1. X
  2. Yiding Jiang
Log inSign up
Yiding Jiang
398 posts
user avatar
Yiding Jiang
@yidingjiang
Research @GoogleDeepMind | Prev: PhD @mldcmu, AI resident @GoogleAI, BS @Berkeley_EECS. Trying to understand stuff.
yidingjiang.github.io
Joined December 2015
648
Following
2,547
Followers
RepliesRepliesMediaMedia

New to X?

Sign up now to get your own personalized timeline!

Create account

By signing up, you agree to the Terms of Service and Privacy Policy, including Cookie Use.

Terms·Privacy·Cookies·Accessibility·Ads Info·© 2026 X Corp.
Don't miss what's happening
People on X are the first to know.
Log inSign up
  • Pinned
    user avatar
    Yiding Jiang
    @yidingjiang
    Jan 7
    Information theory often gives unintuitive conclusions when it comes to data. Many of these inconsistencies can be resolved elegantly if we limit the amount of computation the observers can use. Very happy to finally introduce our work on epiplexity! 1/🧵
    user avatar
    Marc Finzi
    @m_finzi
    Jan 7
    1/🧵 We are very excited to release our new paper! From Entropy to Epiplexity: Rethinking Information for Computationally Bounded Intelligence arxiv.org/abs/2601.03220 with amazing team @ShikaiQiu @yidingjiang @Pavel_Izmailov @zicokolter @andrewgwils
    Image
    11K
  • user avatar
    Yiding Jiang
    @yidingjiang
    Jun 26, 2025
    A mental model I find useful: all data acquisition (web scrapes, synthetic data, RL rollouts, etc.) is really an exploration problem 🔍. This perspective has some interesting implications for where AI is heading. Wrote down some thoughts:
    yidingjiang.github.io
    The Era of Exploration
    This post explores the idea that the next breakthroughs in AI may hinge more on how we collect experience through exploration, and less on how many parameters and data points we have.
    37K
  • user avatar
    Yiding Jiang
    @yidingjiang
    Oct 21, 2024
    Selecting good pretraining data is crucial, but rarely economical. Introducing ADO, an online solution to data selection with minimal overhead. 🧵 1/n
    Image
    75K
  • user avatar
    Yiding Jiang
    @yidingjiang
    Jan 27, 2022
    It’s that time of the year again…
    Image
  • user avatar
    Yiding Jiang
    @yidingjiang
    Nov 27, 2022
    Hierarchical RL aims to break complex tasks into simpler subtasks, but how do we judge the quality of decomposition w/ minimal prior knowledge? In this new work at NeurIPS (openreview.net/forum?id=D4fuQ…), we show that good decompositions arise from compression. 🧵 1/
    Image
  • user avatar
    Yiding Jiang
    @yidingjiang
    Apr 8, 2024
    Models with different randomness make different predictions at test time even if they are trained on the same data. In our latest ICLR paper (oral), we investigate how models learn different features, and the effect this has on agreement and (potentially) calibration. 1/
    Image
    27K
  • user avatar
    Yiding Jiang
    @yidingjiang
    Oct 13, 2023
    Extremely honored to receive the Google PhD fellowship! Many thanks to my advisor @zicokolter and collaborators, and to @GoogleAI for the generous support.
    user avatar
    Google AI
    @GoogleAI
    Oct 13, 2023
    In 2009, Google created the PhD Fellowship Program to recognize and support outstanding graduate students pursuing exceptional research in computer science and related fields. Today, we congratulate the recipients of the 2023 Google PhD Fellowship! goo.gle/3PYfLXl
    23K
  • user avatar
    Yiding Jiang
    @yidingjiang
    Jul 5, 2023
    An agent is unlikely to be tested in the exact same environment it’s trained in. In our new work with @zicokolter and @robertarail, we show that exploration during training plays a crucial role in zero-shot generalization to new environments. 🧵 1/ arxiv.org/pdf/2306.05483…
    Image
    18K
  • user avatar
    Yiding Jiang
    @yidingjiang
    May 20, 2025
    Data selection and curriculum learning can be formally viewed as a compression protocol via prequential coding. New blog (with @AllanZhou17 ) about this neat idea that motivated ADO but didn’t make it into the paper.
    yidingjiang.github.io
    A compression perspective on curriculum learning
    We describe a unified framework for data selection and curriculum learning via compression.
    18K
  • user avatar
    Yiding Jiang
    @yidingjiang
    Jul 13, 2025
    I will be at ICML next week. If you are interested in chatting about anything related to generalization, exploration, and algorithmic information theory + computation, please get in touch 😀 (DM or email)! My coauthors and I will be presenting 2 papers 👇:
    Image
    7K
  • user avatar
    Yiding Jiang
    @yidingjiang
    Apr 20, 2025
    I will be at #ICLR2025 until Monday. Looking forward to meeting old and new friends. If you want to chat about generalization / RL / curriculum learning / compression & algorithmic info theory (or anything really 😬), please DM me! Otherwise, I will be presenting 2 papers:
    Image
    7K
  • user avatar
    Yiding Jiang
    @yidingjiang
    Jul 4, 2022
    It turns out that for neural networks, agreement degrades the same way as accuracy does under distribution shift👀. This observation gives us a powerful way to estimate OOD accuracy w/ unlabeled data and a collection of different models, even if the models are not calibrated!
    user avatar
    Christina Baek
    @_christinabaek
    Jul 4, 2022
    Estimating out-of-distribution (OOD) performance is hard because labeled data is expensive. Can we predict OOD performance w/ only _unlabeled data_? In our work (arxiv.org/pdf/2206.13089…), we show this can be done using models’ agreement. w/ @yidingjiang, Aditi R., @zicokolter
    Image
    00:00
  • user avatar
    Yiding Jiang
    @yidingjiang
    Oct 14, 2024
    A short paper (maybe too short for ArXiv 🫠) on improving generalization in RL. We showed that efficient exploration combined with fairly simple architectural changes and scaling can significantly improve generalization performance and sample efficiency on Procgen!
    user avatar
    aj
    @anndvision
    Oct 14, 2024
    the results arxiv doesn’t want you to know about : small changes yield big gains on the ProcGen generalization benchmark a 37.9% reduction in the optimality gap ! 📄 andrewjesson.com/assets/report.… 💻 github.com/anndvision/vso… 🪡
    Image
    Image
    6.1K
  • user avatar
    Yiding Jiang
    @yidingjiang
    May 10, 2025
    Replying to @MinqiJiang
    I think it also shows how bad the existing exploration methods are. RL works now because pretraining did the hard part of exploration for it. There could be some room in figuring out how to put more compute into more efficient exploration during RL beyond just sampling more.
    3.1K
Advertisement
Advertisement