Wow, we can now leave comments on any line of a research paper on arXiv.org thanks to @alphaXiv_forum. Amazing work rehaan and Raj Palleti @ ICML 2025. Cannot wait to use it

worldmodels.github.io this is the most beautiful interactive paper i have seen till now. any better ones?

If you’re interested in manifold sampling or SMC come say hi, I’ll talk about Integrator Snippets a robust alternative to standard SMC samplers and show how they can be used to sample from concentrated probability measures! Joint work with Christophe Andrieu

🔥DanteGPT ❄️ I've trained a small GPT-like model on Dante Alighieri's Divina Commedia and deployed it using Flask and Heroku. The model is inspired by Andrej Karpathy's NanoGPT. All feedback is welcome, enjoy! 🤖 dante-gpt-39ee77b9fb5d.herokuapp.com 💻 github.com/MauroCE/DanteG…

I am attending ISBA 2024 ISBA j-ISBA ; If you want to chat about estimating expectations that have intractable normalizing constants, with Monte Carlo, when it is efficient and when not, come to my poster on Friday.

Our paper on SBI using diffusion methods attracted huge crowd! Jack is doing an excellent job explaining it!! Come to poster 1410! ICML Conference

It’s time to say thank you and goodbye to Google DeepMind. I had the immense fortune of working there for 10 years. They were undoubtedly the most exciting years in the history of AI, and I feel that I grew beyond all my expectations thanks to my uniquely smart, generous and

Late chunking - a paper I co-authored with Jina AI - shows that switching the order in which you chunk and embed your document gives better retrieval results 🤯 Thank you Michael Günther and Isabelle Mohr for letting me help! 📃 Full paper: arxiv.org/pdf/2409.04701

🚀 2025 is packed with opportunities to solve high-impact problems 🚀 🛑Combat terrorism & disinformation with ML 🏆Highly-skilled & friendly team 🏢New office in the 💛 of London 🤖Train/deploy/eval SOTA multimodal models ⚖️ Drive fair & responsible AI app.welcometothejungle.com/jobs/gpXoHwCc

Why 'I don’t know' is the true test for AGI—it’s a strictly harder problem than text generation! This magnificent 62-page paper (arxiv.org/abs/2408.02357) formally proves AGI hallucinations are inevitable, with 50 pages (!!) of supplementary proofs.