XNews

First Systematic Book on RLHF Covers DPO, RLVR, and Synthetic Data

A post announces the first systematic book on RLHF, which follows the main line of 'instruction fine-tuning → reward model → reinforcement learning'. It covers topics including DPO, RLVR, synthetic data, over-optimization, and model personality shaping. The book is available to read online, and its project code is open-sourced.

Source posts