user avatar
Liam Fedus
Periodic Labs
@LiamFedus
Building industrial-scale science at @periodiclabs Past: VP of Post-Training @OpenAI; Google Brain
San Francisco, CA
periodic.com
Joined October 2012
1,194
Following
35K
Followers
Posts
  • Pinned
    user avatar
    Liam Fedus
    Periodic Labs
    @LiamFedus
    Sep 30, 2025
    Today, @ekindogus and I are excited to introduce @periodiclabs. Our goal is to create an AI scientist. Science works by conjecturing how the world might be, running experiments, and learning from the results. Intelligence is necessary, but not sufficient. New knowledge is
    3.6M
  • user avatar
    Liam Fedus
    Periodic Labs
    @LiamFedus
    May 13, 2024
    GPT-4o is our new state-of-the-art frontier model. We’ve been testing a version on the LMSys arena as im-also-a-good-gpt2-chatbot 🙂. Here’s how it’s been doing.
    3.3M
  • user avatar
    Liam Fedus
    Periodic Labs
    @LiamFedus
    Mar 17, 2025
    This is what I sent to my colleagues at OpenAI: Hi all, I made the difficult decision to leave OpenAI as an employee, but I’m looking to work closely together as a partner going forward. Contributing to the mission of OpenAI and working with world-class teams to create and
    510K
  • user avatar
    Liam Fedus
    Periodic Labs
    @LiamFedus
    Sep 12, 2024
    Happy to release a couple of our reasoning models today (🍓)! At @OpenAI , these new models are becoming a larger contributor to the development of future models. For many of our researchers and engineers, these have replaced a large part of their ChatGPT usage.
    218K
  • user avatar
    Liam Fedus
    Periodic Labs
    @LiamFedus
    Dec 2, 2022
    Was having so much fun, I forgot to Tweet about it: I joined OpenAI! I’m continuously awed by the people, the technology, and the ambition. We’re just scratching the surface with ChatGPT. If you’re interested, get in touch!
  • user avatar
    Liam Fedus
    Periodic Labs
    @LiamFedus
    Jun 14, 2022
    Today we're releasing all Switch Transformer models in T5X/JAX, including the 1.6T param Switch-C and the 395B param Switch-XXL models. Pleased to have these open-sourced! github.com/google-researc… All thanks to the efforts of James Lee-Thorp, @ada_rob, and @hwchung27
    GitHub - google-research/t5x
    From github.com
  • user avatar
    Liam Fedus
    Periodic Labs
    @LiamFedus
    May 13, 2024
    Replying to @LiamFedus
    Not only is this the best model in the world, but it's available for free in ChatGPT, which has never before been the case for a frontier model.
    135K
  • user avatar
    Liam Fedus
    Periodic Labs
    @LiamFedus
    May 13, 2024
    Replying to @LiamFedus
    But the ELO can ultimately become bounded by the difficulty of the prompts (i.e. can’t achieve arbitrarily high win rates on the prompt: “what’s up”). We find on harder prompt sets — and in particular coding — there is an even larger gap: GPT-4o achieves a +100 ELO over our prior
    229K
  • user avatar
    Liam Fedus
    Periodic Labs
    @LiamFedus
    Nov 20, 2023
    OpenAI is nothing without its people
    92K
  • user avatar
    Liam Fedus
    Periodic Labs
    @LiamFedus
    Jun 16, 2022
    Presenting our survey on emergent abilities in LLMs! arxiv.org/abs/2206.07682 What's it about? Certain downstream language tasks exhibit an interesting behavior: eval curves are flat/random up to a certain model scale, until -- poof -- things start to work. 1/7
  • user avatar
    Liam Fedus
    Periodic Labs
    @LiamFedus
    Sep 12, 2024
    As part of today, we’re also releasing o1-mini. This is an incredibly smart, small model that can also reason before it’s answer. o1-mini allows us at @OpenAI to make high-intelligence widely accessible. openai.com/index/openai-o… On the AIME benchmark, o1-mini re-defines the
    136K
  • user avatar
    Liam Fedus
    Periodic Labs
    @LiamFedus
    May 13, 2024
    GPT-4o is the first model to exceed human performance on MathVista
    user avatar
    Pan Lu
    @lupantech
    May 13, 2024
    🚨 BREAKING: @OpenAI's new GPT-4o model outperforms humans on MathVista for the first time! 📊 Scores: Human avg: 60.3 GPT-4o: 63.8 📖 Learn more: OpenAI : openai.com/blog/gpt-4o/ MathVista: mathvista.github.io
    46K
  • user avatar
    Liam Fedus
    Periodic Labs
    @LiamFedus
    Jan 12, 2021
    Pleased to share new work! We design a sparse language model that scales beyond a trillion parameters. These versions are significantly more sample efficient and obtain up to 4-7x speed-ups over popular models like T5-Base, T5-Large, T5-XXL. Preprint: arxiv.org/abs/2101.03961
  • user avatar
    Liam Fedus
    Periodic Labs
    @LiamFedus
    May 6, 2022
    Fun following LLM retrieval progress. One recent work is Memorizing Transformers which increases context length up to 262k by an external memory of (keys, values) for that document. - Matches quality of Transformers 5x larger - Can fine-tune a prior pre-trained models to use it