AI300
  1. Andrej Karpathy / AI researcher
    2461.
    @Jess_Riedel @goakhmad Yeah I think granted, it’s why I call them prompts.›
    LLM score 30 · 9 months ago
  2. Andrej Karpathy / AI researcher
    2462.
    @ashleybcha @goakhmad I should add that 1 even lotr used plenty of CGI overall and just like AI now it is a powerful tool in the box.›
    LLM score 20 · 9 months ago
  3. Andrej Karpathy / AI researcher
    2463.
    @ashleybcha @goakhmad I think it’s right that you’re grouping CGI with (current-capability etc) AI because they are both fundamentally technologies that trades wins in time/cost against increase in slop.›
    LLM score 80 · 9 months ago
  4. Xiang Fu / Researcher at Periodic Labs
    2464.
    @khoomeik Checkout old PGM works if this is your thing›
    LLM score 20 · 9 months ago
  5. Xiang Fu / Researcher at Periodic Labs
    2465.
    Scaling laws are often presented as a story of data, model size, and compute.›
    LLM score 80 · 9 months ago
  6. Emmanuel Ameisen / Anthropic Interpretability Researcher
    2466.
    By decomposing weights using loss curvature, you can identify components used for memorization vs generalization.›
    LLM score 90 · 9 months ago
  7. Andrej Karpathy / AI researcher
    2467.
    @soumithchintala Great run Soumith! Of all the deep learning framework transitions I've been through where I re-wrote ~all of my code (matlab -> caffe -> numpy -> torch -> pytorch), the PyTorch one was most pleasant and now significantly longest lasting.›
    LLM score 60 · 9 months ago
  8. Niklas Muennighoff / AI Researcher at Stanford
    2468.
    New, extremely challenging visual reasoning benchmark "MIRA" where current models fail...great resource to research reasoning with imgs/video🌌 https://t.co/1ecwLKoOkv https://t.co/X56ckYBK6O
    LLM score 80 · 9 months ago
  9. Tri Dao / Chief Scientist at Together
    2469.
    Tons of effort from IBM and vLLM folks to make these hybrid models go fast.›
    LLM score 20 · 9 months ago
  10. Tri Dao / Chief Scientist at Together
    2470.
    Thank you @schmidtsciences for the 2025 #AI2050 Early Career Fellowship supporting my work on self-improving AI systems: as AI gets better, it should help human experts design better model architectures and faster training & inference systems https://t.co/1BqPcvg1R8
    LLM score 40 · 9 months ago
  11. Richard Song / DeepMind Research Scientist
    2471.
    Reliable performance is crucial for doing great math.›
    LLM score 85 · 9 months ago
  12. Noam Brown / OpenAI Research Scientist
    2472.
    @dtklp1 @hughbzhang Demonstrated expertise in an area I know well, or a recommendation from someone I trust, still carries a lot of weight.›
    LLM score 20 · 9 months ago
  13. Noam Brown / OpenAI Research Scientist
    2473.
    In 2019 @hughbzhang sent me a detailed personalized cold email asking to intern with me.›
    LLM score 20 · 9 months ago
  14. Noam Brown / OpenAI Research Scientist
    2474.
    @offchan420 Yes but the point is specifying the reward in a way that leads pure self play to optimal real-world behavior is virtually impossible.
    LLM score 80 · 9 months ago
  15. Naman Jain / Researcher at Cursor
    2475.
    We added LLM judge based hack detector to our code optimization evals and found models perform non-idiomatic code changes in upto 30% of the problems 🤯 https://t.co/CPwEQvRQJC
    LLM score 85 · 9 months ago
  16. Andrej Karpathy / AI researcher
    2476.
    @GoodFaithOnly @binarybits When I say AGI I mean a "feature complete" build of a remote worker with a lot of 9s (think: a bit like current Autopilot FSD, maybe plus a few more iterations).›
    LLM score 70 · 9 months ago
  17. Emmanuel Ameisen / Anthropic Interpretability Researcher
    2477.
    Neat use of attribution graphs to identify features in models which are responsible for jailbreak resistance and refusals.›
    LLM score 85 · 9 months ago
  18. Hang Gao / ex MTS at xAI
    2478.
    The value of eval lies in a distributional look at the model performance.›
    LLM score 92 · 9 months ago
  19. Victoria Lin / Thinking Machines Researcher
    2479.
    Very interesting read ☕ When poking different frontier models (e.g., GPT-5 vs Gemini), I’ve often noticed surprising similarity on non-STEM questions.›
    LLM score 85 · 9 months ago
  20. Xiang Fu / Researcher at Periodic Labs
    2480.
    Second @simonbatzner. FF architecture was no longer a bottleneck for AI materials discovery since nequip.›
    LLM score 80 · 9 months ago
  21. Noam Brown / OpenAI Research Scientist
    2481.
    @lineardiff @ericzelikman I don’t view that as incompatible.›
    LLM score 20 · 9 months ago
  22. Emmanuel Ameisen / Anthropic Interpretability Researcher
    2482.
    Striking result, which changed how I think about LLMs:›
    LLM score 85 · 9 months ago
  23. Niklas Muennighoff / AI Researcher at Stanford
    2483.
    new paper "ATLAS" – for details: https://t.co/MJO9mvlS90; it also covers a new scaling law, pretrain vs finetune for multilingual & more :)›
    LLM score 70 · 9 months ago
  24. Niklas Muennighoff / AI Researcher at Stanford
    2484.
    To scale data-constrained LLMs, repeating & denoising objectives can help.›
    LLM score 85 · 9 months ago
  25. Tri Dao / Chief Scientist at Together
    2485.
    State space architecture for state of the art voice model! https://t.co/QjdOdrIFSM
    LLM score 80 · 9 months ago
  26. Emmanuel Ameisen / Anthropic Interpretability Researcher
    2486.
    Cool paper using attribution graphs to automatically detect which reasoning steps contain mistakes!›
    LLM score 85 · 9 months ago
  27. Lilian Weng / Thinking Machines Cofounder
    2487.
    On-policy distillation provides an elegant way to use the teacher model as a process reward model to provide dense reward while preventing SFT style "OOD shock" during rollout.›
    LLM score 90 · 9 months ago
  28. Mira Murati / Thinking Machines CEO
    2488.
    Combining the benefits of RL and SFT with on-policy distillation, a promising approach for training small models for domain performance and continual learning.›
    LLM score 85 · 9 months ago
  29. Xiang Fu / Researcher at Periodic Labs
    2489.
    To adapt to today's world, crank down discount factor and crank up learning rate
    LLM score 75 · 9 months ago
  30. John Schulman
    2490.
    Happy to share a new paper! Designing model behavior is hard -- desirable values often pull in opposite directions.›
    LLM score 80 · 9 months ago
More