Excited to share our recent work DreamPRM, a multi-modal LLM reasoning method achieving first place on the MathVista leaderboard. DreamPRM is an LLM-agnostic framework that can be applied to any multi-modal LLM for improving its reasoning capabilities. It is a bi-level

Thank you all!.

I am thrilled to be promoted to Associate Professor with tenure at UC San Diego. I am deeply grateful for the outstanding work of my students over the past five years, the strong support of colleagues, the valuable help of collaborators, and the unwavering support of my family!

RT @SimularAI: 🚨 Excited to host Prof. Pengtao Xie @cmuptx at our next Simular Seminar — happening next Thursday! 🎉. He’ll share bold insig….

I am excited to receive the NIH Maximizing Investigators' Research Award (MIRA) for Early Stage Investigators. I am deeply grateful to my students for their great work over the past few years and to my collaborators for their strong support and help!.

Paper link:

We are happy to release scLong, the largest single-cell transcriptomics foundation model to date, with 1 billion parameters and a context length of 28,000 genes. It was pretrained on 1300 billion tokens from 48 million cells. Pretrained model and code:

RT @yuyinzhou_cs: Our ICML'21 workshop on "Interpretable Machine Learning for Health" is accepting submissions now!. .

This framework has achieved significant improvement in tasks such as neural architecture search and domain adaptation.

This framework formulates humans' classroom learning skills, such as learning by teaching, learning by progressive examination, interleaving learning, small-group learning, etc., into machines' learning skills, which are leveraged to train better models.

We recently propose a new machine learning framework, Skillearn: machine learning inspired by humans' learning skills. Project page:

Our workshop "Trustworthy AI for Healthcare" organized at AAAI will be held tomorrow (8:30 am - 6:50 pm PST on Feb 9). The workshop is featured with 13 speakers and 23 accepted papers. Welcome to attend! Zoom link:

We are organizing an ICLR21 workshop: Machine Learning for Preventing and Combating Pandemics, featured with 9 invited talks. We welcome submissions of short papers. The deadline is Feb 26.

Our NeurIPS workshop on Self-supervised Learning - Theory and Practice will be held tomorrow: 8:50 am to 6:40 pm PST on Dec 12. Welcome to attend. The workshop is featured with 13 invited talks and 45 accepted papers. Workshop webpage:

We are organizing an AAAI workshop: Trustworthy AI for healthcare, featured by 14 invited speakers. We welcome submissions of short papers. The deadline is Nov 9.

RT @marinkazitnik: We are very excited to be co-organizing a workshop at #AAAI2021 on Trustworthy #AI for #Healthcare! We have a stellar li….

We are organizing a NeurIPS 2020 workshop: Self-supervised Learning -Theory and Practice. We invite submissions of short papers. The submission deadline is Oct 10. The workshop is featured with 13 invited talks. The website of the workshop is

I am happy to share our recent work "Differentially-private Federated Neural Architecture Search".Code:

I am happy to host postdocs via the CIFellow program, in the research direction of sample-efficient learning, NLP, ML for healthcare. Please let me know if you are interested.