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

Agent Safety on ASSEBench

92.04Accuracy

DRAFT (Qwen3Guard-Gen-4B)

48.46459.77771.0982.403Feb 11, 2026Mar 1, 2026Mar 19, 2026Apr 6, 2026Apr 24, 2026May 12, 2026May 30, 2026
Updated 1d ago

Evaluation Results

MethodLinks
2026.02
92.0490.55-88.77
2026.02
91.5789.75-87.19
2026.05
91.591.44--
2026.02
91.3888.98-86.04
2026.02
90.8188.0996.0681.33
2026.02
89.9787.2393.1882
2026.02
89.7286.91-84.62
2026.02
89.6986.2597.4877.33
2026.05
89.6390.09--
2026.05
89.1288.37--
2026.05
89.0289.44--
2026.02
88.8686.0190.4482
2026.05
88.6687.81--
2026.05
87.9986.95--
2026.02
87.1983.4590.6277.33
2026.05
85.6384.73--
2026.05
85.1985.7--
2026.02
84.480.6983.5778.01
2026.02
84.2277.06-68.9
2026.02
84.1277.9992.6667.33
2026.02
83.0177.1588.0368.67
2026.05
82.3382.92--
2026.05
81.5380.53--
2026.02
81.0974.15-68.34
2026.02
81.0671.4396.5956.67
2026.02
80.8281.84-79.33
2026.02
80.1772.27-64.45
2026.05
79.6978.17--
2026.02
79.3970.1688.7858
2026.05
79.3776.63--
2026.05
79.3181.08--
2026.02
78.8369.687.0658.03
2026.02
77.9979.41-70.67
2026.05
77.5874.6--
2026.02
77.1667.284.0456.02
2026.02
76.5663.41-57.17
2026.05
76.378.44--
2026.02
74.6771.6-62.56
2026.02
74.1972.44-67.01
2026.02
73.5459.2383.1346
2026.05
72.7465.6--
2026.05
72.1969--
2026.02
71.5952.3487.537.33
2026.02
71.3152.0986.1537.33
2026.02
71.0658.92-60.3
2026.05
70.8174.9--
2026.02
69.5267.85-60.51
2026.05
68.5474.62--
2026.05
65.5168.37--
2026.02
65.1839.02-26.67
2026.02
64.7657.91-57.33
2026.02
63.7965.24-81.33
2026.02
63.2344.57-41.09
2026.02
62.6434.36-23.66
2026.02
61.5525.69-16.08
2026.02
60.4539.8354.6531.33
2026.02
59.6118.0859.2610.67
2026.02
59.3329.81-20.67
2026.02
58.9635.9850.6427.91
2026.02
58.814.4150.928.39
2026.02
58.6949.87-49.85
2026.05
57.4156.16--
2026.02
56.8259.9548.9577.33
2026.02
54.257.4946.6374.92
2026.02
54.1255.9446.3770.48
2026.02
51.2556.3645.0275.33
2026.05
51.0265.19--
2026.02
50.5650.4443.0660.87
2026.02
50.1450.4243.1360.67