List your company

Binary Classification on Unsafe Bench (test)

90.28Macro-F1SFT-Single
72.0876.80581.5386.255Mar 12, 2026

Evaluation results

MethodLinks
SFT-Singletraining=Single-task SFT2026.0390.28
SFT-Unifiedtraining=Unified multi-task SFT2026.0389.49
RL-Singletraining=Single-task RL2026.0386.92
MT-RL-Judgetraining=Multi-task RL2026.0385.22
Off-the-shelf2026.0372.78