If you want to pursue academic excellence and depth instead of chasing fashion, please participate conferences like CPAL. One condition for this conference is that it will always be held on the campus of a host university so that the conference will be small and focused. So one

【5 Days Until #HKU Info Day 2024 🐼】 Talk #2 talk under the "Data Science in 1,800 Seconds for Beginners" series will be Prof Chao Huang's session! Save your seat🕺: hku.au1.qualtrics.com/jfe/form/SV_6n… Seize the chance to get IDS souvenirs- See you soon! #hkudatascience #hkuinfoday2024

【Grand Opening of IDS New Premises on Oct 29 – Share the Joy with Us for Its Promising Future Dev't amidst Its 3rd Birthday!🎉🙌🏻】 Let’s revisit the memorable day with our team - hope to see your faces at different research collaborative opportunities😉 #HKU #hkudatascience

🚀 Excited to Share Our Recent Work 🌟GraphAgent🌟 — An innovative AI agent that efficiently integrates structured and unstructured data! 📚 👉 Paper link: arxiv.org/abs/2412.17029 👉 GitHub page: github.com/HKUDS/GraphAge… #GraphAgent is our newly proposed agent workflow that

✨ What Makes Aria-UI Stand Out • Diverse Instruction Understanding. • Context-Aware Grounding. • Lightning-Fast: 3.9B active parameters deliver top-tier speed and efficiency. • Versatile & Scalable: Handles any screen size, resolution, or layout with ease.

A new year gift from the HKU IDS. The opening ceremony of the Advanced Technology Fellow Program of the HKU.

🔍 Seeking a simpler RAG solution? 📢 Introducing 🔥MiniRAG🔥: The Ultra-Lightweight RAG Framework #MiniRAG reimagines Retrieval-Augmented Generation for resource-constrained environments, offering exceptional performance with small language models (SLMs). Ready to simplify

Happy Chinese new year, from the School of Computing and Data Science, the University of Hong Kong!

VideoRAG: Retrieval-Augmented Generation with Extreme Long-Context Videos Presents a RAG framework for processing and understanding extremely long-context videos through a dual-channel architecture leveraging graphbased knowledge & multimodal retrieval 📝arxiv.org/abs/2502.01549

VideoRAG: How Can AI Truly Understand Hours of Video Content? 👉 WHY THIS MATTERS Processing and analyzing long videos (lectures, documentaries, multi-part series) has been a persistent challenge for AI systems. Traditional methods fragment content into short clips, losing

VideoRAG: Retrieval-Augmented Generation with Extreme Long-Context Videos arxiv.org/abs/2502.01549 Project page: github.com/HKUDS/VideoRAG

🚀 Introducing our 🔥VideoRAG🔥 "Retrieval-Augmented Generation with Extreme Long-Context Videos" 🎥📺 💥 Breaking the Video Length Limitation in Understanding Extreme Long Videos 📄 Paper: arxiv.org/abs/2502.01549 🔗 Code: github.com/HKUDS/VideoRAG 💻 Efficient Extreme

酷，一款可用于理解超长视频的RAG框架：VideoRAG，可以处理数百小时的视频内容，可以同时理解多个视频 支持数百小时的视频输入，可以同时分析多个视频 高效率的知识索引，可将数百小时的视频提炼成结构化的知识图谱，理解视频内容间的关联，提供准确回复

VideoRAG，超长上下文视频处理理解，利用单块 NVIDIA RTX 3090 GPU（24G）来理解数百小时的视频内容，通过自适应多模态融合机制优化检索效率，保持多视频上下文中的语义连贯性。

【CITERS2025】HKU IDS is one of the co-organizers to the CITER Symposium this year! The keynote speeches will broadcast live👉 check out the program: citers2025.cite.hku.hk/program-overvi… Don't miss out! The event is happening today and tomorrow🥳

datascience.hku.hk/2025/04/hku-id…

🎬 Excited to Launch Vimo: Chat with Your Videos Like Never Before! 🚀 ⭐ Fully open-source on GitHub: github.com/HKUDS/Vimo Transform how you interact with video content using cutting-edge AI technology. Whether it's a 30-second clip or an extensive 100+ hour lecture series,

Vimo! Vimo! Chat with Your Videos~~🚀 "Say goodbye to video context limits! 🎬 With Vimo, you can chat with videos of any length. From short clips to 100+ hour documentaries—Vimo's got you covered. As always, everything is open-sourced: github.com/HKUDS/ViMo (now with over