Howard Chen


I am a fifth year PhD student in the Princeton NLP Group and the Princeton Language + Intelligence center, co-advised by Professor Danqi Chen and Professor Karthik Narasimhan.

My research goal is to build controllable agents that can operate on a long interactive horizon and evolve over time. I am particularly interested in memory — how we can design proper structure and algorithm to make it elastic and can self-reorganize (easily updatable), robust (insensitive to perturbation), and persistent (minimal forget). I am also interested in related topics such as unlearning, mechanistic understanding, and safety from the memory perspective.

Prior to Princeton, I was an ML researcher at ASAPP and research assistant at Cornell NYC Tech. I recieved my M.Eng. from Cornell University and B.S. from National Taiwan University.

Please see my google scholar page for the up-to-date publications.

howardchen@cs.princeton.edu

news

Jun 5, 2023 I will be interning at Meta AI over the summer working with Asli Celikyilmaz and Jason Weston.
Dec 25, 2022 New preprint Controllable Text Generation with Language Constraints is out!
Sep 21, 2022 Our work on grounded language agent has been accepted to NeurIPS 2022! You can check out our demo here.
May 28, 2022 Our work on examining the robustness of extractive rationale models is accepted to NAACL 2022!
Apr 6, 2021 Check out the article that covers our newly released ABCD dataset.