After yesterday's Salon de Refusés paper presentation and today's EIML workshop paper presentation with @jkobialka.bsky.social, we will conclude with an oral by @erpel.bsky.social at the MedEurIPS workshop tomorrow 🔥 Very happy that we had so many great opportunities to present our research! ❤️EurIPS

EurIPS was pretty fun! Thanks to @lacerbi.bsky.social , @conorhassan.bsky.social , and all organizers of the Amortized ProbML team for the great workshop and for inviting me as a speaker! 🙌 Also great to be able to present our spotlight paper on sparsity on the main stage 😍

If you are at ELLIS UnConference on Dec 2 (pre @euripsconf.bsky.social ), join us at our Amortized ProbML workshop! Come for our great speakers: @paulbuerkner.com Yingzhen Li @davidruegamer.bsky.social @liza-semenova.bsky.social & poster session; stay for our hot takes on amortized probML!

Updated website: sites.google.com/view/amortiz... Co-organized with Cen-You (Scott) Li, @conorhassan.bsky.social, @desirivanova.bsky.social as well as @samikaski.bsky.social & FCAI

Some personal news: I've joined sophont.med to help build the next generation of open medical foundation models. We've relaunched medarc.ai, our open science research community. Join us if you want to help advance open medical AI. And we are hiring.

I have not announced it widely but I will be starting a professorship in TU Darmstadt (Germany) next year. I will start accepting PhD students from Sep 2026 (but I will keep my group and position in Tokyo as well). Here is a first call for PhD positions at our new Center of Excellence at TUDa […]

The abstract submission deadline for the ELLIS Workshop on Amortized Probabilistic ML has been extended to October 25th! You still have time to submit a short abstract on your work in: Amortized Inference, Neural processes, Prior-fitted networks, Sequential Decision-Making, and Foundation models.

**Submission DL extended to Oct 25!** You are still in time to submit a short abstract on amortized (simulator-based) inference (SBI), neural processes, prior-fitted networks (PFNs), amortized experimental design, foundation models, and related applications. Call: sites.google.com/view/amortiz...

Looks like an awesome workshop!

Reminder that Amortized inference Workshop submissions at the ELLIS Unconference are still open until **Oct 16, 2025**. Only a short abstract (½ page), so go ahead! Workshop: Dec 2, 2025, co-located with EurIPS. Website: sites.google.com/view/amortiz...

🔸 Organizers: Cen-You Li, @conorhassan.bsky.social, Desi R. Ivanova, Luigi Acerbi (+ @samikaski.bsky.social as Advisory Member) 📣 Call for contributions open until Oct 10 👉 Details: eurips.cc/ellis

submit your posters and attend this exciting workshop!

Want to work on Trustworthy AI? 🚀 I'm seeking exceptional candidates to apply for the Digital Futures Postdoctoral Fellowship to work with me on Uncertainty Quantification, Bayesian Deep Learning, and Reliability of ML Systems. The position will be co-advised by Hossein Azizpour or Henrik Boström.

EurIPS workshop on AmOrTIzED InfErENcE!!!1!

Do you like to train neural networks to solve all your nasty probabilistic inference and sequential design problems? Do you love letter salads such as NPs, PFNs, NPE, SBI, BED? Then no place is better than the Amortized ProbML workshop we are organizing at #ELLIS UnConference.

submit your posters and attend this exciting workshop!

We'll discuss tricky questions such as: • When does amortization fail badly even though it's rarely talked about? • Which applications can these neat methods actually work for? • Can we really deploy these methods in the real world? • ... and more!

Short abstract submission is open (DL Oct 16, 2025 AoE) Accepted abstracts will be presented as posters. 🌍 Location: Copenhagen (co-located with #EurIPS) 📅 Date: 2 Dec, 2025 🔗 Website: sites.google.com/view/amortiz... 📝Call for abstracts: sites.google.com/view/amortiz...

