Hi everyone,
I’d love to get your thoughts on a question I have about agents and Reinforcement Learning.
I’ve read in several articles that Reinforcement Learning (RL) is often the most effective way to improve Agent’s capabilities, more so than Supervised Learning (SL). My previous projects didn’t involve agents, so my experience is mostly limited to supervised learning (specifically, fine-tuning with question-answer pairs).
I’m now planning a new project that will likely involve an agentic workflow. This leads to my main question: is it actually feasible for an individual or a small company to use RL to train an Agent and continuously improve its performance?
If it is, could you recommend any frameworks that are relatively easy for a newcomer to get started with?
Thanks in advance for any insights!