About Xubin Ren

Xubin Ren

Hi, I’m Xubin Ren. I’m a Ph.D. candidate at the Institute of Data Science, The University of Hong Kong, advised by Prof. Chao Huang and Prof. Benjamin C. M. Kao.

I began doing research at Wuhan University, where I studied Computer Science and Technology and worked with Prof. Zhenzhong Chen at the IIP Lab. Since then, my questions have moved toward large language models, information retrieval, and AI agents: how can a system find the right context, reason with it, and still be useful in the real world?

I like research most when it grows beyond a paper. nanobot is my attempt to turn that instinct into an open-source system people can actually use.

Along the way, I’ve been honored as an AI25 AI Innovation Pioneer Under 25, a BAAI InnoVibe 2026 Most Notable Academic Rising Star, and a CAAI Rising Scientist.

xubinrencs AT gmail.com GitHub Twitter LinkedIn Google Scholar

Recent updates

  1. Grateful to be named an AI25 AI Innovation Pioneer Under 25!
  2. Grateful to receive the BAAI InnoVibe 2026 Most Notable Academic Rising Star award!
  3. One long paper, MiniRAG, is accepted by ACL'26!
  4. One full paper is accepted by KDD'26!
  5. Grateful to receive the Ant InTech Scholarship - Future!
  6. One survey paper is accepted by ACM TIST!
  7. Two papers are accepted by EMNLP'25!
  8. Grateful to receive the WAIC Youth Outstanding Paper Nomination Award!
  9. One survey paper is accepted by ACM Computing Surveys!
Earlier updates
  1. Grateful to be selected as the CAAI Rising Scientist!
  2. One paper is accepted by EMNLP'24!
  3. One survey is accepted by KDD'24!
  4. One full paper and a tutorial are accepted by WWW'24!
  5. Two full paper (oral presentation) are accepted by WSDM'24!
  6. One long paper is accepted by CIKM'23!
  7. Graduated from Wuhan University with Honor!
  8. Two full papers are accepted by SIGIR'23!
  9. Awarded the HKPF and HKU-PS! Thanks all!
  10. One paper is accepted by ICLR'23!
  11. One paper is accepted by ACM TOIS!
  12. Awarded the National Scholarship for the fourth time! Thanks all!
  13. Awarded the CCF Elite Collegiate Award! Thanks all!
  14. Awarded the Finalist Winner in Mathematical Contest in Modeling (MCM)! Thanks all!

What I build

nanobot is where most of that thinking has converged. It began as an experiment in how little machinery a useful personal agent really needs. The project has since outgrown that early constraint: today it is an open-source, self-hosted agent framework that runs in a WebUI, terminal, and chat apps, with tools, memory, MCP integrations, multi-agent workflows, and automation. The goal is no longer to hit a line-count target; it is to keep the core legible while the system and its community grow.

nanobot downloads nanobot GitHub stars nanobot GitHub forks code · docs

The path to nanobot ran through Vimo, where the question moved closer to everyday use: could someone talk to a video that was far too long for a model’s context window? We built a desktop application around that interaction, while the retrieval ideas behind it developed into VideoRAG.

Built with PyTorch VideoRAG GitHub stars VideoRAG GitHub forks paper · code

Before either of them, there was SSLRec. Self-supervised recommendation had become a crowded research area, and I wanted one place where ideas could be reproduced and compared instead of living in isolated paper repositories. It became a shared PyTorch framework—and my first lesson that research code can be useful beyond its original experiment.

Built with PyTorch SSLRec GitHub stars SSLRec GitHub forks paper · code

I like building things that start as research questions and grow into something people can actually use.

Research

Publications & Awards →

My long-term goal is artificial general intelligence: systems that can acquire knowledge, adapt through experience, and keep improving after deployment.

My research began with a narrower question: what can a model learn from behavior when supervision is sparse? DCCF (SIGIR 2023) learned the different intents behind user–item interactions, while SSLRec (WSDM 2024 · Oral) turned this line of work into a shared framework for building and evaluating self-supervised recommenders.

With RLMRec (WWW 2024) and EasyRec (EMNLP 2025), I began using language to give those behavioral representations more semantic structure. RLMRec aligned LLM-generated profiles with collaborative signals; EasyRec pushed the same idea toward text-based zero-shot recommendation.

The next bottleneck was not only learning from the data a model already had, but finding the right evidence when the necessary context was too large or missing. VideoRAG (KDD 2026) brought retrieval to extremely long videos; MiniRAG (ACL 2026 · Long) asked how much of that capability could remain when the model and the system had to stay lightweight.

That brings me back to the question I started with, now at the scale of agents: how can a deployed system turn new interactions into lasting knowledge and skills without forgetting what it already knows? This is the continual-learning problem I want to work on—and, for me, a necessary step toward AGI.

Education

The University of Hong Kong

Ph.D. in Data Science
2023 — Present

Wuhan University

B.Eng. in Computer Science and Technology
2018 — 2023

Selected honors

  • AI25 AI Innovation Pioneer Under 25, 2026
  • BAAI InnoVibe Academic Rising Star, 2026
  • CAAI Rising Scientist, 2025
  • Hong Kong PhD Fellowship, 2023
  • National Scholarship, 2019–2022

See all awards →

Academic service

Reviewer: Information Fusion, ACM TOIS, ACM TKDD, ACM TKDE, ACM TIST, and IEEE TNNLS.

Program committees: NeurIPS, ICLR, ICML, KDD, WWW, WSDM, CIKM, RecSys, and ACL ARR.