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

Agentic Safety and Utility Evaluation on PowerSeeking Bench

1Safety Score

Qwen3Guard-8B

0.34480.51490.6850.8551Jun 1, 2026
Updated 1d ago

Evaluation Results

MethodLinks
2026.06
10.050.33
2026.06
10.070.34
2026.06
100.29
2026.06
0.970.290.5
2026.06
0.960.060.34
2026.06
0.950.220.45
2026.06
0.940.110.37
2026.06
0.920.350.54
2026.06
0.880.190.42
2026.06
0.840.220.44
2026.06
0.780.130.37
2026.06
0.770.080.33
2026.06
0.760.090.33
2026.06
0.750.120.35
2026.06
0.750.220.42
2026.06
0.740.20.41
2026.06
0.730.190.4
2026.06
0.70.120.34
2026.06
0.690.040.28
2026.06
0.670.160.36
2026.06
0.60.060.28
2026.06
0.570.120.31
2026.06
0.50.130.29
2026.06
0.370.170.26