1. X
  2. Ziyang Luo
Log inSign up
Ziyang Luo
248 posts
user avatar
Ziyang Luo
@ChiYeung_Law
Working on Agentic RL @SFResearch | WizardCoder Creator | Optimizing for the long-term reward.
Singapore
chiyeunglaw.github.io
Joined January 2018
3,435
Following
1,859
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
    Ziyang Luo
    @ChiYeung_Law
    Jul 30
    🚨 Salesforce AI Research just sets a new SOTA on OSWorld 2.0! ⭐️ We did it by making our computer-use agent look less, not more. Everyone's racing the same way: better vision, sharper screen-reading, more pixels. We went the opposite direction. StateAct barely looks at the
    646
  • user avatar
    Ziyang Luo
    @ChiYeung_Law
    Jul 28
    Thanks AK for sharing our work.
    user avatar
    AK
    @_akhaliq
    Jul 28
    StateAct Program State, before Pixels, for Long-Horizon Computer-Use Agents paper: huggingface.co/papers/2607.22…
    997
  • user avatar
    Ziyang Luo
    @ChiYeung_Law
    Jul 15
    Impressive work, especially for long-horizon RL!
    user avatar
    jietang
    Z.ai
    @jietang
    Jul 14
    Our Single-rollout Asynchronous Optimization (SAO), is able to train stably for one thousand steps and consistently outperform GRPO and its variants on agentic coding and reasoning benchmarks, such as SWE-Bench Verified, BeyondAIME, and IMOAnswerBench. arxiv.org/pdf/2607.07508
    936
  • user avatar
    Ziyang Luo
    @ChiYeung_Law
    Jun 9
    The model sets the ceiling; the harness sets the floor. Both push the agent forward.
    user avatar
    Dongxi 东锡 NLP
    @dongxi_nlp
    Jun 8
    Article cover image
    Article
    The Harness Is The Product
    Content move to subscriber only....
    2.1K
  • user avatar
    Ziyang Luo
    @ChiYeung_Law
    Jun 8
    Building a stable, efficient async agentic RL training pipeline is so hard and interesting. 🤣
    4.2K