user avatar
Chenxi Liu
@chenxi116
TBD @Meta. Previously @GoogleDeepMind Gemini post-trainer. Opinions my own.
chenxi116.github.io
Joined August 2015
288
Following
1,582
Followers
Posts
  • user avatar
    Chenxi Liu
    @chenxi116
    Dec 12, 2024
    We've obviously trained this model for a little while, so today was just a normal day at work. But seeing the somewhat-familiar numbers, now not from raw internal docs but from nicely-formatted CEO's tweet, is an odd feeling unlike any other. Amazing stories. Insane team work.
    user avatar
    Sundar Pichai
    Google
    @sundarpichai
    Dec 11, 2024
    We’re kicking off the start of our Gemini 2.0 era with Gemini 2.0 Flash, which outperforms 1.5 Pro on key benchmarks at 2X speed (see chart below). I’m especially excited to see the fast progress on coding, with more to come.  Developers can try an experimental version in AI
    42K
  • user avatar
    Chenxi Liu
    @chenxi116
    Aug 23, 2020
    My recent blog post on what I think are some "good practices" for a PhD career: cs.jhu.edu/~cxliu/2020/ke… The period between PhD defense and job start seems the perfect time for writing about such a topic.
  • user avatar
    Chenxi Liu
    @chenxi116
    Jan 22, 2025
    More to come, in several senses :)
    user avatar
    Demis Hassabis
    @demishassabis
    Jan 21, 2025
    Our latest update to our Gemini 2.0 Flash Thinking model (available here: goo.gle/4jsCqZC) scores 73.3% on AIME (math) & 74.2% on GPQA Diamond (science) benchmarks. Thanks for all your feedback, this represents super fast progress from our first release just this past
    20K
  • user avatar
    Chenxi Liu
    @chenxi116
    Jul 30, 2018
    Check out our “Progressive Neural Architecture Search” arxiv.org/abs/1712.00559, accepted as oral at #ECCV! Our algorithm automatically discovered PNASNet-5, one of the best models on ImageNet (top 1 82.9%), using only a fraction of the computation than leading RL and EA methods.
  • user avatar
    Chenxi Liu
    @chenxi116
    Dec 19, 2024
    🚨Gemini thinks! With thoughts visible! And multimodal! 2.0 flash wasn't an end; it was truly a beginning Gemini ships! And rocks!
    user avatar
    Logan Kilpatrick
    Google AI Studio
    @OfficialLoganK
    Dec 19, 2024
    Replying to @OfficialLoganK
    It’s still an early version, but check out how the model handles a challenging puzzle involving both visual and textual clues: (2/3)
    00:00
    6.3K
  • user avatar
    Chenxi Liu
    @chenxi116
    Jun 25, 2024
    The team I'm in has a public opening: boards.greenhouse.io/deepmind/jobs/…. This has never happened before, and I was surprised to see it. So apply if you want to work together!
    job-boards.greenhouse.io
    DeepMind
    23K
  • user avatar
    Chenxi Liu
    @chenxi116
    Nov 22, 2024
    we ate a pudding, then carried on sprinting Gemini-Exp-1121 and Gemini-Exp-1114 score much higher than Gemini-1.5-Pro-002
    user avatar
    Arena.ai
    @arena
    Nov 21, 2024
    Woah, huge news again from Chatbot Arena🔥 @GoogleDeepMind’s just released Gemini (Exp 1121) is back stronger (+20 points), tied #1🏅Overall with the latest GPT-4o-1120 in Arena! Ranking gains since Gemini-Exp-1114: - Overall #3 → #1 - Overall (StyleCtrl): #5 -> #2 - Hard
    7.8K
  • user avatar
    Chenxi Liu
    @chenxi116
    Jun 12, 2020
    Today I passed my PhD defense! Huge thank you to my awesome committee @YuilleAlan, Greg Hager, @drfeifei and fabulous collaborators: it has been surreal. This date best defines my life so far; but I can’t wait for more life-defining dates that will outshine this one 😈
  • user avatar
    Chenxi Liu
    @chenxi116
    Jul 30, 2018
    Replying to @PyTorch @TensorFlow and 4 others
    It's github.com/chenxi116/PNAS… with released code for model conversion :)
    GitHub - chenxi116/PNASNet.pytorch: PyTorch implementation of PNASNet-5 on ImageNet
    From github.com
  • user avatar
    Chenxi Liu
    @chenxi116
    Mar 27, 2019
    Check out Weight Standardization (arxiv.org/abs/1903.10520)! We bring normalization from activations to weights, and it pushes GroupNorm over BatchNorm even in the large batch setting. Best part: it’s only 2 more lines of code in @PyTorch and @TensorFlow!
  • user avatar
    Chenxi Liu
    @chenxi116
    Nov 13, 2020
    I am hosting an intern at @Waymo Research summer 2021. Topic is autonomous driving perception and/or #AutoML. Please apply and/or RT :) waymo.com/joinus/2435634/
  • user avatar
    Chenxi Liu
    @chenxi116
    Dec 7, 2024
    we sipped a boba tea, then, you guessed it, carried on sprinting
    user avatar
    Jeff Dean
    @JeffDean
    Dec 6, 2024
    What a way to celebrate one year of incredible Gemini progress -- #1🥇across the board on overall ranking, as well as on hard prompts, coding, math, instruction following, and more, including with style control on. Thanks to the hard work of everyone in the Gemini team and
    3.3K
  • user avatar
    Chenxi Liu
    @chenxi116
    Feb 9, 2019
    It is important to acknowledge the success of deep learning. It is also important not to get too carried away.
    user avatar
    The Gradient
    @gradientpub
    Feb 9, 2019
    "Deep learning has blown past its competition" but will it continue to scale towards increasingly difficult problems? @chenxi116 and Alan discuss what might lie ahead for visual deep learning. thegradient.pub/the-limitation…
  • user avatar
    Chenxi Liu
    @chenxi116
    Sep 5, 2019
    Thank you @GoogleAI for this honor!!
    user avatar
    Google AI
    @GoogleAI
    Sep 5, 2019
    Reflecting our continued commitment to supporting and building relationships with the academic community, we are excited to announce the 2019 class of Google PhD Fellows for North America, Europe, Asia, Australia and the Middle East! goo.gle/2kx3U7J