List your company

LLM Red-teaming on GFN-defended Target Model

0.33Unsuccessful Attack Rate (UA)PPO
-1.3910.2221.8333.44May 1, 2026

Evaluation results

MethodLinks
PPO2026.050.330.03
SFT2026.050.670.07
PPO + Curiosity2026.0510.1
DPO2026.0520.2
Rainbow Teaming2026.053.330.33
GFN2026.0554.69
ICL2026.055.330.52
Jailbreak R1CoT=enabled2026.0530.332.96
S-GFN2026.0543.3322.53