List your company

Safety discovery in text-based scenarios on Deploy Pipeline Off Switch

47REPO-Safe
44.9245.464646.54Apr 25, 2026

Evaluation results

MethodLinks
EPO-SafeModel=Claude Sonnet 4.62026.0447470
CoTModel=Claude Sonnet 4.62026.0447475
StaticModel=Claude Sonnet 4.62026.0447475
Reward-OnlyModel=Claude Sonnet 4.62026.0446460
Reward-OnlyModel=Gemini 3 Flash2026.0446465
CoTModel=Gemini 3 Flash2026.0446465
StaticModel=Gemini 3 Flash2026.0446465
EPO-SafeModel=Gemini 3 Flash2026.0445450