sail-sg/understand-r1-zero
A research project analyzing and replicating R1-Zero-like training approaches for language models using reinforcement learning.
Velocity · 7d
+2.8
★ / day
Trend
→steady
star history
This repository contains a paper, models, and codebase for studying R1-Zero-like training methodologies for large language models. The work investigates reinforcement learning approaches to develop reasoning capabilities in LLMs, providing analysis and implementation insights into this training paradigm. The project includes trained models and training scripts released for reproducibility.