trl

Official implementation of GDPO: Group reward-Decoupled Normalization Policy Optimization for Multi-reward RL Optimization

Turn any research paper into a commercialization report — 6 AI agents, TRL/MRL scoring, patent landscape, market intelligence, verified citations. DeepSeek / OpenAI / Claude.

LLaMA-TRL: Fine-tuning LLaMA with PPO and LoRA

Notus is a collection of fine-tuned LLMs using SFT, DPO, SFT+DPO, and/or any other RLHF techniques, while always keeping a data-first approach

An implementation of GRPO for Unsloth's VLMs training

基于DPO算法微调语言大模型，简单好上手。

Code repository dedicated to experimenting and research with tiny reasoning language model

Distill teacher chains-of-thought into a LoRA adapter via a strict boxed-answer format contract + two-phase Train→Nudge (silver-medal NVIDIA Nemotron reasoning recipe, as a tested library).

Agentic RL 零基础中文教程：24 章从概念到 GRPO 实战，含 TRL 最小可跑示例 | Beginner-friendly Agentic RL tutorial with hands-on GRPO project

simpleR1: A Simple Framework for Training R1-like Models

realize the reinforcement learning training for gpt2 llama bloom and so on llm model

使用trl、peft、transformers等库，实现对huggingface上模型的微调。

Different post-training techniques for LLMs, including: SFT, DPO and Online RL

Config-driven LLM fine-tuning with safety evaluation, EU AI Act compliance, 6 alignment methods, and one-command bundled quickstart templates.

Reasoning-first post-training for tiny language models (140M) on a single GPU.

Domain-neutral LLM post-training pipeline for CPT, Fact-SFT, optional DPO, adapter merge, quality evaluation, GGUF/ONNX export, and OpenAI-compatible inference.

DPO/SafeDPO/OPAD training + eval for teaching tool-using LLMs to refuse falsely-benign MCP exploits

Minimal code to train reasoning model with reinforcement learning.

Process-supervised RL for a multi-step reasoning agent — DAPO + a learned Process Reward Model (PRM) training a Qwen3-8B Planner. A modern, from-scratch rebuild of the AgentFlow paper (ICLR 2026).

A production-ready Python script that automatically generates high-quality supervised fine-tuning (SFT) datasets for chat-based language models using Google's Gemini 2.0 Flash API. Compatible with TRL (Transformer Reinforcement Learning) and Hugging Face Transformers.