List your company

Workflow Steering Defense on ASB-Bench

68TASRG-Safeguard
14.9628.7342.556.27May 12, 2026

Evaluation results

MethodLinks
G-SafeguardModel=Gemini-2.5-Flash2026.056871
ARGUSModel=Qwen-3.5-Flash2026.055359
w/o DefenseModel=GPT-4o-mini2026.055259
w/o DefenseModel=Gemini-2.5-Flash2026.055163
G-SafeguardModel=Qwen-3.5-Flash2026.055064
w/o DefenseModel=Qwen-3.5-Flash2026.054962
G-SafeguardModel=GPT-4o-mini2026.054459
ARGUSModel=DeepSeek-V32026.054164
G-SafeguardModel=DeepSeek-V32026.054153
w/o DefenseModel=DeepSeek-V32026.054054
ARGUSModel=GPT-4o-mini2026.053647
FLOWGUARDModel=Gemini-2.5-Flash2026.053434
FLOWGUARDModel=Qwen-3.5-Flash2026.053346
FLOWGUARDModel=GPT-4o-mini2026.052227
FLOWGUARDModel=DeepSeek-V32026.052230
ARGUSModel=Gemini-2.5-Flash2026.051722