Jeff shows a great example on how a senior author presents their contribution! 'Minor co-author' as opposed to popular 'Equal Contribution Senior Co-Adviser.

Super excited about the new MatQuant work! Allows training a quantized model where 2bit weights are nested within 4bits and so on. This enables "reading" off accurate models that can have 2bit quantization in the first layer, 4bit in the second layer etc. Along with the

✨New Paper: Presenting Interleaved Gibbs Diffusion (IGD), a novel generative framework for mixed continuous-discrete data, focusing on constrained generation. From 3-SAT and molecule design to layout generation, IGD advances diffusion models by capturing complex inter-variable

Think you know Gemini? 🤔 Think again. Meet Gemini 2.5: our most intelligent model 💡 The first release is Pro Experimental, which is state-of-the-art across many benchmarks - meaning it can handle complex problems and give more accurate responses. Try it now →

Gemini 2.5 Pro #1 across ALL categories, tied #1 with Grok-3/GPT-4.5 for Hard Prompts and Coding, and edged out across all others to take the lead 🏇🏆

The pricing for 2.5 Pro is out. Here is the pareto performance of price : lmsys elo visualized as a rainbow.

My first work Google DeepMind accepted to ICML!

Our recent work Matryoshka Quantization just nested its way into #ICML ! 🪆

First Video from the Learning Together Series on Matryoshka machine learning is live now! Aditya covered everything on matryoshka starting with embeddings, transformers and quantization.

🚨Breaking: Google DeepMind’s latest Gemini-2.5-Pro is now ranked #1 across all LMArena leaderboards 🏆 Highlights: - #1 in all text arenas (Coding, Style Control, Creative Writing, etc) - #1 on the Vision leaderboard with a ~70 pts lead! - #1 on WebDev Arena, surpassing Claude

We’ve developed Gemini Diffusion: our state-of-the-art text diffusion model. Instead of predicting text directly, it learns to generate outputs by refining noise, step-by-step. This helps it excel at coding and math, where it can iterate over solutions quickly. #GoogleIO

Pocket powerhouse admist I/O awesomeness! Gemma 3n E4B & E2B are insane models, optimized for on-device while rivaling frontier models. It's a 🪆Matryoshka Transformer (MatFormer)🪆: Natively elastic b/w 4B & 2B pareto-optimally! ⭐️: free models with ZERO training cost! 🧵👇

Interesting work on reducing reward hacking. Trains a reward model that is aware of the causal attributes pertaining to evaluation.

📢Now open, Gemma 3n weights & it is natively flexible, first of its kind, thanks to MatFormer🪆 Any model between E4B & E2B with ZERO training near Pareto -- we found a bunch! Find a better E3B than what we released, I will send you a 🪆😉 Find the colab for extraction 🧵👇🪆

Hi, we'll be presenting MaGNeTS (arxiv.org/pdf/2502.00382) on 15th July at #ICML2025 📍East Exhibition Hall A-B #3209 🕦 11 AM - 1:30PM Excited to discuss about nested transformers and decode time scaling for visual generation!

Puranjay will present our poster on nested bitwise models or MatQuant, so if you are ICML and interested in the topic, do bother him :) Puranjay is going on the grad-school market this cycle. So if you are looking for a brilliant, hardworking student with good ML+LLM exposure,

Puranjay will be presenting MatQuant on the 16th. Do check it out!

🪆 Matryoshka is extremely general & applicable to every component in our modern ML/DL stack. It can't get more fundamental that 🪆 in bit space to enable elastic quantization! Drop by the poster and say hi to Puranjay (on behalf of Pranav Nair Jeff Dean Prateek Jain & me).

Check out our #ICML2025 poster on Matryoshka Quantization tomorrow, being presented by PURANJAY DATTA on behalf of all the authors.

FYI, as Prateek Jain points out, PURANJAY DATTA is applying to grad schools. x.com/jainprateek_/s…