List your company

Safety Classification on ASSEBench (test)

89.97AccuracyTRACE
48.983659.624370.26580.9057May 30, 2026

Evaluation results

MethodLinks
TRACEBackbone=Qwen2.5-7B-Instruct2026.0589.9787.2393.1882
MAGEBackbone=Qwen2.5-7B-Instruct2026.0588.8686.0190.4482
TRACEBackbone=Qwen3-4B-Instruct-25072026.0587.1983.4590.6277.33
TRACEBackbone=Qwen2.5-3B-Instruct2026.0584.480.6983.5778.01
MAGEBackbone=Qwen3-4B-Instruct-25072026.0583.0177.1588.0368.67
MAGEBackbone=Qwen2.5-3B-Instruct2026.0581.0671.4396.5956.67
SFTBackbone=Qwen3-4B-Instruct-25072026.0579.3970.1688.7858
SFTBackbone=Qwen2.5-7B-Instruct2026.0571.5952.3487.537.33
SFTBackbone=Qwen2.5-3B-Instruct2026.0571.3152.0986.1537.33
BaseBackbone=Qwen3-4B-Instruct-25072026.0554.257.4946.6374.92
BaseBackbone=Qwen2.5-7B-Instruct2026.0554.1255.9446.3770.48
BaseBackbone=Qwen2.5-3B-Instruct2026.0550.5650.4443.0660.87