Really cool work! This was definitely on my ICLR highlights list this year.

Bytes Are All You Need: Transformers Operating Directly On File Bytes paper page: huggingface.co/papers/2306.00… ByteFormer, achieves an ImageNet Top-1 classification accuracy of 77.33% when training and testing directly on TIFF file bytes using a transformer backbone with

Well, I guess I'll have to switch to mindlessly scrolling through the "Discover" articles that Chrome inserts into new empty tabs.

Check out our work that just won the Best Paper at the Ubiquitous Robotics Conference 2023. jennyzzt.github.io/gta/

Every linux box has a story.

Congratulations Dieter!

🚨Is it possible to devise an intuitive approach for crowdsourcing trainable data for robots without requiring a physical robot🤖? Can we democratize robot learning for all?🧑‍🤝‍🧑 Check out our latest #CoRL2023 paper-> AR2-D2: Training a Robot Without a Robot

Time traveller from past: Wow! You guys have AI! How does it work? Guy from future: No idea. TTFP: Who does know? GFF: Literally no one. We just told this thing stuff we like and eventually it learned to do what we want. TTFP: Wow, that must be so great! GFF: No I hate it.

Announcing MatFormer - a nested🪆(Matryoshka) Transformer that offers elasticity across deployment constraints. MatFormer is an architecture that lets us use 100s of accurate smaller models that we never actually trained for! arxiv.org/abs/2310.07707 1/9

Whats a good intro to genetic algorithms and other gradient-free methods?

Fun holiday project github.com/aaronwalsman/b4. (I think) I solved Bodega Brawl (bodegabrawl.com).

Please spread the word!

Calling all Harvard University undergrads interested in the study of intelligence: our spring research program’s application deadline is Jan. 28! More details: bit.ly/3GruS76 #KempnerInstitute Ella Batty Sham Kakade Bernardo Sabatini

Extremely excited to be joining Kempner Institute at Harvard University as a research fellow this fall! If you are in the area and want to talk about research or future collaborations, do reach out!

Congratulations to #KempnerInstitute’s Marinka Zitnik on receiving the U.S. National Science Foundation CAREER award! Zitnik’s project develops #AI that can analyze complex networks of data with applications for therapeutics and #drugdiscovery: bit.ly/3xXPu64 #ML Harvard Medical School

Congratulations to our #KempnerInstitute faculty and students who will be showcasing their work at ICLR 2026 this week! #ICLR2024   *indicates Kempner presenter Naomi Saphra Gabriel Kreiman Melanie Weber Sham Kakade Andy Keller George Alvarez talia konkle Cengiz Pehlevan Sam Gershman

Excited about Zoey Chen's new work on real2sim for robotics! We present URDFormer, a technique to learn models that go from RGB images to full articulated scene URDFs in sim by "inverting" pre-trained generative models. These can be used to train robots for the real-world! 🧵(1/8)

We’re thrilled to announce #KempnerInstitute’s first cohort of post-bac scholars! These 4 recent college grads will undertake extensive mentored research & coursework with 2 years of funding & professional development. Read more: bit.ly/3KhCJ9x @harvardGSAS #AI #NeuroAI

Recent work with an amazing team at the Kempner Institute at Harvard University and friends! I want to highlight the contributions of our undergrads Joseph Bejjani, Chase Van Amburg and Vincent Wang, who are all applying for PhD programs this fall. They crushed this one, keep an eye out for them!

🎉 Excited to share that our paper “Convergent Functions, Divergent Forms” will be presented at NeurIPS 2025🤖 in San Diego! We present LOKI, a compute-efficient framework for co-evolving robot morphologies🦾 and control policies⚙️. LOKI discovers diverse, high-performing robot