wxzhang-photo

Wenxuan Zhang (张雯轩)

Senior Algorithm Engineer (算法专家)
Language Technology Lab
Singapore R&D Center
DAMO Academy, Alibaba Group

Contact: isakzhang [at] gmail.com
More about me: Google Scholar | Blog Posts | LinkedIn | X (Twitter)

About me

I am currently a research scientist at Language Technology Lab, Alibaba DAMO Academy. Prior to that, I obtained my Ph.D. degree from The Chinese University of Hong Kong, under the supervision of Prof. Wai Lam.

My research aims to advance NLP models that are inclusive, supporting diverse languages and cultures (e.g., multilingual large language models), while also trustworthy through techniques that improve safety and robustness of language models. I am passionate about ensuring NLP benefits people equally by making systems accessible, controllable and reliable.

Current research projects are (full publications can be found here or Google Scholar):
  • [Multilingual] Understanding and enhancing multilingual capabilities of LLMs: SeaLLMs
  • [Safety] Ensuring safety of LLMs from development to deployment: multilingual jailbreak
  • [Evaluation] Comprehensive and fair evaluation of foundation models: M3Exam, Auto-Arena
Feel free to drop me an email if you want to collaborate. Research intern positions are also available based in Singapore or China (Hangzhou).

News

  • [June 2024] SeaLLMs was awarded the "Best Innovate for Impact Award" by The International Telecommunication Union (ITU) of United Nations.
  • [May 2024] M3Exam was officialy used to evaluate GPT-4o's multilingual and multimodal capability.
  • [Apr 2024] I am honored to speak at SCS and IMDA about our SeaLLMs project and the evaluation / safety measures we've implemented during the development.
  • [Jan 2024] Two papers accepted to ICLR 2024: multilingual jailbreak challenge & plug-and-play policy planner for LLM-based dialog agent.
  • [Nov 2023] We release the first LLM named SeaLLMs dedicated to languages in Southeast Asia region.
  • [Sep 2023] Two papers accepted to NeurIPS 2023.
  • [June 2023] We release M3Exam, a novel benchmark sourced from real and official human exam questions for evaluating LLMs in a multilingual, multimodal, and multilevel context. [paper] [data]

Professional Service

  • Area Chair: EMNLP 2024
  • Regular PC Member (or Reviewer): ACL, EMNLP, ACL Rolling Review, COLING; NeurIPS, AAAI, IJCAI; SIGKDD, WWW, WSDM
  • Journal Reviewer: ACM Transactions on Information Systems (TOIS), IEEE Transactions on Knowledge and Data Engineering (TKDE), ACM Transactions on the Web (TWEB), Neurocomputing, Transactions on Audio, Speech and Language Processing (TASLP)

Teaching & Talk