Superlative initiative! Massive kudos to the organisation team for getting this going in such a short time!

NeurIPS is endorsing EurIPS, an independently-organized meeting which will offer researchers an opportunity to additionally present NeurIPS work in Europe concurrently with NeurIPS. Read more in our blog post and on the EurIPS website: blog.neurips.cc/2025/07/16/n... eurips.cc

Tomorrow, Tuesday (July 1st) from 4pm to 5pm (UK time). “SDE Matching: Scalable and Simulation-Free Training of Latent Stochastic Differential Equations" (arxiv.org/abs/2502.02472) 🚀 Join via Zoom 🔥 t.co/N1C3UFukxd

🚨🚀 Come hear @gbarto.bsky.social talk about SDE Matching tomorrow! SDE Matching is a highly efficient and scalable training framework for Latent/Neural SDEs. You no longer have to discretize or simulate your SDE models when fitting them to data. #SDE #Diffusion #FlowMatching #ML

Blog post: lacerbi.github.io/blog/2025/ju...

New blog post! You can train a neural network to *just do things* -- such as *predict the optimum of a function*. But how do you get a big training dataset of "functions with known optima"? Read the blog post to find out! (link 👇)

I want to share my latest (very short) blog post: "Active Learning vs. Data Filtering: Selection vs. Rejection." What is the fundamental difference between active learning and data filtering? Well, obviously, the difference is that: 1/11

Website for more info: acerbilab.github.io/amortized-co...

@huangdaolang.bsky.social will be presenting our work tomorrow on pretrained/amortized transformers for all sorts of inference tasks, go say hi! #AISTATS2025 Poster session 1 Place: Hall A-E Time: Sat 3 May 3 p.m. — 6 p.m. Poster number: 23

11/10 This work is part of ongoing research at our group @univhelsinkics.bsky.social & FCAI on efficient & robust probabilistic methods for science. Interested in surrogate based and amortized inference? We often have open PhD/postdoc positions! Check us out: www.helsinki.fi/en/researchg...

10/10 Read the full paper #AABI2025 Proceedings Track! Work by @chengkunli.bsky.social, @abacabadabacaba.bsky.social, Petrus Mikkola, & yours truly. 📜 Paper: arxiv.org/abs/2504.11554 💻 Code: github.com/acerbilab/no... 📄 Overview: acerbilab.github.io/normalizing-... Let us know what you think!

1/10🔥 New paper alert in #AABI2025 Proceedings! Normalizing Flow Regression (NFR) — an offline Bayesian inference method. What if you could get a full posterior using *only* the evaluations you *already* have, maybe from optimization runs?

Equivariant Denoisers Cannot Copy Graphs: Align Your Graph Diffusion Models Najwa Laabid · Severi Rissanen · Markus Heinonen · Arno Solin · Vikas Garg Hall 3 + Hall 2B #194 🗓️ Fri 25 Apr 3 p.m. +08 — 5:30 p.m. +08 📄 arxiv.org/abs/2405.17656

Our #ICLR2025 poster "Free Hunch: Denoiser Covariance Estimation for Diffusion Models Without Extra Costs" (Severi Rissanen, Markus Heinonen, Arno Solin) 🗓️ Hall 3 + Hall 2B #140, Thu 24 Apr 3 p.m. +08 — 5:30 p.m. +08 📄 Preprint: arxiv.org/abs/2410.11149

Our #ICLR2025 poster "Discrete Codebook World Models for Continuous Control" (Aidan Scannell, Mohammadreza Nakhaeinezhadfard, Kalle Kujanpää, Yi Zhao, Kevin Luck, Arno Solin, Joni Pajarinen) 🗓️ Hall 3 + Hall 2B #415, Thu 24 Apr 10 a.m. +08 — 12:30 p.m. +08 📄 Preprint: arxiv.org/abs/2503.00653

This is a new blog looking at the individual optimizations that went into serving the DeepSeek model class in SGLang. I have been observing the SGLang repo for a few months now, and it's crazy how quickly they integrate new optimized features. It's a very cool open-source project!

