We can then develop practical risk-averse algorithms with convergence guarantees. Empirically, our method learns CVaR-sensitive policies that can shrink the lower tail of the return distribution,…