Generative AI
PAAI envisions to power the last mile of GenAI by building intelligent systems through collaborative GenAI and systems that are low-resource, privacy-preserving, and continuously evolving.
We advance GenAI through four core research areas:
- Unconstrained Model Fusion for Enhanced LLM Reasoning
- Ultra-Efficient Low-Bit LLMs: Train & Development
- LLM with Ultra-Low Resource and Strong Reasoning Capability
Powering the Last Mile of GenAI
Federated Learning
PAAI envisions to become a world-leading hub for next-generation Federated Learning (FL), a privacy-preserving distributed AI paradigm that empowers multi-party collaboration without data exposure, enabling collaborative innovation and data democracy to deploy sustainable and responsible AI.
We advance FL through these core research areas:
- Advancing core FL algorithms for addressing efficiency, privacy and utility in FL systems, and building world-class open-source platforms
- Integrating multi-modal Foundation Models into an FL Agent framework to orchestrate privacy-enhanced distributed collaboration
- Deploying high-performance frameworks in key industrial sectors (e.g. medical and financial) to solve critical data and intelligence silos through secure data collaboration
Becoming a world-leading hub for next-generation FL
