Share your thoughts, 1 month free Claude Pro on usSee more

PointGoal on Safety Gym

0.64Normalized Reward

Reference policy (πref)

Updated 13d ago

Evaluation Results

Method	Links
Reference policy (πref) 2026.05		0.64	2.82
PPL 2026.05		0.46	0.95
PREFINE 2026.05		0.4	0.24
CPQ 2026.05		0.25	0.2
SafeDICE 2026.05		0.24	0.26
BC 2026.05		0.22	0.82