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

    Alex Dimakis Thanks Thinky Machines and John Schulman

    Berkeley professor publicly thanks two accounts for support and AI tools.

    AD
    MS
    BR
    5 Sources, 27d ago, first seen 27d ago

    TLDR

    Alex Dimakis, a Professor of EECS at UC Berkeley whose research covers machine learning, information theory, and generative AI, and who co-founded Bespoke Labs AI, made a reply on the platform X. He tagged the accounts @thinkymachines and @johnschulman2. The message expressed thanks for support and for the great tools built by those accounts. This forms the visible evidence in the provided source lines for the cluster.

    Combined views

    76.5K

    5 Sources, first seen 27d ago

    Combined views

    76.5K

    5 Sources, first seen 27d ago

    462 likes
    462 likes
    14 comments
    520 saves
    78 reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    14 comments
    520 saves
    78 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

    5 Sources

    @AlexGDimakisHow to post-train a model to personalize it on your code repo? In our latest research in Bespoke Labs, we post-trained a model to improve its performance on a given Github repository. Starting from Inkling base, we use supervised fine-tuning (SFT) with trajectories coming from a strong teacher model, and reinforcement learning (GRPO) on repository-specialized environments that we curated. SFT gave a 52pp improvement in performance on the held-out fontTools evaluation set. Further RL training lifts the total improvement to 57pp compared to the base Inkling model. In addition to the in-distribution evaluation our post-trained Inkling shows good performance on Terminal-Bench 2.1 and SWE-Bench Lite while becoming 40% more token efficient due to post-training. Read our full research blog post here: https://bespokelabs.ai/blog/personalizing-inkling-for-your-code-repository-with-post-training Many thanks to Thinking Machines Lab for their credit contribution that helped support this research.
    @madiatorCheck out our blog post on how we did SFT and RL to improve the performance of open models on specific code repositories. This should be a great blueprint for everyone! https://bespokelabs.ai/blog/personalizing-inkling-for-your-code-repository-with-post-training Thanks to @tinkerapi for making all of this easy.
    @xlr8harder@AlexGDimakis What do you do for reasoning traces when you are extracting from Claude? Is the SFT set available?
    @rbhar90RT @AlexGDimakis: How to post-train a model to personalize it on your code repo? In our latest research in Bespoke Labs, we post-trained a…

    5 Sources

    @AlexGDimakisHow to post-train a model to personalize it on your code repo? In our latest research in Bespoke Labs, we post-trained a model to improve its performance on a given Github repository. Starting from Inkling base, we use supervised fine-tuning (SFT) with trajectories coming from a strong teacher model, and reinforcement learning (GRPO) on repository-specialized environments that we curated. SFT gave a 52pp improvement in performance on the held-out fontTools evaluation set. Further RL training lifts the total improvement to 57pp compared to the base Inkling model. In addition to the in-distribution evaluation our post-trained Inkling shows good performance on Terminal-Bench 2.1 and SWE-Bench Lite while becoming 40% more token efficient due to post-training. Read our full research blog post here: https://bespokelabs.ai/blog/personalizing-inkling-for-your-code-repository-with-post-training Many thanks to Thinking Machines Lab for their credit contribution that helped support this research.
    @madiatorCheck out our blog post on how we did SFT and RL to improve the performance of open models on specific code repositories. This should be a great blueprint for everyone! https://bespokelabs.ai/blog/personalizing-inkling-for-your-code-repository-with-post-training Thanks to @tinkerapi for making all of this easy.
    @xlr8harder@AlexGDimakis What do you do for reasoning traces when you are extracting from Claude? Is the SFT set available?
    @rbhar90RT @AlexGDimakis: How to post-train a model to personalize it on your code repo? In our latest research in Bespoke Labs, we post-trained a…