Tomek Korbak comments on RL with KL penalties is better seen as Bayesian inference

[Tomek Korbak]

good catch, yes, thanks!