))
Illustrating Reinforcement Learning from Human Feedback (RLHF)
Manage episode 429711881 series 3498845
Innhold levert av BlueDot Impact. Alt podcastinnhold, inkludert episoder, grafikk og podcastbeskrivelser, lastes opp og leveres direkte av BlueDot Impact eller deres podcastplattformpartner. Hvis du tror at noen bruker det opphavsrettsbeskyttede verket ditt uten din tillatelse, kan du følge prosessen skissert her https://no.player.fm/legal.
This more technical article explains the motivations for a system like RLHF, and adds additional concrete details as to how the RLHF approach is applied to neural networks.
While reading, consider which parts of the technical implementation correspond to the 'values coach' and 'coherence coach' from the previous video.
A podcast by BlueDot Impact.
Learn more on the AI Safety Fundamentals website.
Kapitler
1. Illustrating Reinforcement Learning from Human Feedback (RLHF) (00:00:00)
2. RLHF: Let’s take it step by step (00:03:16)
3. Pretraining language models (00:03:51)
4. Reward model training (00:05:46)
5. Fine-tuning with RL (00:09:26)
6. Open-source tools for RLHF (00:16:10)
7. What’s next for RLHF? (00:18:20)
8. Further reading (00:21:17)
83 episoder
Del
