Jonathan Geuter

📅 𝗦𝗮𝘁, 𝗝𝘂𝗹 𝟭𝟭 (𝗔𝗱𝗮𝗽𝘁𝗙𝗠 𝗪𝗼𝗿𝗸𝘀𝗵𝗼𝗽, 𝗔𝗦𝗘𝗠 𝗕𝗮𝗹𝗹𝗿𝗼𝗼𝗺 𝟮𝟬𝟮): Understanding Layer Patching in Model Size Interpolation

📅 𝗪𝗲𝗱, 𝗝𝘂𝗹 𝟴, 𝟮.𝟯𝟬𝗽𝗺 (𝗻𝗼𝘄!), 𝗛𝗮𝗹𝗹 𝗔, #𝟭𝟳𝟭𝟯: 🐆Stop Training for the Worst: Progressive Unmasking Accelerates Masked Diffusion Training

I'm in Seoul for ICML! 🇰🇷 If you're around and would like to grab a coffee, let me know :) I'll also present these two papers (today and Saturday):

Day 2 of #ICML2026 is packed with #KempnerInstitute presentations on diffusion models, LLM reasoning, world models for robotics, and much more. Don't miss Wednesday's sessions! #AI

Super excited to start as an AI Research Intern at @IFM_MBZUAI in Sunnyvale next week (and to get back to the Bay after 7 years)! Working with @ssahoo_ and lots of other great people, hope to help train some large open-source models! (If you're in the Bay hit me up!)

📄Paper: "𝗦𝘁𝗼𝗽 𝗧𝗿𝗮𝗶𝗻𝗶𝗻𝗴 𝗳𝗼𝗿 𝘁𝗵𝗲 𝗪𝗼𝗿𝘀𝘁: 𝗣𝗿𝗼𝗴𝗿𝗲𝘀𝘀𝗶𝘃𝗲 𝗨𝗻𝗺𝗮𝘀𝗸𝗶𝗻𝗴 𝗔𝗰𝗰𝗲𝗹𝗲𝗿𝗮𝘁𝗲𝘀 𝗠𝗮𝘀𝗸𝗲𝗱 𝗗𝗶𝗳𝗳𝘂𝘀𝗶𝗼𝗻 𝗧𝗿𝗮𝗶𝗻𝗶𝗻𝗴" (arxiv.org/abs/2602.10314) (updated version coming soon!)

Thank you to everyone involved in this project: my amazing first-co-first-author @Jaeyeon_Kim_0 , @elmelis , @ShamKakade6 , @sitanch

Our paper "𝗦𝘁𝗼𝗽 𝗧𝗿𝗮𝗶𝗻𝗶𝗻𝗴 𝗳𝗼𝗿 𝘁𝗵𝗲 𝗪𝗼𝗿𝘀𝘁: 𝗣𝗿𝗼𝗴𝗿𝗲𝘀𝘀𝗶𝘃𝗲 𝗨𝗻𝗺𝗮𝘀𝗸𝗶𝗻𝗴 𝗔𝗰𝗰𝗲𝗹𝗲𝗿𝗮𝘁𝗲𝘀 𝗠𝗮𝘀𝗸𝗲𝗱 𝗗𝗶𝗳𝗳𝘂𝘀𝗶𝗼𝗻 𝗧𝗿𝗮𝗶𝗻𝗶𝗻𝗴" got accepted to ICML! See you in Seoul 🥳 🇰🇷 More👇

📅 𝗦𝘂𝗻, 𝗔𝗽𝗿 𝟮𝟲 (𝗠𝗠 𝗜𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝗰𝗲, 𝗥𝗼𝗼𝗺 𝟮𝟬𝟰𝗖): 🐆 𝗣𝗨𝗠𝗔: We show that adapting the forward pass of MDMs at training time to the inference time trajectories can speed up pretraining by up to 2.3x. w/ @Jaeyeon_Kim_0 @elmelis @ShamKakade6 @sitanch

📅 𝗙𝗿𝗶, 𝗔𝗽𝗿 𝟮𝟰, 𝟯.𝟭𝟱𝗽𝗺 (𝗣𝟯-#𝟴𝟭𝟴): 🪃 𝗕𝗼𝗼𝗺𝗲𝗿𝗮𝗻𝗴 𝗗𝗶𝘀𝘁𝗶𝗹𝗹𝗮𝘁𝗶𝗼𝗻: You can now create a family of LLMs without any additional training from a single student-teacher pair! w/ @SaraKangaslahti @nihalcanrun @marco_fumero @FrancescoLocat8 @elmelis

I'm super excited to present three of our papers at ICLR in Rio! 🇧🇷 🏖️ If you're around, come say hi, or let me know if you'd like to grab a coffee :) (We'd have to try find one that's not entirely made of sugar though, still searching). Posters in the thread👇

Our Data-Centric ML group is at ICLR 🇧🇷this week. I couldn't make it this year 😰, but @SaraKangaslahti, @JonathanGeuter, @rach_it_ are there. Find them, say hi. Quick rundown 👇

🚨🚨🚨 Now you can stop training your masked diffusion models ''for the worst''. We propose 🐆PUMA🐆--Progressive UnMAsking, a simple modification of the forward masking process that speeds up the masked diffusion training.

@KempnerInst + Boomerang Distillation at UniReps!

#NeurIPS2025 workshops are underway! Today’s workshop presentations from the #KempnerInstitute cover a range topics, including the use of #AI in linguistics and the study of communication & sensing in electric fish. Presenters tagged below 👇 (1/2)

@Uber @Uber trying to reach out here, as your customer support keeps ending chats and is not of much help -- I got charged for an Uber Eats order that contained the wrong items, and I'm trying to get a refund, but your customer support refuses. Can you please help out? Thanks

Uber rides make the perfect opening act 🎶

📄 New preprint alert: We study 🪃Boomerang Distillation🪃, a surprising phenomenon that allows generating a family of pre-trained LLMs of intermediate sizes from a single teacher–student pair — 𝐧𝐨 𝐞𝐱𝐭𝐫𝐚 𝐭𝐫𝐚𝐢𝐧𝐢𝐧𝐠 𝐫𝐞𝐪𝐮𝐢𝐫𝐞𝐝! 🧵👇

@Jaeyeon_Kim_0 @shahkulin98 @ShamKakade6 @sitanch 'Universal Neural Optimal Transport' @JonathanGeuter · Gregor Kornhardt · Ingimar Tomasson · Vaios Laschos doi.org/10.48550/arXiv… (5/20)

If you're at #AISTATS2025, check out the presentation by Jonathan Geuter, in collaboration with @Clement_Bonet_ , @Korba_Anna and @elmelis. 'DDEQs: Distributional Deep Equilibrium Models through Wasserstein Gradient Flows' openreview.net/forum?id=rFfNu… #AI #statistics #ML