My PhD student Chi Han ’s work is always theoretically grounded, simple, lightweight and elegant. I’m very proud of the recent work LMSwitch for generative language model conditioning
Congratulations to Dr. Prof. Manling Li for passing her PhD defense without revision!
Existing large vision-language models have close to random performance on action related tasks. Zhenhailong and team have developed a novel learning framework PAXION to learn action knowledge and patch it into large vision-language models. The performance on action knowledge related tasks is improved from ~50% to ~80%.
Upcoming keynote speeches: ``Large Neural Models' Self-Learning Symbolic Knowledge'' at Seventh International Workshop on Symbolic-Neural Learning, and ICML2023 Workshop on Knowledge Reasoning; ``Joint Text and Molecule Learning for Drug Discovery'' at Korean NLP Symposium 2023
``Large Neural Models' Self-Learning Symbolic Knowledge'' at Tsinghua University, Chinese Academy of Sciences and Peking University.
``Joint Text and Molecule Learning for Drug Discovery'' at Center for the Transformation of Chemistry, ``Joint Text and Molecule Learning for Drug Discovery'' at NeurIPS2023 Workshop on Graph Learning Frontiers.
The event extraction community has been stuck in using an 18 years old ACE data set with only 33 event types. It’s time to move on … we have developed a new dataset with 3,465 event types and 208K sentences, along with a high performing new system to extract them.
Check out this paper on General-Purpose Event Detection, led by the beyond amazing UIUC PhD student Zoey Li (she is on job market).
I was asked by our college of engineering to share experiences on seeking and securing funding. I’m sharing some slides publicly in case they are useful for young PIs.
I have many post-doc, research scientist, RAs, software engineer positions and project managers available to work on information extraction related research on news domain and scientific domain, please drop me an email if interested.