When and why do modular representations emerge in neural networks? @stefanofusi.bsky.social and I posted a preprint answering this question last year, and now it has been extensively revised, refocused, and generalized. Read more here: doi.org/10.1101/2024... (1/7)

once again being driven insane by ML conference submissions

Very much looking forward to this series of workshops on the computational ingredients of reasoning. We have an amazing lineup of speakers from diverse backgrounds, and there will be lots of opportunities for discussion. Please consider attending!

🧠 Ingrédients computationnels du raisonnement : participez au prochain semestre thématique #IVADO qui se tiendra à Montréal de janvier à mars 2026. ➡️ Inscriptions event.fourwaves.com/fr/thematics... @glajoie.bsky.social @taylorwwebb.bsky.social @lampinen.bsky.social

For those attending the #CogInterp workshop at NeurIPS, please check out our work on visual symbolic mechanisms led by @rassouel.bsky.social and @thisisadax.bsky.social. We find that visual feature binding in VLMs is supported by emergent symbolic mechanisms.

1/ Why does RL struggle with social dilemmas? How can we ensure that AI learns to cooperate rather than compete? Introducing our new framework: MUPI (Embedded Universal Predictive Intelligence) which provides a theoretical basis for new cooperative solutions in RL. Preprint🧵👇 (Paper link below.)

1/9 🚨Thrilled to share "Caption This, Reason That", a #NeurIPS2025 Spotlight! 🔦 Meet us at #2112, 3 Dec 11 a.m. We analyze VLM limitations through the lens of Cognitive Science (Perception, Attention, Memory) and propose a simple "Self-Captioning" method that boosts spatial reasoning by ~18%. 🧵👇

Our paper “The cost of thinking is similar between large reasoning models and humans” is now out in PNAS! 🤖🧠 w/ @fepdelia.bsky.social, @hopekean.bsky.social, @lampinen.bsky.social, and @evfedorenko.bsky.social Link: www.pnas.org/doi/10.1073/... (1/6)

LLMs are trained to compress data by mapping sequences to high-dim representations! How does the complexity of this mapping change across LLM training? How does it relate to the model’s capabilities? 🤔 Announcing our #NeurIPS2025 📄 that dives into this. 🧵below #AIResearch #MachineLearning #LLM

this paper takes me by surprise a bit. of coz, we all know Ned's been thinking along these lines for decades: philpapers.org/rec/BLOBVC but is he really gonna seriously publish a new paper on this now, given all the AI hype & debates re: how unscientific some popular views on C are these days? 1/

Very excited to share that our work (together with co-first author Shanka Subhra Mondal and @neuroai.bsky.social ) on a brain-inspired architecture for planning with LLMs is now out in Nature Communications! www.nature.com/articles/s41... (thread below)

Very nice commentary arguing that binding is still a problem, for both biological and artificial neural networks www.sciencedirect.com/science/arti...

Very happy to announce that our paper “Sensory Horizons and the Functions of Conscious Vision” is now out as a target article in BBS!! @smfleming.bsky.social and I present a new theory of the evolution and functions of visual consciousness. Article here: doi.org/10.1017/S014.... A (long) thread 🧵

🔍 Large language models, similar to those behind ChatGPT, can predict how the human brain responds to visual stimuli New study by @adriendoerig.bsky.social @freieuniversitaet.bsky.social with colleagues from Osnabrück, Minnesota and @umontreal-en.bsky.social Read the whole story 👉 bit.ly/3JXlYmO

Very excited to release a new blog post that formalizes what it means for data to be compositional, and shows how compositionality can exist at multiple scales. Early days, but I think there may be significant implications for AI. Check it out! ericelmoznino.github.io/blog/2025/08...

Our new preprint explores how advances in AI change how we think about the role of symbols in human cognition. As neural networks show capabilities once used to argue for symbolic processes, we need to revisit how we can identify the level of analysis at which symbols are useful.

New position paper! We argue that symbolic and neural network models are not in opposition to each other, but occupy different levels of analysis, and also outline a new research agenda for better understanding the relationship between them. Please check out the paper / thread for more details!

