Matrix multiplication is the basic computational operation that powers modern computing (including AI). Yet, the theoretical fastest speed at which computers can multiply matrices (omega ω) is still unknown and has been a longstanding challenge for complexity theory and computer science.
Today, we announce a new record for omega (ω<2.371177). This is the result of a great team effort between @GoogleDeepMind, our academic collaborators, and our Gemini-powered coding agent AlphaEvolve! 🧮
arxiv.org/abs/2608.16884v1
We introduce 🌸✨ AlphaEvolve ✨🌸, an evolutionary coding agent using LLMs coupled with automatic evaluators, to tackle open scientific problems 🧑🔬 and optimize critical pieces of compute infra ⚙️
deepmind.google/discover/blo…
📢New Paper on Reward Modelling📢
Ever wondered how to choose the best comparisons when building a preference dataset for LLMs?
Our latest paper revives classic statistical methods to do it optimally!
Here’s a 🧵on how it works 👇
arxiv.org/abs/2502.04354
Image-generation diffusion models can draw arbitrary visual-patterns. What if we finetune Stable Diffusion to 🖌️ draw joint actions 🦾 on RGB observations?
Introducing 𝗚𝗘𝗡𝗜𝗠𝗔
paper, videos, code, ckpts: genima-robot.github.io/
🧵Thread⬇️
🚨Important update from our Robot Learning Lab in London.
Following recent news, we’re moving on after a wonderful 2 years…
Today, we unveil 4 big pieces of research from our incredible team. Check out the compilation video and thread below to see our final work! 📽️👇
We introduce C3, which significantly improves COOL-CHIC compression performance, approaching the SOTA neural codec (MLIC+) while requiring 200x fewer FLOPs to decode. We also extend C3 to videos 🎥
COOL-CHIC (Ladune et al., 2023), learns a decoder per image, as well as a latent grid and an entropy model *per image*. This dramatically improves compression at a very low decoding cost
arxiv.org/abs/2212.05458
COIN (Dupont et al, 2021) learns a small decoder *per image*, leading to low decoding cost. However, compression performance is weak
arxiv.org/abs/2103.03123
Traditional neural compression models are based on autoencoders trained on datasets of natural images or videos. While these achieve good compression, the decoder is often large as it needs to generalize to arbitrary images, leading to expensive decoding
arxiv.org/abs/2202.06533
We build neural codecs from a *single* image or video, achieving compression performance close to SOTA models trained on large datasets, while requiring ~100x fewer FLOPs for decoding ⚡ #CVPR2024c3-neural-compression.github…
We present #FunSearch in @Nature today - a system combining LLMs with evolutionary search to generate new discoveries in math and computer science! 👩🔬🔬✨
Introducing FunSearch in @Nature: a method using large language models to search for new solutions in mathematics & computer science. 🔍
It pairs the creativity of an LLM with an automated evaluator to guard against hallucinations and incorrect ideas. 🧵 dpmd.ai/x-funsearch
We construct neural processes by iteratively transforming a simple stochastic process into an expressive one, similar to flow/diffusion-based models, but in function space!
Join us at our #NeurIPS2023 poster session: neurips.cc/virtual/2023/post… on Wednesday morning!
Introducing Manifold Diffusion Fields (MDF), our new work on learning generative models over fields defined on curved geometries. This is joint work with our intern @Ahmed_AI035 (who hasn’t even started his PhD yet!) and @jsusskin at @Apple MLR arxiv.org/abs/2305.15586 🧵
Manifold Diffusion Fields
present Manifold Diffusion Fields (MDF), an approach to learn generative models of continuous functions defined over Riemannian manifolds. Leveraging insights from spectral geometry analysis, we define an intrinsic coordinate system on the manifold via the eigen-functions of the Laplace-Beltrami Operator. MDF represents functions using an explicit parametrization formed by a set of multiple input-output pairs. Our approach allows to sample continuous functions on manifolds and is invariant with respect to rigid and isometric transformations of the manifold. Empirical results on several datasets and manifolds show that MDF can capture distributions of such functions with better diversity and fidelity than previous approaches
paper page: huggingface.co/papers/2305.1…