about-this-book
about this book
This book is a comprehensive overview of what RLHF is, how it is done, and how it fits into the broader methods of modern post-training. RLHF is both complicated to implement and intensely interdisciplinary. The book is a horizontal cut through all the fields and pieces that go into RLHF: reinforcement learning, theory, code, philosophy, data, UX, product, and so on, with a few key chapters that go into great detail on the core methods—the places where complexity is part of the process.