huggingface/alignment-handbook

Aligning Large Language Models with Human: A Survey

AllenAI's post-training codebase

Post-training with Tinker

Reference implementation for DPO (Direct Preference Optimization)

[NeurIPS 2024] SimPO: Simple Preference Optimization with a Reference-Free Reward

Train transformer language models with reinforcement learning.

A course on aligning smol models.

Xwin-LM: Powerful, Stable, and Reproducible LLM Alignment

Code and documentation to train Stanford's Alpaca models, and generate the data.