1. X
  2. Pin-Yu Chen
Log inSign up
Pin-Yu Chen
1,190 posts
user avatar
Pin-Yu Chen
@pinyuchenTW
Principal research scientist@IBM Research & Chief Scientist@RPI-IBM AI Research Collaboration & PI@MIT-IBM AI Lab. IJCAI Computers & Thought Award Winner.
New York, USA
pinyuchen.com
Joined April 2018
962
Following
3,631
Followers
RepliesRepliesMediaMedia

Log in or sign up for X

See what’s happening and join the conversation

Continue with phone
or
Log in with username or email
Terms·Privacy·Cookies·Accessibility·Ads Info·© 2026 X Corp.
  • Pinned
    user avatar
    Pin-Yu Chen
    @pinyuchenTW
    Jul 11, 2025
    [@icmlconf thread] I'll be giving an #ICML2025 expo talk on computational safety for #GenAI (4pm, July 13th) and an invited talk (No Free Lunch in AI Safety) at DIG-BUGS workshop on July 19th icml.cc/virtual/2025/w… See you in Vancouver! @IBMResearch @trustworthy_ml @AISecHub
    Talk: https://icml.cc/virtual/2025/46777
Paper: https://arxiv.org/abs/2502.12445
    3.5K
  • user avatar
    Pin-Yu Chen
    @pinyuchenTW
    Jul 9
    And here is the @IBMResearch blog post about this work, along with extended discussions. "How the wrong training environment can teach AI models to misbehave" research.ibm.com/blog/how-the-w… @techethicslab
    user avatar
    Pin-Yu Chen
    @pinyuchenTW
    Jul 5
    Replying to @pinyuchenTW
    (7/n) We design 4 types of vulnerability games to show that capability-driven models tend to exploit system loopholes, a form of shortcut learning from imperfect training environments. This causes misalignment and is transferable across tasks. Paper: openreview.net/forum?id=Rbeke…
    2.8K
  • user avatar
    Pin-Yu Chen
    @pinyuchenTW
    Jul 5
    [#ICML2026 thread] (1/n) Sharing our oral presentations and posters @icmlconf, including programming model internals, adversarial testing and failure diagnosis of LLMs, audio deepfake detection, capability-driven misalignment, MAMBA, and computational safety.
    1.3K
  • user avatar
    Pin-Yu Chen
    @pinyuchenTW
    May 29
    A call to all researchers in adversarial machine learning and agentic AI: The AdvML-Frontiers x CoTMA workshop is coming to @COLM_conf 2026 👉 advml-frontier.github.io The AdvML Rising Star Awards are also back! 👉sites.google.com/view/advml/adv… #AgenticAI #AdvML
    4.2K
  • user avatar
    Pin-Yu Chen
    @pinyuchenTW
    Apr 23
    [#ICLR2026 thread] (1/n) We are presenting 3 works on dynamic benchmarking #GenAI, building safety guardrails for #AgenticAI, and detecting AI edits in images @iclr_conf
    676