Nat Lambert finishes an RLHF book with a 10-hour course and training code
natolambert · x · 2026-07-21
Nat Lambert says his book on Reinforcement Learning from Human Feedback is finished.
- He wrote it to be the guide he wished he had while learning to fine-tune, align, and post-train models after ChatGPT.
- The release includes a 10+ hour full course, slide decks, working code for the training chapters, and an example model completions library.
- A free web version is available, while printed copies from Manning should ship in 1–2 weeks and Amazon copies a bit later.
- He says the book distills lessons learned while building Olmo, and was written nights and weekends since 2024.
Related event: Nathan Lambert Completes RLHF Book After Two Years with Courses and Code(8 posts)→
More from Research
- Project APE launches CRED to test whether LLMs can verify research errors — soumitrashukla9 · 2026-07-22
- Project APE finds verifier reliability drops when papers contain multiple errors — soumitrashukla9 · 2026-07-22
- Project APE says verifier costs fell about 90x in a year as Chinese open models lead — soumitrashukla9 · 2026-07-22
- OpenAI-linked paper says capability RL can make models more reward-seeking — MariusHobbhahn · 2026-07-22
- Project APE builds its verifier benchmark from 100 AI-written papers with injected errors — soumitrashukla9 · 2026-07-22
- Paper proposes a CRED taxonomy and benchmark to measure research-error detectors — soumitrashukla9 · 2026-07-22