Arvind Narayanan (@randomwalker.bsky.social) brings us a talk “What will be left for us to work on?” At #icml2026

The PC Chairs are presenting the welcome remarks. #ICML2026 has over 20,000 in-person attendees!

The impact of Claude: it - has become very cheap to produce submissions (quality?) - remains very expensive to evaluate them by humans. Possible solutions are limited (I don't advocate ... yet) - Limit submissions - LLM assistance for reviews - Partially restrict authors to submit after rejections

Searching for something new to read? All ICML 2026 papers are now public: openreview.net/group?id=ICM... (including all 6551 accepted papers and also rejects which opt-in)

[More #ICML2026] I'll do many things for good friends… even present a Spotify poster after working for a competitor for nearly 8 years! 😁 Come check out our joint work with Walid Bendada, titled "Size- and Dispersion-Corrected Two-Level Softmax Sampling," on July 10 at the SPIGM@ICML workshop.

We will present our position paper, titled "Regulating Algorithms Is Not Enough. A Study on Content Discovery in Online Platforms," next week at #ICML2026 in Seoul, South Korea! 🇰🇷 More details: www.linkedin.com/posts/guilla...

TMLR has been facing an significant uptick in the number of submissions since the start of 2026. This is placing an extreme burden on our amazing team of reviewers & action editors. To ease this burden, TMLR will be implementing submission quotas, effective July 1. 1/n medium.com/@TmlrOrg/ann...

Connect at the ELLIS Sites ICML Fest...in Munich 🇩🇪 on 23 June during the @munichcenterml.bsky.social Munich AI Day 2026, meet the local AI/ML community & engage with the broader ELLIS network with the ELLIS Unit Munich & @zuseschooleliza.bsky.social. 🔗 More info → https://bit.ly/4xoJtcN #ICML2026

The alpha version of my new book "Optimal Transport for Machine Learners" is out, with in particular an online version with interactive figures www.gpeyre.com/ot4ml/

Grateful to have joined such a fantastic panel yesterday for a discussion on "AI and Humans" at Tech for Good 2026, organized by La French Tech Shanghai!

📢 Second Call for Research and Practice Notes: ACM #RecSys2026 📄 Details & submission guidelines can be found here: recsys.acm.org/recsys26/cal...

I am pleased to serve as a panelist at the upcoming Tech for Good 2026 event on Al, organized by La French Tech Shanghai and taking place on June 11th.

Facebook just opened up their meshflow repository, which seems to take a point cloud and a reference image and return a mesh with clean geometry in ~1 second! mesh-flow.github.io github.com/facebookrese...

Better late than never: Sophie Germain’s name is to be engraved on the Eiffel Tower in recognition of her groundbreaking work in mathematics, after a lifetime spent fighting for acceptance in the sexist environment of the late 18th and early 19th centuries. 🗼📐

Happy to share this new LLM paper, just accepted at ECML-PKDD 2026! 🥳 This joint work with my former colleagues from the French music streaming service Deezer presents an LLM-based system automatically describing the “Daily Mix” music playlists recommended to millions of users every day. (1/2)

👀

value-driven transport! a new framework for generative modeling, combining elements of * optimal control / RL * optimal transport * stochastic primal-dual optimization thread about our new work with @pmorenoz.bsky.social (@upf.edu) & Adrian Müller (@ethz.ch) 1/

More good news for ICML 2026! 🎉 Our work with Walid Bendada (Spotify), titled “Size- and Dispersion-Corrected Two-Level Softmax Sampling”, will be presented this July at the ICML 2026 Workshop on Structured Probabilistic Inference & Generative Modeling. More details online soon. See you in Seoul.

Excited to share Sakana AI’s new #ICML2026 paper in collaboration with NVIDIA: "Sparser, Faster, Lighter Transformer Language Models" arxiv.org/abs/2603.23198 This work introduces new open-source GPU kernels and data formats for faster inference and training of sparse transformer LLMs: 🧵 Thread 👇

📢 Second Call for RecSys Challenge: ACM #RecSys2026 📄 Details & submission guidelines can be found here: RecSys 2026 Challenge Website: www.recsyschallenge.com/2026/ RecSys 2026 Website: recsys.acm.org/recsys26/cha...

Today NeurIPS is announcing our official satellite event in Paris. After responding to the call from Ellis following the success of EurIPS in December, we are pleased to reach a new milestone by joining forces with the NeurIPS organizing committee for the 2026 edition.

FAQ on NeurIPS Europe: NeurIPS Europe is an official NeurIPS 2026 satellite event taking place in Paris, France, alongside the main conference in Sydney and the other satellite event in Atlanta. NeurIPS authors can present their papers at any of the three locations, subject to space availability.

