A not-so-surprising finding: with asynchronous RL, agents learn faster and achieve significantly higher returns. Still, it's nice to see it empirically verified.

→ View original post on X — @jiqizhixin