Can LLMs reason by analogy like humans? We investigate this question in a new paper published in the Journal of Memory and Language (link below). This was a long-running but very rewarding project. Here are a few thoughts on our methodology and main findings. 1/9

Is the Language of Thought == Language? A Thread 🧵 New Preprint (link: tinyurl.com/LangLOT) with @alexanderfung.bsky.social, Paris Jaggers, Jason Chen, Josh Rule, Yael Benn, @joshtenenbaum.bsky.social, ‪@spiantado.bsky.social‬, Rosemary Varley, @evfedorenko.bsky.social 1/8

My first, first author paper, comparing the properties of memory-augmented large language models and human episodic memory, out in @cp-trendscognsci.bsky.social! authors.elsevier.com/a/1lV174sIRv... Here’s a quick 🧵(1/n)

After five years of confused staring at Greek letters, it is my absolute pleasure to finally share our (with @smfleming.bsky.social) computational model of mental imagery and reality monitoring: Perceptual Reality Monitoring as Higher-Order inference on Sensory Precision ✨ osf.io/preprints/ps...

Very excited for this symposium! We have an amazing lineup of speakers exploring the intersection between cog sci and mechanistic interpretability. If you’re at CogSci and interested in the ways in which mechanisms in LLMs might inform cognitive theories, please check it out!

📯 Come visit our #ICML25 Spotlight Poster and meet @taylorwwebb.bsky.social to discuss our work: "Toward an Algorithmic Evaluation and Understanding of Generative AI." Paper: openreview.net/forum?id=eax... Poster: icml.cc/media/Poster...

If you're at ICML, check our work on AlgEval, toward algorithmic understanding of generative AI. I couldn't make it in person but am excited to say @taylorwwebb.bsky.social is there presenting our spotlight paper. P.S. If you see Taylor congratulate him on his professorship! arxiv.org/abs/2507.07544

Come check out our poster session at 11 tomorrow to find out how LLMs approximate symbol systems for abstract reasoning icml.cc/virtual/2025...

Excited to announce the first workshop on CogInterp: Interpreting Cognition in Deep Learning Models @ NeurIPS 2025! 📣 How can we interpret the algorithms and representations underlying complex behavior in deep learning models? 🌐 coginterp.github.io/neurips2025/ 1/4

Excited to share a new preprint w/ @annaschapiro.bsky.social! Why are there gradients of plasticity and sparsity along the neocortex–hippocampus hierarchy? We show that brain-like organization of these properties emerges in ANNs that meta-learn layer-wise plasticity and sparsity. bit.ly/4kB1yg5

Super excited to share this one!! Meta-learning sparsity and learning rate gives rise to brain-like gradients of complementary learning systems. So complementary learning systems emerge organically through behavior optimization, and it's not just two of them!!

Exciting new preprint from the lab: “Adopting a human developmental visual diet yields robust, shape-based AI vision”. A most wonderful case where brain inspiration massively improved AI solutions. Work with @zejinlu.bsky.social @sushrutthorat.bsky.social and Radek Cichy arxiv.org/abs/2507.03168

🧠 Can a neural network build a spatial map from scattered episodic experiences like humans do? We introduce the Episodic Spatial World Model (ESWM)—a model that constructs flexible internal world models from sparse, disjoint memories. 🧵👇 [1/12]

Excited to share that our work on emergent symbolic mechanisms will be presented at ICML. Please check out the paper, where we have *significantly* expanded the results. In short, we find evidence for emergent symbol processing across several models and tasks (details below).

Pleased to share our ICML Spotlight with @eberleoliver.bsky.social, Thomas McGee, Hamza Giaffar, @taylorwwebb.bsky.social. Position: We Need An Algorithmic Understanding of Generative AI What algorithms do LLMs actually learn and use to solve problems?🧵1/n openreview.net/forum?id=eax...

