Stefano Sarao Mannelli@stefsmlab

Assistant Professor @chalmersuniv and @goteborgsuni 🌱 ML + CogSci + StatPhys 🔎 @stefsm.bsky.social 🦋

 Gothenburg, Sweden
stefsmlab.github.io/
Joined January 2010
29 Photos and videos
📢 We’re now accepting applications for the 2026 School on Analytical Connectionism dedicated this year to Language Acquisition. 📍 Gothenburg, Sweden 🗓️ August 17–28, 2026 ☠️ Apply by April 17! 🔗 analytical-connectionism.net… 👇 Meet the experts joining us this summer!
1
15
42
10,407
Great news from the lab: our work (👇) has been accepted to NeurIPS, and we will present it in Sydney! See you there! 🦘🤖
Stefano Sarao Mannelli@stefsmlab
May 8
New preprint out! What happens when diffusion models are trained on heterogeneous data (different classes/populations)? Do they generalise and memorise everything at the same time? The answer is no, and it has a clear dependence on the distribution's properties. 🧵1/6
3
182
Stefano Sarao Mannelli retweeted
Gabriel Ocana-Santero
@OcanaSantero
Sep 30
Something that became super obvious working on our cross-species biomarkers is that epilepsy patients are very heterogeneous on neural dynamics. Some of them sit much closer to one of our mouse models than to the others. 1/5
1
1
2
481
Another great iteration of the Structured Learning Workshop! Looking forward to the discussions :) 📅 Oct 27-28 📍 Göteborg
Simon Olsson@smnlssn
Aug 25
Happy to announce this years Structured Learning Workshop, Oct 27–28 at Chalmers, Gothenburg! Exceptional speakers from academia & industry. Limited spots, apply to attend via the waiting list: chalmers.ungapped.io/Events/…
1
4
380
🙌 Looking forward to it!
Kanishka Misra 🌊@kanishkamisra
Aug 20
Tomorrow, I head to beautiful Gothenburg, Sweden, to lecture at Analytical Connectionism (analytical-connectionism.net…). In a shocking turn of events, I will talk about controlled rearing -- a topic I have never before presented!
Title slide for the talk "The Role of Input in Shaping Linguistic Generalization: Insights from Controlled Rearing of Language Models"

ALT Title slide for the talk "The Role of Input in Shaping Linguistic Generalization: Insights from Controlled Rearing of Language Models"
8
556
Stefano Sarao Mannelli retweeted
Soon Hoe Lim@Shoelim8
Jul 16
Exciting update! We are organizing the 2nd Workshop on Principles of Generative Modeling (PriGM) at #NeurIPS Paris. We have an amazing lineup of speakers ready for December. Paper submissions are welcome. We’d love to see your research! 🙏sites.google.com/view/prigmn…

PriGM@NeurIPS2026

Workshop summary Machine learning theory has long focused on classical supervised learning settings, where a model is trained on input–label pairs drawn from a well-defined data distribution, with...

sites.google.com
3
10
1,351
new preprint: how does curriculum learning affect representation formation in transformers, and why does it matter for AI safety? with Sebi Bruijns, Mia Whitefield, @RubruckJirko @akaijsa and @FazlBarez . arxiv.org/abs/2607.04846

Pretraining Curricula Enable Selective Fine-tuning

Transformers follow implicit curricula whereby some tasks are learned before others. However, how explicit pretraining curricula influence learning, generalization, and the selectivity of...

arxiv.org
3
42
5,861
Interested in any of our works? Come and say hi @icmlconf! Unfortunately I won't be around this year but my collaborators will answer all your questions :)
Cognition, Adaptation and Learning (CAandL) Lab@caandl_lab
Jul 6
We are excited to share that CAandL Lab will feature in 3 papers at #ICML2026 this week. Stop by and say hi if you are interested in any of these! @devonjarvi5, @stefsmlab, @geraudnt with @kleinric, @BenjaminRosman, Damien Harvey, Branden Ingram, and Steven James
1
2
8
1,366
Model collapse is often framed as “models getting worse” In our ICML Spotlight Position paper, we show a high risk of unequal degradation. Rare languages, minority viewpoints, and low-resource communities are likely to be affected first and most severely arxiv.org/abs/2605.04127

Position: the Stochastic Parrot in the Coal Mine. Model Collapse...

