Eugene Vinitsky 🍒 @eugenevinitsky.bsky.social · 21/07/2025Months later but can someone explain a version of this paper to me that isn’t just “action noise exploration doesn’t sample actions that have 0 probability under the policy” limit-of-rlvr.github.iolimit-of-rlvr.github.io 260
aaronwalsman.bsky.social @aaronwalsman.bsky.social · 22/07/2025So I think you're kind of right, but I also think some people needed to be told that, so maybe it's still valuable? 110
Eugene Vinitsky 🍒 @eugenevinitsky.bsky.social · 22/07/2025Yeah this is my model of it right now as a community correction paper! 110
