About me

Keep Learning & Be Positive!

Lijun Wu is currently a Senior Researcher in Microsoft Research AI4Science. He got the Ph.D. degree from Sun Yat-sen University (SYSU), School of Data and Computer Science, and was a member of joint Ph.D. program between SYSU and MSRA, advised by Dr. Tie-Yan Liu and Prof. Jianhuang Lai. He was honored to be awarded with MSRA Ph.D. Fellowship. His team has won 8 champions in WMT19 machine translation competition.

His researches focus on AI4Science (Bio-NLP, Drug Discovery), Large Language Model, Multimodality Learning, Medical Health. He has rich experiences on sequence learning tasks such as neural machine translation. He is also interested in reinforcement learning. He has published many papers in top conferences and journals, such as ICLR, NeurIPS, ACL, TPAMI. He has served as AC/SPC in top conferences, e.g., ACL, EMNLP, NAACL, COLING, AAAI, IJCAI and so on.

You can also refer to the Microsoft Page.

News

🔥2024.3 We (some friends from different domains and I) have released an AI4Science Research Project page, which contains multiple different research projects, check it if you are interested!
🔥2024.3 We have released a comprehensive survey about Leveraging Biomolecule and Natural Language through Multi-Modal Learning: A Survey. Check it!
🔥2023.11 We have released a report on Large Language Models (GPT-4) on Scienctific Discovery, have a check!
🔥2022.4 We have released a comprehensive survey about Non-Autoregressive Generation for Neural Machine Translation and Beyond. Check it!
2024.3 Our BioT5+, a much stronger extension of BioT5 is released.
2024.2 I am honered to give a talk about the LLM in Science Discovery at AGI Leap Summit 2024.
2024.2 One paper is accepted by TPAMI-2024.
2024.1 I am servering as Area Chair for IEEE-CAI-2024.
2023.10 Our BioT5 (pre-trained large language model for bio-chemistry) is accepted by EMNLP-2023.

Awesome Repos

• Awesome-Biomolecule-Language-Cross-Modeling
• Awesome-Bio-Foundation-Models
• Awesome-Docking

Selected Research

• Consistency Training and Dropout
  • R-Drop (sub-model consistency)
  • UniDrop (unified dropout)
  • JANUS (NAT&AT consistency)
• LLM for Science
  • BioT5 (pre-trained large language model for bio-chemistry)
  • BioT5+ (a stronger extension of BioT5)
  • AbGNN (pre-training for antibody design)
• Drug Discovery
  • FABind (Fast and Accurate for Protein-Ligand Binding)
  • DMCG (direct molecular conformation generation)
  • AbGNN (pre-training for antibody design)
• Neural Machine Translation
  • RL4NMT (the first RL for NMT survey)
  • NAT beyond (a comprehensive study of NAT and beyond)
  • BERT-NMT (BERT for NMT)