🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

✨ I'm excited to share our joint work with Google Research titled “Mesh Neural Cellular Automata” appearing in SIGGRAPH 2024. MeshNCA can directly synthesize dynamic textures on 3D meshes in real time, without UV maps. meshnca.github.io (1/n) Xu Yitao Alex Mordvintsev Eyvind Niklasson

I'll be presenting our Template Programs paper at ICML Conference in Vienna next week! If you want to chat about neurosymbolic models for visual data (or anything else) please come by our poster (Wed, 11:30, Hall C 4-9 #201) or send me a message!

Bayesian Inverse Graphics for Few-Shot Concept Learning. arxiv.org/abs/2409.08351

Itai Sher At the risk of making your case harder, speaking as a philosopher I agree with you 🤣 more specifically, I'd say that a proper analysis of probability requires a general theory of modality (possibility/necessity/actuality), and that's incredibly difficult to deliver.

Keras 3.6 is out now! It includes some really useful goodies like the KerasFileEditor for inspecting and editing weights files, 11 new ops, 3 new layers, the `keras.utils.Config` class for managing your experiment parameters. Also various bug fixes and performance improvements!

New ARC-AGI paper ARC Prize w/ fantastic collaborators Wen-Ding Li @ ICLR'25 Keya Hu Zenna Tavares evanthebouncy Basis For few-shot learning: better to construct a symbolic hypothesis/program, or have a neural net do it all, ala in-context learning? cs.cornell.edu/~ellisk/docume…

We are excited to share our #CORL2024 paper (oral) on "Learning Quadruped Locomotion Using Differentiable Simulation" done in collaboration with Sangbae Kim Massachusetts Institute of Technology (MIT). We present a new way to learn to walk in minutes without parallelization, outperforming PPO in sample efficiency!

🚀 Introducing DiffusionRenderer, a neural rendering engine powered by video diffusion models. 🎥 Estimates high-quality geometry and materials from videos, synthesizes photorealistic light transport, enables relighting and material editing with realistic shadows and reflections

Shape from Semantics: 3D Shape Generation from Multi-View Semantics Liangchen Li, Caoliwen Wang, Yuqi Zhou, Bailin Deng, Juyong Zhang tl;dr: diffusion + 3DGS + optimization -> shape generation, which look differently (as requested) from different angles arxiv.org/abs/2502.00360

New preprint: Generalised Parallel Tempering (GePT) Sampling in multimodal & high-dimensional distributions is tough: ⚠️ Standard MCMC mixes slowly but is consistent ⚠️ Neural samplers are flexible but biased GePT combines the best of both worlds—integrating flows & diffusions

Introducing π-0.5! The model works out of the box in completely new environments. Here the robot cleans new kitchens & bedrooms. 🤖 Detailed paper + videos in more than 10 unseen rooms: physicalintelligence.company/blog/pi05 A short thread 🧵

Here's my 3DV talk, in chapters: 1) Intro / NeRF boilerplate. 2) Recent reconstruction work. 3) Recent generative work. 4) Radiance fields as a field. 5) Why generative video has bitter-lessoned 3D. 6) Why generative video hasn't bitter-lessoned 3D. 5 & 6 are my favorites.

New Keras release is out -- includes saved model file sharding for very large models, the Muon optimizers, new ops and many performance improvements

We just posted our paper on ARC-AGI-2 -- covering its design principles, what makes it more challenging, a detailed analysis of human performance, and current model performance.

Here's how to train and fine-tune JAX models using model•fit() with Keras 3.

New paper: World models + Program synthesis by Wasu Top Piriyakulkij 1. World modeling on-the-fly by synthesizing programs w/ 4000+ lines of code 2. Learns new environments from minutes of experience 3. Positive score on Montezuma's Revenge 4. Compositional generalization to new environments

Check out the project website and paper for more details, including a comparison with Mujoco MPC. This work was jointly led by Sharanya Venkatesh and Bibit Bianchini! approximating-global-ci-mpc.github.io 5/5

Our computer vision textbook is now available for free online here: visionbook.mit.edu We are working on adding some interactive components like search and (beta) integration with LLMs. Hope this is useful and feel free to submit Github issues to help us improve the text!

Impressive flex from the University of Utah: Augmented Vertex Block Descent (AVBD) «Vertex Block Descent is a fast physics-based simulation method that is unconditionally stable, highly parallelizable, and capable of converging to the implicit Euler solution»

The new Keras release (3.11.0) is out! Main upgrades: • int4 quantization with all backends • Support for Grain, a data i/o and streaming library inspired by tf-data, that is backend-agnostic • On the JAX side, integration with the NNX library -- if you're a NNX user, you