1. X
  2. Zi Wang, Ph.D.
Log inOpen app
Zi Wang, Ph.D.
127 posts
user avatar
Zi Wang, Ph.D.
@ziwphd
Staff Research Scientist @ Google DeepMind. Previously CS PhD @ MIT CSAIL. Opinions my own. zi-wang.com
Cambridge, MA
ziw.mit.edu
Joined September 2013
197
Following
1,442
Followers
RepliesRepliesMediaMedia

See Zi Wang, Ph.D.’s full profile

Open X App
Don't miss what's happening
People on X are the first to know.
Log inSign up
  • user avatar
    Zi Wang, Ph.D.
    @ziwphd
    May 15
    Check out Proactive Co-Creator on @GoogleAIStudio , a human-AI belief alignment demo I vibe coded: aistudio.google.com/apps/bundled/p… 🧠 See & edit the AI's uncertainty via belief graph. It asks clarifying questions before creating! πŸ“· Try Image βž” Story βž” Video. You can even remix it!
    00:00
    2
    4
    10
    480
  • user avatar
    Zi Wang, Ph.D.
    @ziwphd
    May 2
    I'll present our ACE paper @aistats_conf #Morocco this Sunday, on behalf of @NehaKalibhat @anonymani0 @prasNLP @_beenkim et al. ACE has been a super useful tool for us to explore prompts and gain critical insights on model behavior. Code open-sourced! github.com/google-deepmin…
    user avatar
    Neha Kalibhat
    @NehaKalibhat
    Feb 4
    Thrilled to share that our paper on "Interpreting and Controlling Model Behavior via Constitutions for Atomic Concept Edits" has been accepted at AISTATS 2026! πŸš€πŸš€ Read more about how input mutations can be mapped to interpretable behavioral insights. arxiv.org/abs/2602.00092 🧡
    5
    13
    1.7K
  • user avatar
    Zi Wang, Ph.D.
    @ziwphd
    Apr 28
    Evaluating GenAI is too slow and expensive. Our new paper introduces ProEval, achieving 8-65x fewer samples needed for 1% accuracy! It uses Bayesian quadrature to efficiently estimate model performance and catch failure cases early. More in
    arXiv logo
    arxiv.org
    ProEval: Proactive Failure Discovery and Efficient Performance...
    Evaluating generative AI models is increasingly resource-intensive due to slow inference, expensive raters, and a rapidly growing landscape of models and benchmarks. We propose ProEval, a...
    2
    1
    21
    2.5K
  • user avatar
    Zi Wang, Ph.D.
    @ziwphd
    Feb 5
    Fresh off arXiv! πŸš€ We’re using Atomic Concept Edits as powerful exploration strategies to build rulebooks for model behavior shifts when concepts are manipulated. A new framework for adversarial steering and failure discovery! arxiv.org/pdf/2602.00092 See you at AISTATS 26!
    user avatar
    Neha Kalibhat
    @NehaKalibhat
    Feb 4
    Thrilled to share that our paper on "Interpreting and Controlling Model Behavior via Constitutions for Atomic Concept Edits" has been accepted at AISTATS 2026! πŸš€πŸš€ Read more about how input mutations can be mapped to interpretable behavioral insights. arxiv.org/abs/2602.00092 🧡
    1
    9
    1.1K
  • user avatar
    Zi Wang, Ph.D.
    @ziwphd
    Dec 3, 2025
    πŸ”₯ Proactive Co-Creator is officially LIVE in @GoogleAIStudio! Stop guessing prompts. Start collaborating. Use it now to remix ideas and generate images, stories, and video with an AI that proactively helps you create. πŸ”— Try it here: aistudio.google.com/apps/bundled/p… πŸ“ At #NeurIPS2025?
    8
    27
    14K