%20--%3e%3cdefs%3e%3cstyle%3e%20.st0%20{%20fill:%20%23061b40;%20}%20.st1%20{%20fill:%20%23306af1;%20}%20.st2%20{%20fill:%20%235ce5cf;%20}%20%3c/style%3e%3c/defs%3e%3cg%3e%3cpath%20class='st0'%20d='M55,10.5h9v3h-9v9h12V7.5h-12v3ZM64,19.5h-6v-3h6v3Z'/%3e%3cpolygon%20class='st0'%20points='69%2016.5%2078%2016.5%2078%2019.5%2069%2019.5%2069%2022.5%2081%2022.5%2081%2013.5%2072%2013.5%2072%2010.5%2081%2010.5%2081%207.5%2069%207.5%2069%2016.5'/%3e%3cpolygon%20class='st0'%20points='95%2010.5%2095%207.5%2083%207.5%2083%2022.5%2095%2022.5%2095%2019.5%2086%2019.5%2086%2016.5%2095%2016.5%2095%2013.5%2086%2013.5%2086%2010.5%2095%2010.5'/%3e%3cpath%20class='st0'%20d='M40,1.5v21h11.6l1.4-1.4v-7.6h0c0,0-1.4-1.5-1.4-1.5l1.4-1.4V2.9l-1.4-1.4h-11.6ZM50,19.5h-7v-6h7v6ZM50,10.5h-7v-6h7v6Z'/%3e%3c/g%3e%3cpath%20class='st1'%20d='M23.1,24L14.7,4.8l-4.9,11.2h3.8l-1.8,4H3.3L12.1,0H2C.9,0,0,.9,0,2v20c0,1.1.9,2,2,2h21.1Z'/%3e%3cpath%20class='st2'%20d='M34,0h-16.8l10.6,24h6.2c1.1,0,2-.9,2-2V2C36,.9,35.1,0,34,0ZM32.5,20h-4V4h4v16Z'/%3e%3c/svg%3e)
Light R1
Overview :
Light-R1 is an open-source project developed by Qihoo360, aiming to train long-chain reasoning models through curriculum-style supervised fine-tuning (SFT), direct preference optimization (DPO), and reinforcement learning (RL). This project achieves long-chain reasoning capabilities from scratch through decontaminated datasets and efficient training methods. Its main advantages include open-source training data, low-cost training, and excellent performance in mathematical reasoning. The project background is based on the current training needs of long-chain reasoning models, aiming to provide a transparent and reproducible training method. The project is currently free and open-source, suitable for research institutions and developers.
Target Users :
Target audience includes AI researchers, machine learning engineers, and developers interested in long-chain reasoning models. This project is suitable for research teams and enterprises that want to train high-performance long-chain reasoning models with limited resources. It also provides valuable reference for the open-source community.
Use Cases
The Light-R1-7B-DS model achieved 59.1% accuracy in the AIME24 test, significantly outperforming other similar models.
Through curriculum-style SFT and DPO training, Light-R1-32B achieved 76.6% accuracy on AIME24, surpassing DeepSeek-R1-Distill-Qwen-32B.
Developers can quickly reproduce the Light-R1 training process and make customized improvements based on the open-source training code and dataset.
Features
Provides a training method for long-chain reasoning from scratch, without relying on pre-trained long-chain reasoning capabilities.
Open-sources the complete training dataset and code, facilitating reproduction and improvement by researchers.
Employs curriculum learning, improving model performance through SFT and DPO.
Supports reinforcement learning (RL) training to further optimize model performance.
Exhibits excellent performance in mathematical reasoning, particularly in benchmark tests such as AIME24 and AIME25.
How to Use
1. Clone the Light-R1 project code to your local machine.
%20--%3e%3cdefs%3e%3cstyle%3e%20.st0%20{%20fill:%20%230080ff;%20}%20%3c/style%3e%3c/defs%3e%3cpath%20class='st0'%20d='M16.2,11.1c.4.5.4,1.2,0,1.8l-4.7,7.1h-3.8l5.3-8L7.6,4h3.8l4.7,7.1Z'/%3e%3c/svg%3e)
2. Download and install the project's dependent Python packages.
3. Run the SFT training script using the open-source training dataset.
4. Run the DPO training script based on SFT to further optimize the model.
5. Use the trained model for inference or continue RL training.
Featured AI Tools
Tensorpool
TensorPool is a cloud GPU platform dedicated to simplifying machine learning model training. It provides an intuitive command-line interface (CLI) enabling users to easily describe tasks and automate GPU orchestration and execution. Core TensorPool technology includes intelligent Spot instance recovery, instantly resuming jobs interrupted by preemptible instance termination, combining the cost advantages of Spot instances with the reliability of on-demand instances. Furthermore, TensorPool utilizes real-time multi-cloud analysis to select the cheapest GPU options, ensuring users only pay for actual execution time, eliminating costs associated with idle machines. TensorPool aims to accelerate machine learning engineering by eliminating the extensive cloud provider configuration overhead. It offers personal and enterprise plans; personal plans include a $5 weekly credit, while enterprise plans provide enhanced support and features.
Model Training and Deployment
306.6K
Scireviewhub
SciReviewHub is an AI-powered tool designed to accelerate scientific writing and literature reviews. We leverage AI technology to quickly filter relevant papers based on your research goals and synthesize the most pertinent information into easily understandable and readily usable literature reviews. Through our platform, you can enhance your research efficiency, expedite publication timelines, and achieve breakthroughs in your field. Join SciReviewHub and reshape the future of scientific writing!
Research Tools
284.6K