[qumpis]

I haven't seen RL with decision trees! it sounds really interesting. Any classic results worth looking into?

[markisus]

This is a relatively recent paper you might find interesting. https://arxiv.org/abs/2204.03771

A random forest is also a universal function approximator so anything a neural net can do, so can a random forest (in theory). In practice, neural nets are easier for modern hardware to optimize while I think trees incur computational overhead due to branchiness.