Rohin Shah
TalkRL: The Reinforcement Learning Podcast
The Top Two Approaches
Team cairos used 80 thousand plus labelled images and built some very specific components for this. Team obsidian produced this inverse cue learning method which has seemed like more general, theoretical solution. Even even the top tame did rely on a behavior cloned navigation policy that used the neral network. It shows you if you're just actually trying to get good performance, do you put inor or put in domain knowledge? And how much domain knowledge do you putting in? And a how do you do it?
00:00
Transcript
Play full episode
Remember Everything You Learn from Podcasts
Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.
