Accepted at #ICML2024 🥳 Happy to share “Aligning Transformers with Weisfeiler-Leman” where we advance the SOTA of standard transformers for graphs by making them more expressive, more scalable and equivariant. 📝: arxiv.org/abs/2406.03148 🧑‍💻: github.com/luis-mueller/w…

Wondering where (the theory of) graph machine learning is going? Check out our ICML 2024 position paper: arxiv.org/abs/2402.02287. We outline challenges in understanding expressive power, generalization, optimization, and aligning theory with practice.

Are there foundation models for graphs? TL;DR: there are a few already! In a new blog post with Michael Galkin Haitao Mao Jianan Zhao Zhaocheng Zhu we discuss the main challenges and existing approaches.

🚀We're thrilled to introduce TGB 2.0, a benchmark for learning on temporal knowledge graphs and heterogeneous graphs.🚀 TGB 2.0 extends the Temporal Graph Benchmark (TGB) to temporal knowledge and heterogeneous graphs, offering a collection of diverse, challenging datasets. 1/4

Temporal Graph Benchmark 2.0 now features knowledge graphs and heterogeneous graphs of various sizes and flavors (and leaderboards to show off your numbers ofc)! Check out Julia's announcement

The Learning on Graphs Conference 2024, the leading conference dedicated to graph machine learning, is happening! Submission deadline: September 11th, 2024 Final decision: November 13th, 2024 Don't miss out on this opportunity! More details at logconference.org/cfp/#LoG2024 #GraphML #MachineLearning

Myself and Yoshua Bengio are hiring a postdoctoral researcher Mila - Institut québécois d'IA! For this call, we are prioritizing candidates with experience in reinforcement learning, scientific discovery, or high-impact applications of ML. Apply here (docs.google.com/forms/d/e/1FAI…)

I won't be attending ICML this year (getting a Schengen visa seems to be impossible in California this time of year), will have to deliver all my talks to this purrfectly fluffy audience.

💥 Introducing PINDER & PLINDER With existing evals saturating without clear advances in real life downstream tasks, current progress in AI x Bio is primarily rate-limited by better datasets & evals. In two back-to-back preprints, we address this via fantastic academic-industry

Tune in tomorrow (Monday) for the Graph Learning Tutorial at ICML Conference - Adrian and Ameya prepared an overview on the latest and greatest in Graph ML! 🌶️ Finally, there will be a panel discussion with bald claims and hot takes 😏

I’ll be at #ICML2024 this week presenting our joint work with Remy Lau Dominique Beaini Ladislav Rampasek: “Graph Positional and Structural Encoder” — happy to chat on graph learning theory and applications to chemistry & biology! 🕦Tuesday 23rd, 11:30am 📍 Hall C 4-9 #701 See you there!

Come meet us by our “Graph Positional and Structural Encoder” poster on Tuesday at #ICML2024 !

Wow I didn't expect my presence would be _so large (like "the big brother" 1984 large) - I would've used a background image with cats then!

Having difficulty keeping up with new AI safety research? Great news: My new blog will help with just that! "AI Safety at the Frontier" covers each month's (subjectively) best papers. In July '24, I discuss Safetywashing, SAD, AgentDojo and much more: open.substack.com/pub/aisafetyfr…

We are recruiting postdocs for 2024/25 Real AI @ UCSB ! You want to build next-gen artificial scientific intelligence?🤖 Apply to the UCSB Real AI Initiative! Deadline: Sept 15 Competitive salary Prime location for your future office 🌊 Yao Qin Haewon Jeong University of California

Peer review as a concept has more pros that cons, imo the problem is in the incentives - those days many reviewers serve as gatekeepers and reject works similar to theirs in order to claim sota for a few more months. I think Learning on Graphs Conference 2024 is a good example of working peer review

Studying on developing foundation models on single cell RNA-seq (scRNA-seq) data? The serge continues!💪 We're excited to announce our new work "scCello", a single cell, Cell-ontology guided transcriptome foundation model (TFM) on scRNA-seq data. Paper: arxiv.org/pdf/2408.12373