Reinforcement Learning is hard..

FireCrawl: https://firecrawl.link/caleb-writes-code Why RL doesn't scale and is difficult to set up. Reinforcement Learning has been helping LLMs improve capabilities in post training beyond what we could expect from these systems. And it is growingly becoming more important in robotics, gaming and simulation but where are the bottlenecks and challenges that still exist? What about RL makes it so tricky and what do we need to think about when training the model in RL? #reinforcementlearning #machinelearning #ai Follow me: X: https://x.com/calebfoundry LinkedIn: https://www.linkedin.com/in/calebeom/ TikTok: https://www.tiktok.com/@calebwritescode Chapters 00:00 Intro 00:52 Explore vs Exploit 02:45 Data Scarcity 04:49 Sponsor: FireCrawl 05:49 On vs Off-Policy 07:15 Reward 08:55 Budget 09:38 Challenges 10:22 Trend