1/ We asked GPT-4.5 -- allegedly the model with the best sense of humor, according to the site we do not talk about here -- to write a comic about our recent AISTATS paper on the Amortized Conditioning Engine (ACE). Then gpt-4o drew it. You judge the result... (text continues 👇)

B200 go brrrr! It seems by doubling the TFLOPs you get double the speed. Cool stuff by Antonio and WavespeedAI team to get FLUX-dev inference (SOTA diffusion) in under a second on a B200. datacrunch.io/blog/flux-on...

medium.com/@NoamShazeer... I came across this post from one of the attention GOATs (thnx @conorhassan.bsky.social for term). I know there are more complicated solutions to this but I find this quite a simple solution to implement to keep track of shapes.

1/ Introducing ACE (Amortized Conditioning Engine)! Our new AISTATS 2025 paper presents a transformer framework that unifies tasks from image completion to BayesOpt & simulator-based inference under *one* probabilistic conditioning approach. It's Bayes all the way down!

Multi-Head Latent Attention vs Group Query Attention: We break down why MLA is a more expressive memory compression technique AND why naive implementations can backfire. Check it out!

Registration is open for the following ELLIS Summer Schools: 🇨🇭 Unit Lausanne / @epfl-ai-center.bsky.social on ML & AI for Health 🏴󠁧󠁢󠁥󠁮󠁧󠁿 Unit Manchester on ML for Healthcare & Biology 🇩🇪 @ellisunitjena.bsky.social on AI for Earth & Climate Sciences Get links for more details and registration:

You still have a chance to submit your work to all tracks for AABI. New deadline is February 14 for both Workshop and Proceedings track! #AABI2025 #ML #ICLR2025 #Stats #Bayes

🚀 It's time for a new JAX ecosystem library! This time quite a small one: ESM2quinox. A #JAX + Equinox implementation of the ESM2 protein language model. GitHub: github.com/patrick-kidg... SciML is obviously my whole jam. My open source has largely focused on the differential equations ... [1/2]

📢PSA: #NeurIPS2024 recordings are now publicly available! The workshops always have tons of interesting things on at once, so the FOMO is real😵‍💫 Luckily it's all recorded, so I've been catching up on what I missed. Thread below with some personal highlights🧵

2/2 Papers accepted at #ICLR2025. Congrats to all my co-authors 🥳. Definitely check out these works if you're interested in fine-tuning/composing diffusion models! Papers in thread 🧵 below 👇

I am hiring in my machine learning group, DL February 2, 2025 Postdocs in Helsinki: www.aalto.fi/en/open-posi... Topic: combinations of the following: - foundations of (prob) ML - beyond current distribution shift models - AI-assistance with computationally rational user models - appl: eng bio etc

ModernBERT is officially released on Transformers v4.48.0. You no longer need to install from git to use. If you are plugging ModernBERT into an existing encoder finetuning pipeline, try increasing the learning rate. We've found that ModernBERT tends to prefer a higher LR than older models.

I'm hiring one summer research intern (BSc or MSc level) to work on some part of Bayesian workflow at Aalto (Finland). See my publication and talk list at users.aalto.fi/~ave/ for what we've been doing so far, and the summer topic would be related Apply at www.aalto.fi/en/aalto-sci...

I'm going to slowly repost my math notes from the other site🐦 here🦋; it's the only thing I posted over there that I think may have some long-term value & worth not deleting. These started out as notes for myself, but people seem to appreciate them. 😅 I'll keep track of all of them in this thread.

📢 My team at Meta (including Yaron Lipman and Ricky Chen) is hiring a postdoctoral researcher to help us build the next generation of flow, transport, and diffusion models! Please apply here and message me: www.metacareers.com/jobs/1459691...