reinforcement-learning-from-human-feedback

An Easy-to-use, Scalable and High-performance Agentic RL Framework based on Ray (PPO & DAPO & REINFORCE++ & VLM & TIS & vLLM & Ray & Async RL)

Safe RLHF: Constrained Value Alignment via Safe Reinforcement Learning from Human Feedback

A simulation framework for RLHF and alternatives. Develop your RLHF method without collecting human data.

Super-Efficient RLHF Training of LLMs with Parameter Reallocation

Okapi: Instruction-tuned Large Language Models in Multiple Languages with Reinforcement Learning from Human Feedback

A Survey of Direct Preference Optimization (DPO)

CodeUltraFeedback: aligning large language models to coding preferences (TOSEM 2025)

A repo for RLHF training and BoN over LLMs, with support for reward model ensembles.

Official code for ICML 2024 paper, "RIME: Robust Preference-based Reinforcement Learning with Noisy Preferences" (ICML 2024 Spotlight)

A curated list of papers on post-training for end-to-end autonomous driving: distillation, preference alignment, reinforcement learning, and test-time refinement.

A modular, production-grade framework for Reinforcement Learning from Human Feedback (RLHF) with Proximal Policy Optimization (PPO) and Direct Preference Optimization (DPO).

annotated tutorial of the huggingface TRL repo for reinforcement learning from human feedback connecting equations from PPO and GAE to the lines of code in the pytorch implementation

[AAMAS 2025] Privacy-preserving and Personalized RLHF, with convergence guarantees. The Code contains experiments for training multiple instances of GPT-2 for personalized sentiment aligned text generation.

Shaping Language Models with Cognitive Insights

RLHF-Blender: A Configurable Interactive Interface for Learning from Diverse Human Feedback

Acceleration framework for Human Alignment Learning

[TSMC] Ask-AC: An Initiative Advisor-in-the-Loop Actor-Critic Framework

🤖 Enhance reinforcement learning stability and efficiency with advanced algorithms like TRPO, PPO, DPO, GRPO, DAPO, and GSPO for optimized policy training.

This repository contains the implementation of a Reinforcement Learning with Human Feedback (RLHF) system using custom datasets. The project utilizes the trlX library for training a preference model that integrates human feedback directly into the optimization of language models.

End-to-end RLHF pipeline: reward modeling, PPO/DPO/GRPO, reward signal design, FSDP scaling analysis, and agent evaluation on GPT-2