Close Menu
    Main Menu
    • Home
    • News
    • Tech
    • Robotics
    • ML & Research
    • AI
    • Digital Transformation
    • AI Ethics & Regulation
    • Thought Leadership in AI

    Subscribe to Updates

    Get the latest creative news from FooBar about art, design and business.

    What's Hot

    Prison IP to Showcase ASM and CTI Improvements at GovWare 2025 in Singapore

    October 14, 2025

    OpenAI and Broadcom Announce Strategic Collaboration to Deploy 10 Gigawatts of Customized AI Accelerators by 2029

    October 14, 2025

    Futures of Work ~ Constructing Higher Techniques for Survivors of Exploitation

    October 14, 2025
    Facebook X (Twitter) Instagram
    UK Tech InsiderUK Tech Insider
    Facebook X (Twitter) Instagram
    UK Tech InsiderUK Tech Insider
    Home»Machine Learning & Research»Rethinking JEPA: Compute-Environment friendly Video SSL with Frozen Academics
    Machine Learning & Research

    Rethinking JEPA: Compute-Environment friendly Video SSL with Frozen Academics

    Oliver ChambersBy Oliver ChambersOctober 11, 2025No Comments2 Mins Read
    Facebook Twitter Pinterest Telegram LinkedIn Tumblr Email Reddit
    Rethinking JEPA: Compute-Environment friendly Video SSL with Frozen Academics
    Share
    Facebook Twitter LinkedIn Pinterest Email Copy Link


    Video Joint Embedding Predictive Architectures (V-JEPA) study generalizable off-the-shelf video illustration by predicting masked areas in latent house with an exponential transferring common (EMA)-updated trainer. Whereas EMA prevents illustration collapse, it complicates scalable mannequin choice and {couples} trainer and pupil architectures. We revisit masked-latent prediction and present {that a} frozen trainer suffices. Concretely, we (i) practice a goal encoder with a easy pixel-reconstruction goal beneath V-JEPA masking, then (ii) freeze it and practice a pupil to foretell the trainer’s latents on masked areas. This results in a two-stage, unregularized scheme that we discuss with as SALT (Static-teacher Uneven Latent Coaching). SALT decouples optimization into pixel reconstruction (trainer) and masked latent prediction (pupil), rising transparency, effectivity, and scalability whereas preserving the flexibility of illustration to generalize beneath frozen analysis. Empirically, our pupil fashions outperform just lately proposed V-JEPA 2 encoders beneath frozen spine analysis throughout various benchmarks. They’re additionally extra compute-optimal: at matched pretraining FLOPs, our technique achieves larger probing accuracy, and its scaling curves dominate V-JEPA’s accuracy-FLOPs Pareto frontier. Lastly, we discover that pupil high quality is remarkably strong to trainer high quality: high-performing college students emerge even with small, sub-optimal lecturers. This factors to a compute finances allocation that ought to overwhelmingly favor the coed. These outcomes place SALT as a easy, scalable, and compute-efficient various to EMA-based self-distillation for video illustration studying.

    • † Work performed whereas at Apple
    Share. Facebook Twitter Pinterest LinkedIn Tumblr Email
    Oliver Chambers
    • Website

    Related Posts

    Remodeling the bodily world with AI: the subsequent frontier in clever automation 

    October 14, 2025

    Constructing Pure Python Net Apps with Reflex

    October 14, 2025

    The Architect’s Dilemma – O’Reilly

    October 14, 2025
    Top Posts

    Evaluating the Finest AI Video Mills for Social Media

    April 18, 2025

    Utilizing AI To Repair The Innovation Drawback: The Three Step Resolution

    April 18, 2025

    Midjourney V7: Quicker, smarter, extra reasonable

    April 18, 2025

    Prison IP to Showcase ASM and CTI Improvements at GovWare 2025 in Singapore

    October 14, 2025
    Don't Miss

    Prison IP to Showcase ASM and CTI Improvements at GovWare 2025 in Singapore

    By Declan MurphyOctober 14, 2025

    Torrance, United States, October 14th, 2025, CyberNewsWirePrison IP at Sales space J30 | Sands Expo…

    OpenAI and Broadcom Announce Strategic Collaboration to Deploy 10 Gigawatts of Customized AI Accelerators by 2029

    October 14, 2025

    Futures of Work ~ Constructing Higher Techniques for Survivors of Exploitation

    October 14, 2025

    Remodeling the bodily world with AI: the subsequent frontier in clever automation 

    October 14, 2025
    Stay In Touch
    • Facebook
    • Twitter
    • Pinterest
    • Instagram
    • YouTube
    • Vimeo

    Subscribe to Updates

    Get the latest creative news from SmartMag about art & design.

    UK Tech Insider
    Facebook X (Twitter) Instagram
    • About Us
    • Contact Us
    • Privacy Policy
    • Terms Of Service
    • Our Authors
    © 2025 UK Tech Insider. All rights reserved by UK Tech Insider.

    Type above and press Enter to search. Press Esc to cancel.