Model collapse, the degradation in performance that arises when generative models are trained on the outputs of prior models, is an increasing concern as artificially generated content...

arxiv.org
Devon Jarvis@devonjarvi5
Jun 10
I'm excited to share our position paper that has been accepted at ICML as a Spotlight paper. In this work we (@kleinric, @BenjaminRosman, Steven James and @stefsmlab) make a call to action for more focus on model collapse in the AI Fairness community arxiv.org/pdf/2605.04127
1
22
3,328
Stefano Sarao Mannelli retweeted
Alessandro Favero@alesfav
May 29
AI needs vastly more data than we do. One idea might close the gap: don't predict raw signals (tokens), predict your own abstract latent representation (JEPA, data2vec). With @DanKorchinski @MatthieuWyart, on a toy model, we prove how much that helps: the gap is exponential. 🧵
14
76
526
53,882
Stefano Sarao Mannelli retweeted
Devon Jarvis@devonjarvi5
May 28
I’m excited to share that our paper “Compositionality and systematicity emerge from iterated learning in deep linear networks” has been published at PNAS. This work was conducted with @kleinric @BenjaminRosman and @SaxeLab. Some highlights below. pnas.org/doi/full/10.1073/pn…

Compositionality and systematicity emerge from iterated learning in deep linear networks | PNAS

Humans have a remarkable ability to systematically generalize—reasoning about new situations by combining aspects of previous experiences. Language...

pnas.org
Wits University
@WitsUniversity
May 27
New research from the University of the Witwatersrand, South Africa, is shedding light on how language evolves, in both humans and artificial intelligence models. The study explores the role of culture and “iterated learning”, showing how language becomes more structured over generations in both human development and large-scale AI language models. 🔗 Read More: ow.ly/42Vr50Z4BjH #WitsForGood #WitsResearch #ResearchForGood
1
11
21
1,655
New preprint out! What happens when diffusion models are trained on heterogeneous data (different classes/populations)? Do they generalise and memorise everything at the same time? The answer is no, and it has a clear dependence on the distribution's properties. 🧵1/6
Flavio Nicoletti@flaNick94
May 8
We’re happy to share our latest work on diffusion models, now out as an arXiv preprint. During score learning, different classes in a dataset are generalized and memorized at different times. What drives these differences? Find out here: arxiv.org/abs/2605.06367
3
18
84
12,626
more replies
When we tested the theory on a U-Net trained on Fashion MNIST, it achieved incredibly good predictions! Even when comparing the RF model on a GMM matching the 1st and 2nd moments from Fashion MNIST, with the real training, the qualitative agreement is remarkable. 🧵5/6
1
2
2
218
Check out the full paper for more details (and insights)! 📄 arxiv.org/html/2605.06367 👯 = @flaNick94, @the_chma, Enrico Ventura, @saglietti_luca, me. 🧵6/6
1
3
235
Two Analytical Connectionism-related updates: 1. ⏰ 1 week left to apply! Interested in language + AI & cognition? Don’t miss it: analytical-connectionism.net… 2. 📜 Lecture notes from the first two editions are finally out: proceedings.mlr.press/v320/

2026 School on Analytical Connectionism

A 2-week summer course hosted at Chalmers University of Technology on analysing language acquisition and higher-level cognition.

analytical-connectionism.net
Stefano Sarao Mannelli@stefsmlab
Feb 18
📢 We’re now accepting applications for the 2026 School on Analytical Connectionism dedicated this year to Language Acquisition. 📍 Gothenburg, Sweden 🗓️ August 17–28, 2026 ☠️ Apply by April 17! 🔗 analytical-connectionism.net… 👇 Meet the experts joining us this summer!
5
14
2,633
Stefano Sarao Mannelli retweeted
Simon Olsson@smnlssn
Apr 8
Excited to share TITO is out today in Science Advances science.org/doi/10.1126/scia…
1
23
91
5,432
Deadline is on April 17th! Apply before it's too late ⏰
Andrew Saxe@SaxeLab
Apr 2
Very excited by this year's Analytical Connectionism Summer School! A dream lineup of speakers on the topic of language acquisition in minds and machines Bursaries available to cover costs Aug 17 – Aug 28, 2026 Gothenburg Details: analytical-connectionism.net…
2
217
Stefano Sarao Mannelli retweeted
Andrew Saxe@SaxeLab
Apr 2
Postdoc opening! Come work with us on deep learning theory relevant to AI safety Deadline: 7 Apr 2026 Details and application: ucl.ac.uk/work-at-ucl/search…+

UCL – University College London

UCL is consistently ranked as one of the top ten universities in the world (QS World University Rankings 2010-2022) and is No.2 in the UK for research power (Research Excellence Framework 2021).

ucl.ac.uk
36
130
12,227
Load more