GAIR

🏛️ About GAIR Lab

GAIR is part of SII, focusing on Generative Artificial Intelligence Research, with joint effort from SJTU.

Our mission is to create cutting-edge generative AI technologies that are intelligent, ethical, transparent, and accountable, aligned with human values and serving the common good.

Our homepage: https://github.com/GAIR-NLP | https://plms.ai/

---

🔥 Latest Release

Deep Cognition

🔗 Resources

• 🌐 Website: Deep Cognition

AlphaGo Moment for Model Architecture Discovery

🔗 Resources

• 💻 GitHub: Source Code & Implementation

MegaScience: Pushing the Frontiers of Post-Training Datasets for Science Reasoning

🔗 Resources

• 📖 Paper: MegaScience: Pushing the Frontiers of Post-Training Datasets for Science Reasoning
• 🤗 HuggingFace: MegaScience
• 💻 GitHub: Source Code & Implementation

ResearcherBench: Evaluating Deep AI Research Systems on the Frontiers of Scientific Inquiry

🔗 Resources

• 📖 Paper: ResearcherBench: Evaluating Deep AI Research Systems on the Frontiers of Scientific Inquiry
• 🌐 Website: ResearcherBench
• 💻 GitHub: Source Code & Implementation

📚 Previous Works

OctoThinker: Mid-training Incentivizes Reinforcement Learning Scaling

🔗 Resources

• 📖 Paper: OctoThinker: Mid-training Incentivizes Reinforcement Learning Scaling
• 🤗 HuggingFace: OctoThinker Models
• 💻 GitHub: Source Code & Implementation

PC-Agent-E: Efficient Agent Training for Computer Use

🔗 Resources

• 💻 GitHub: PC-Agent-E Repository
• 📖 Paper: PC-Agent-E Research Paper
• 🤗 HuggingFace: PC-Agent-E Model

Thinking with Generated Images

🔗 Resources

• 💻 GitHub: Thinking with Generated Images Repository
• 📖 Paper: Thinking with Generated Images Research Paper
• 🤗 HuggingFace: Thinking with Generated Images Model