• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    AI

    Kangwook Lee Shares WHALE Paper on Model Training

    Lee, CAIO at KRAFTON, links to arXiv paper on joint harness-weight training.

    CF
    OK
    DP
    6 Sources, 28d ago, first seen 28d ago

    TLDR

    Kangwook Lee posted that with data, one can and should train both model weights and harness. The tweet links to the paper WHALE: A SIMPLE RECIPE FOR JOINT HARNESS–WEIGHT OPTIMIZATION by Haechan Kim et al. from KRAFTON, KAIST and Stanford. It includes a screenshot of the paper and states that a full thread is coming. Dimitris Papailiopoulos retweeted the post from Lee, who is CAIO at KRAFTON.

    Combined views

    65.2K

    6 Sources, first seen 28d ago

    Combined views

    65.2K

    6 Sources, first seen 28d ago

    872 likes
    872 likes
    13 comments
    830 saves
    184 reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    13 comments
    830 saves
    184 reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet

    6 Sources

    @Kangwook_LeeWith data, you can/should train both model weight & harness. 🐳: https://arxiv.org/abs/2609.00196 A full blown thread is on its way 🙂
    @DimitrisPapailRT @Kangwook_Lee: With data, you can/should train both model weight & harness. 🐳: https://arxiv.org/abs/2609.00196 A full blown thread is on its w…
    @yoonholeeeHarness optimization is sample-efficient but plateaus. What should you do if you can afford to update the model too? Introducing WHALE: a simple recipe for jointly optimizing an LLM's weights and harness. Blog: https://krafton.ai/blog/whale/ Paper: https://arxiv.org/abs/2609.00196
    @lateinteractionRT @yoonholeee: Harness optimization is sample-efficient but plateaus. What should you do if you can afford to update the model too? Intro…
    @chelseabfinnRT @yoonholeee: Harness optimization is sample-efficient but plateaus. What should you do if you can afford to update the model too? Intro…
    @cwolferesearchGreat visualization and a nicely-executed idea. Many tools are available for model specialization. A lot of work focuses on harnesses or finetuning separately, but why not use them together? Bullish on efficient recipes for optimizing models in custom / task-specific harnesses!

    6 Sources

    @Kangwook_LeeWith data, you can/should train both model weight & harness. 🐳: https://arxiv.org/abs/2609.00196 A full blown thread is on its way 🙂
    @DimitrisPapailRT @Kangwook_Lee: With data, you can/should train both model weight & harness. 🐳: https://arxiv.org/abs/2609.00196 A full blown thread is on its w…
    @yoonholeeeHarness optimization is sample-efficient but plateaus. What should you do if you can afford to update the model too? Introducing WHALE: a simple recipe for jointly optimizing an LLM's weights and harness. Blog: https://krafton.ai/blog/whale/ Paper: https://arxiv.org/abs/2609.00196
    @lateinteractionRT @yoonholeee: Harness optimization is sample-efficient but plateaus. What should you do if you can afford to update the model too? Intro…
    @chelseabfinnRT @yoonholeee: Harness optimization is sample-efficient but plateaus. What should you do if you can afford to update the model too? Intro…
    @cwolferesearchGreat visualization and a nicely-executed idea. Many tools are available for model specialization. A lot of work focuses on harnesses or finetuning separately, but why not use them together? Bullish on efficient recipes for optimizing models in custom / task-specific harnesses!