Contre l'homophobie et la transphobie à l'École www.education.gouv.fr/contre-l-hom...

📢 [RecSys 2026] Final Call for Industry Contributions: ACM #RecSys2026 📄 Details & submission guidelines can be found here: recsys.acm.org/recsys26/cal...

Deep Learning course! 👨🏼‍🏫 SJTU Paris Elite Institute of Technology

Visiting most of the leading Chinese AI labs, I'm struck by a culture that's extremely well suited to building LLMs with fewer resources, but one happening in a very different ecosystem, more companies at play, almost no data industry, etc. Full report: www.interconnects.ai/p/notes-from...

See you at #ICML2026 🥳🎉

Representing my MSc research students who did some cool science in Shanghai!

Geoffrey Hinton's distillation paper might be the most successful workshop (!!) paper of all time. NIPS Workshop on Deep Learning and Representation Learning. The workshop topic alone would not make sense anymore now, it is basically "ICLR".

Deadline to apply to the postdoc position in our group at #USyd: tonight (Sydney time), in ~15h! usyd.wd105.myworkdayjobs.com/USYD_EXTERNA...

Wonderful collaboration with Xiaohongshu (known internationally as RedNote), a leading Chinese content-driven platform, on a multimodal LLM now powering Explore Feed recommendations for hundreds of millions of users. Project led by Yubin Zhang, with many collaborators. 📝: arxiv.org/pdf/2603.01590

Thrilled to see the full MusicSem paper finally online! 🎉 51 pages detailing our new semantically rich language-audio dataset of music descriptions to support multimodal music representation learning. Project led by Rebecca Salganik with many collaborators worldwide. 📝 arxiv.org/pdf/2602.17769

Always a pleasure to visit and stroll around the University of Tokyo.

Last week, I was deeply honored to become the first non-Chinese recipient of the Xiaomi Young Scholar Award (小米青年学者奖) at SJTU. Backed by Xiaomi, this award aims to recognize and support outstanding young professors at SJTU who have secured national-level research funding in the past year.

Stay tuned... 🎥

Final exam day for my students today 💪

Wild stuff: The ACM Digital Library now shows *by default* an AI-generated summary for open-access research papers. This comes instead of a paper's abstract and has just been rolled out without an opt-in or notification. Visible if you have a subscription: dl.acm.org/doi/full/10....

One of my core beliefs is that community is SO SO SO important. I might go as far to say that it's the only thing that matters, everything else positive is downstream. That's why I'm proud that, in times of crisis, so many of our community members & institutions have stepped up

Unfortunately it appears much of the academic community has reconstituted itself on LinkedIn

Check out my poster today (Thurs) at 11am--2pm session. Exhibit Hall C,D,E Poster Location: #602 "Fisher meets Feynman: score-based variational inference with a product of experts" (NeurIPS spotlight) with Robert Gower, David Blei, and Lawrence Saul @flatironinstitute.org #NeurIPS2025

✨ #NeurIPS2025 paper: Scalable Feature Learning on Huge Knowledge Graphs for Downstream Machine Learning Combining contrastive learning and message passing markedly improves features created from embedding graphs, scalable to huge graphs. It taught us a lot on graph feature learning 👇 1/10

🧠 How do neurons encode information? We know HOW MUCH, but what about WHAT information they encode? Our new work uses diffusion models to decompose neural information down to individual stimuli & features! 🎯Spotlight at #NeurIPS2025 🌟📄 arxiv.org/abs/2505.11309

#NeurIPS2025 is just a few days away! Don’t miss Rebecca Salganik’s presentation of MusicSem, our new dataset of musical descriptions on Reddit capturing musical semantics, at the AI4Music workshop: aiformusicworkshop.github.io

Teaching and engaging with students has been the most fulfilling aspect of my return to academia. This month was particularly special, as I had the pleasure of presenting our SJTU Paris Elite Institute of Technology to high school students! 🎓🇨🇳🇫🇷 More details: www.linkedin.com/feed/update/...

📢🚨Folks, you can only submit 25 papers this year to CVPR (25 submissions per author cap). So, please plan accordingly.

It was a pleasure to speak at our Physical AI seminar, last week at the SJTU Paris Elite Institute of Technology in Shanghai 🇨🇳🇫🇷

What if we did a single run and declared victory

EARL@RecSys2025 organizing team!

Even on Friday afternoon, the last stretch of the #RecSys2025 conference, Hugues Bouchard packed the room with his talk on audiobook retrievability at Spotify… the Best Paper Award at the #EARL2025 workshop! Congratulations! 🎉 @recsys.bsky.social @spotify.com