CJReinforce/PURE
Official code for the paper, "Stop Summation: Min-Form Credit Assignment Is All Process Reward Model Needs for Reasoning"
This project helps AI researchers and machine learning engineers fine-tune large language models (LLMs) to improve their reasoning abilities, particularly for complex mathematical problems. It takes an existing LLM and a dataset of mathematical prompts with process rewards, then outputs a more accurate and efficient LLM for solving reasoning tasks. The end-user is typically an expert working on advanced AI model development.
160 stars.
Use this if you are developing highly capable LLMs for reasoning tasks and need to efficiently fine-tune them using process-supervised or verifiable rewards to achieve state-of-the-art accuracy with fewer resources.
Not ideal if you are looking for a pre-trained, off-the-shelf LLM or if your primary goal is not to advance reasoning capabilities through novel fine-tuning techniques.
Stars
160
Forks
7
Language
Python
License
—
Category
Last pushed
Oct 23, 2025
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/transformers/CJReinforce/PURE"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
agentscope-ai/Trinity-RFT
Trinity-RFT is a general-purpose, flexible and scalable framework designed for reinforcement...
OpenRLHF/OpenRLHF
An Easy-to-use, Scalable and High-performance Agentic RL Framework based on Ray (PPO & DAPO &...
zjunlp/EasyEdit
[ACL 2024] An Easy-to-use Knowledge Editing Framework for LLMs.
huggingface/alignment-handbook
Robust recipes to align language models with human and AI preferences
hyunwoongko/nanoRLHF
nanoRLHF: from-scratch journey into how LLMs and RLHF really work.