You might like
Excited to announce Lucid V1: A world model that can emulate Minecraft environments in real-time on consumer hardware! 🔥 play here: lucidv1-demo.vercel.app post: ramimo.substack.com/p/lucid-v1-a-w… repo: github.com/SonicCodes/luc…
Watch @alex_damian_ give a talk about this paper here: youtube.com/watch?v=04E8r7…
Even with full-batch gradients, DL optimizers defy classical optimization theory, as they operate at the *edge of stability.* With @alex_damian_, we introduce "central flows": a theoretical tool to analyze these dynamics that makes accurate quantitative predictions on real NNs.
Bamberger and Jones et al., "Carré du champ flow matching: better quality-generalisation tradeoff in generative models" Geometric regularization of the flow manifold. Boils down to adding anisotropic Gaussian Noise to flow matching training. Neat idea, enhances generalization.
TRM is one of the best papers I've read in the past years - it truly shows the unfiltered process of a researcher: 1) See awesome paper, get hyped about it 2) Read it - looks cool 3) Run it - doesn't work 4) Find glaring mistakes 5) Fix the issues x.com/jm_alexia/stat…
New paper 📜: Tiny Recursion Model (TRM) is a recursive reasoning approach with a tiny 7M parameters neural network that obtains 45% on ARC-AGI-1 and 8% on ARC-AGI-2, beating most LLMs. Blog: alexiajm.github.io/2025/09/29/tin… Code: github.com/SamsungSAILMon… Paper: arxiv.org/abs/2510.04871
I'm convinced that people with all the extremely bull take on RL after sutton's interview have never touched data directly, and simply don't know the importance of data. "Data is so very important" and "RL is the only way" is secretly mutually exclusive statement. Remember how…
There comes many times where a man chooses between printing his papers or using zotero on his ipad
We wrote a book about representation learning! It’s fully open source, available and readable online, and covers everything from theoretical foundations to practical algorithms. 👷♂️ We’re hard at work updating the content for v2.0, and would love your feedback and contributions
Introducing 𝐂𝐡𝐞𝐚𝐩𝐑𝐞𝐬𝐞𝐚𝐫𝐜𝐡, 2 models (1 thinking and one non-thinking) distilled from Tongyi DeepResearch 30B. The thinking model outperforms non-search models 10-100x the size, while being runnable on a 12gb GPU. The non-thinking model is competitive on SimpleQA…
Im looking to up my “draw explanation” game, any suggestions on how can i be more than an ipad kid
Completed my ~500mile solo bike trip yesterday afternoon, it was originally inspired by riley’s blog (though they’ve shut down the highway through big sur since) Thanks for answering my endless list of questions @rtwlz & @aidanmantine
Im biking from sf to la , if theres anyone at king city today lets group up please apparently death rates are very high for solo biking
Like, I think there’s a strong case to ask for better citation. But it should be recognized that using residuals to get ImageNet SOTA has value in of itself. Otherwise it’s as if Fermat walked on stage with Andrew Wiles and said “guys what’s the big deal, I told you this already”
United States Trends
- 1. Bears 89.7K posts
- 2. Jake Moody 13.7K posts
- 3. Snell 24.4K posts
- 4. Caleb 49.5K posts
- 5. Bills 142K posts
- 6. Falcons 51.4K posts
- 7. Josh Allen 26.7K posts
- 8. Jayden 23K posts
- 9. phil 176K posts
- 10. #BearDown 2,381 posts
- 11. Happy Birthday Charlie Kirk 4,796 posts
- 12. Swift 291K posts
- 13. Joji 30.1K posts
- 14. Ben Johnson 4,448 posts
- 15. Turang 4,331 posts
- 16. #Dodgers 15.4K posts
- 17. Troy Aikman 6,552 posts
- 18. Roki 6,130 posts
- 19. Bijan 33K posts
- 20. Brewers 48.6K posts
Something went wrong.
Something went wrong.