Transformer-based neural networks achieve impressive performance on coding, math & reasoning tasks that require keeping track of variables and their values. But how can they do that without explicit memory? 📄 Our new ICML paper investigates this in a synthetic setting! 🎥 youtu.be/Ux8iNcXNEhw 🧵 1/13

Thrilled to share the new paper from the lab out today in @nathumbehav.nature.com, led by the great @jetrach.bsky.social! "Mental graphs structure the storage and retrieval of visuomotor associations" www.nature.com/articles/s41...

Now out in PNAS Nexus, a response to critiques of emergent analogy in LLMs. We find that difficulty on some ‘counterfactual’ task variants is best explained by auxiliary task demands, and that code-augmented models address these limitations. academic.oup.com/pnasnexus/ar...

Excited to share a new preprint w/ @michael-lepori.bsky.social & Michael Franke! A dominant approach in AI/cogsci uses *outputs* from AI models (eg logprobs) to predict human behavior. But how does model *processing* (across layers in a forward pass) relate to human real-time processing? 👇 (1/12)

Preprint Alert 🚀 Can we simultaneously learn transformation-invariant and transformation-equivariant representations with self-supervised learning? TL;DR Yes! This is possible via simple predictive learning & architectural inductive biases – without extra loss terms and predictors! 🧵 (1/10)

Top-down feedback is ubiquitous in the brain and computationally distinct, but rarely modeled in deep neural networks. What happens when a DNN has biologically-inspired top-down feedback? 🧠📈 Our new paper explores this: elifesciences.org/reviewed-pre...

[📄] Are LLMs mindless token-shifters, or do they build meaningful representations of language? We study how LLMs copy text in-context, and physically separate out two types of induction heads: token heads, which copy literal tokens, and concept heads, which copy word meanings.

New work out now in TMLR: Evaluating Compositional Scene Understanding in Multimodal Generative Models. We find that multimodal models have improved in their abilities to both generate and interpret compositional visual scenes, but still lag behind humans. arxiv.org/abs/2503.23125

Reasoning Algorithms: Online workshop open to all, with a bunch of speakers from cognitive development, primate cognition, neuroscience, and computational modeling 🧪🧠 obssr.od.nih.gov/news-and-eve...

Reinforcement learning has led to amazing breakthroughs in reasoning (e.g., R1), but can it discover truly new behaviors not already present in the base model? A new paper with Zak Mhammedi and Dhruv Rohatgi: The Computational Role of the Base Model in Exploration arxiv.org/abs/2503.07453

PhD position in cognitive computational neuroscience! Join us, & investigate how we can endow domain-specific models of vision (eg DNNs) with domain-general processes such as metacognition or working memory. All details => www.kuleuven.be/personeel/jo... #PsychSciSky #Neuroscience #Neuroskyence

Why Tononi et al's defense of IIT fails to convince me. medium.com/@kording/86f...

I am happy to be a signatory to this updated critique of the integrated information theory (IIT) of consciousness. Despite much media attention, I agree that its 'core claims are untestable even in principle' and it is therefore unscientific. www.nature.com/articles/s41...

New version of "the letter" in Nature Neuroscience. Like many others in the field, I signed because I believe that IIT threatens to deligitimize the scientific study of consciousness: www.nature.com/articles/s41....

LLMs have shown impressive performance in some reasoning tasks, but what internal mechanisms do they use to solve these tasks? In a new preprint, we find evidence that abstract reasoning in LLMs depends on an emergent form of symbol processing arxiv.org/abs/2502.20332 (1/N)

I’m very excited to finally see this one in print! Led by the incomparable @brissend.bsky.social www.nature.com/articles/s41... We find that cognitive processes (e.g. attention, working memory) undergo error-based adaptation in a manner reminiscent of sensorimotor adaptation. Read on! (1/n)

1/13 New Paper!! We try to understand why some LMs self-improve their reasoning while others hit a wall. The key? Cognitive behaviors! Read our paper on how the right cognitive behaviors can make all the difference in a model's ability to improve with RL! 🧵

preprint updated - www.biorxiv.org/content/10.1... Each of us perceives the world differently. What may underlie such individual differences in perception? Here, we characterize the lateral prefrontal cortex's role in vision using computational models ... 1/ 🧠📈 🧠💻