=Random2026.050.0970.678
Qwen30BPrompt Selection Protocol=Selected2026.050.0990.629
InternVLPrompt Selection Protocol=Selected2026.050.1170.678
MiniCPMPrompt Selection Protocol=Mean2026.050.1190.608
Qwen30BPrompt Selection Protocol=Mean2026.050.120.623
Qwen3-VL 4BPrompt Selection Protocol=Mean2026.050.1230.689
Qwen3-VL 8BPrompt Selection Protocol=Mean2026.050.1250.623
GemmaPrompt Selection Protocol=Mean2026.050.1290.654
LlamaPrompt Selection Protocol=Selected2026.050.130.696
Qwen30BPrompt Selection Protocol=Random2026.050.1920.898
InternVLPrompt Selection Protocol=Random2026.050.1950.857
Qwen3-VL 4BPrompt Selection Protocol=Selected2026.050.221.198
Qwen3-VL 4BPrompt Selection Protocol=Random2026.050.2291.463
Qwen3-VL 8BPrompt Selection Protocol=Random2026.050.2411.612
LlamaPrompt Selection Protocol=Random2026.050.2640.954
GemmaPrompt Selection Protocol=Random2026.050.2682.061
Qwen3-VL 8BPrompt Selection Protocol=Selected2026.050.3661.267
GemmaPrompt Selection Protocol=Selected2026.050.3911.765