conv.

All stories
AIRising · 10h

Stanford/NVIDIA researchers unveil CLM, a contrastive-learning model for fast agent action selection

CLM-8B matches larger 'System Two' model Jev's accuracy while running up to 9× faster on computer-use, gaming and tool-calling tasks.

What to know

  • CLM pairs a state encoder and an action encoder trained with a contrastive InfoNCE objective, letting the model score and select actions directly via embedding similarity rather than generation.
  • CLM-8B reportedly matches the accuracy of the larger Jev model across computer-use, gaming, and tool-calling benchmarks while running up to 9× faster.
  • Speed gains are largest in tasks with many candidate actions (e.g., WikiRacing) or high action reuse across states (e.g., a T-Rex game benchmark).
  • The work comes from a Stanford/NVIDIA Research team including Jacky Kwok, Christopher Ré, and Azalia Mirhoseini; announced via a Notion writeup and a viral X post.

Jacky Kwok Lead researcher, Stanford UniversityChristopher Ré Senior co-author, Stanford UniversityAzalia Mirhoseini Senior co-author, Stanford University / NVIDIA ResearchHangoo Kang Co-author, Stanford UniversityTarun Suresh Co-authorJon Saad-Falcon Co-author

Stanford/NVIDIA researchers unveil CLM, a contrastive-learning model for fast agent action selection
x.com

How it unfolded 3 developments, newest first · click a bar or a number to jump posts

Peak 1 piece in one quarter hour at Yesterday, 7 PM; 4 pieces over 11 hours (1 article · 3 posts) Yesterday, 7:10 PM — quietYesterday, 7:25 PM — 1 piece · 1 post — X 1Yesterday, 7:40 PM — quietYesterday, 7:55 PM — quietYesterday, 8:10 PM — quietYesterday, 8:25 PM — quietYesterday, 8:40 PM — quietYesterday, 8:55 PM — quietYesterday, 9:10 PM — quietYesterday, 9:25 PM — quietYesterday, 9:40 PM — quietYesterday, 9:55 PM — quietYesterday, 10:10 PM — quietYesterday, 10:25 PM — quietYesterday, 10:40 PM — quietYesterday, 10:55 PM — quietYesterday, 11:10 PM — quietYesterday, 11:25 PM — quietYesterday, 11:40 PM — quietYesterday, 11:55 PM — quietToday, 12:10 AM — 1 piece · 1 post — Hacker News 1Today, 12:25 AM — quietToday, 12:40 AM — quietToday, 12:55 AM — quietToday, 1:10 AM — quietToday, 1:25 AM — 1 piece · 1 article — Google News 1Today, 1:40 AM — quietToday, 1:55 AM — quietToday, 2:10 AM — quietToday, 2:25 AM — quietToday, 2:40 AM — 1 piece · 1 post — Hacker News 1Today, 2:55 AM — quietToday, 3:10 AM — quietToday, 3:25 AM — quietToday, 3:40 AM — quietToday, 3:55 AM — quietToday, 4:10 AM — quietToday, 4:25 AM — quietToday, 4:40 AM — quietToday, 4:55 AM — quietToday, 5:10 AM — quietToday, 5:25 AM — quietToday, 5:40 AM — quietToday, 5:55 AM — quiet 123
8 PM10 PMtoday2 AM4 AMnow · 6:10 AM ET
  1. 3

    Coverage of CLM spreads further on Hacker News

    A second Hacker News submission linking to the original X announcement appears, indicating continued pickup of the CLM announcement, though with minimal additional discussion.

  2. 2

    Research team publishes CLM details and benchmark results

    A Notion page from the Stanford/NVIDIA team (Jacky Kwok, Hangoo Kang, Tarun Suresh, Jon Saad-Falcon, Marco Pavone, Christopher Ré, Azalia Mirhoseini) explains CLM's contrastive state/action encoder design and reports CLM-8B matching Jev's performance at up to 9× the speed across computer-use, gaming, and tool-calling tasks; the writeup circulates on Hacker News.

    “Across computer-use, gaming, and tool-calling tasks, CLM-8B performs on par with Jev while running up to 9× faster.”
    — Stanford/NVIDIA research team, Paper authors · source
  3. 1

    Jacky Kwok announces CLM-8B on X

    Lead researcher Jacky Kwok posts an introduction of Contrastive Language Model (CLM), describing it as an 'ultra-fast System One Model' pre-trained on internet-scale data that is up to 9× faster than Jev with comparable performance.

    “Introducing Contrastive Language Model (CLM): an ultra-fast System One Model trained with a contrastive learning objective that connects states and actions.”
    — @jackyk02
    1. first by MarkTechPost, 4h ago

    • Introducing Contrastive Language Model (CLM): an ultra-fast System One Model trained with a contrastive learning objective that connects states and actions. CLM-8B is pre-trained on internet-scale data and delivers up to 9× faster inference than Jev ⚡ while achieving comparable

      @jackyk02X10h ago1.9k▲view on X ↗