What a pleasant surprise! Thanks for covering our work 🙈 "Help Me Help the AI": Understanding How Explainability Can Support Human-AI Interaction Sunnie S. Y. Kim ☀️ Elizabeth Anne Watkins, PhD Olga Russakovsky Ruth Fong Andrés Monroy-Hernández Learn more 👉 sunniesuhyoung.github.io/XAI_Trust/

A long-held belief in Comp-Neuro is that as DNNs improve at object recognition, they will also become better models of object-selective inferotemporal (IT) cortex neurons. In our new paper, we find that this is no longer the case. Read on to learn more! serre-lab.github.io/neural_harmoni…

✨NEW PREPRINT✨ Visual cognition in the brain is dynamic. Time to consider time 🥁 in models! We present a novel human-like reaction time metric computed from stable recurrent vision models and study 𝘵𝘦𝘮𝘱𝘰𝘳𝘢𝘭 human-model alignment. Read on…🤓 arxiv.org/abs/2306.11582 1/n

📅 Last month, we presented 𝗘𝗩𝗔 at #CVPR2023 the first attribution method using 𝐅𝐨𝐫𝐦𝐚𝐥 𝐌𝐞𝐭𝐡𝐨𝐝𝐬! We leverage recent advances in formal methods to propagate bounds through a neural network, exploring a potentially infinite number of perturbations. 🧵

I am at #ICML2023 to present my latest work. Is the human performance better than that of diffusion models on the one-shot drawings task ? Attend my oral presentation today to have the answer ! More details below : x.com/VictorBoutin/s…

Mark your calendars: The application portal of our #ELLISPhD Program will open in October! Reach many European top #ML labs with only a single application & benefit from exchanges in our network! #JoinELLISforEurope More ➡️ ellis.eu/news/ellis-phd… #PhD #AI #PhDProgram Elise AI

👋 Explain big vision model with 𝐂𝐑𝐀𝐅𝐓 🪄🐰 A method that 𝙖𝙪𝙩𝙤𝙢𝙖𝙩𝙞𝙘𝙖𝙡𝙡𝙮 extracts the most important concepts for your favorite pre-trained vision model. e.g., we automatically discover the most important concepts on a ResNet50 for rabbits: eyes, ears, fur. 🧶

Enjoyed writing this Spotlight article for Trends in Cognitive Sciences w/ Claudia Damiano on @WilmaBrainbridge and Trent’s PNAS paper. We share our excitement for a collaboration between artists and AI to better understand memorability. authors.elsevier.com/a/1hk5N4sIRvPN…

Exciting news! Our paper was accepted as a **Spotlight** to #NeurIPS2023. We compute a human-like reaction time metric from stable recurrent vision models. Check out the 🧵below! arxiv.org/abs/2306.11582 w/ Lakshmi Govindarajan Alekh Karkada Ashok Aarit Ahuja David Sheinberg and Thomas Serre

👋👨‍🍳🍵 After a year of cooking up a secret project, I'm thrilled to officially reveal: The 𝐋𝐄𝐍𝐒 𝐏𝐫𝐨𝐣𝐞𝐜𝐭. By combining modern tools of Explainable AI, how much can we explain a ResNet50? 🧶

David Picard Caroline Petitjean Yann LeCun Yes. Check out MaCo which specifically makes these visualizations work for ViTs arxiv.org/abs/2306.06805 Also their accompanying website is pretty cool: serre-lab.github.io/Lens/ Thomas Fel Thomas Serre

Woohoo! 🎉 It's not every day your work gets a shoutout from the one and only Lucas Beyer (bl16)! I must mention that our visualizations were particularly spectacular on FlexiViT (and other Big Vision models as well); they have a secret recipe... github.com/google-researc…

🎭Recent work shows that models’ inductive biases for 'simpler' features may lead to shortcut learning. What do 'simple' vs 'complex' features look like? What roles do they play in generalization? Our new paper explores these questions. arxiv.org/pdf/2407.06076 #Neurips2024

🚀🎉 Thrilled to share our #Neurips2024 paper: “Latent Representation Matters: Human-like Sketches in One-shot Drawing Tasks”! We pit humans vs. regularized Latent Diffusion Models in the one-shot drawing task. Who’s the best sketch master? 🖌️🤖 (1/5) arxiv.org/abs/2406.06079

I’ll be at NeurIPS Conference this year, sharing some work on explainability and representations. If you’re attending and want to chat, feel free to reach out !👋

🌌🛰️Wanna know which features are universal vs unique in your models and how to find them? Excited to share our preprint: "Universal Sparse Autoencoders: Interpretable Cross-Model Concept Alignment"! arxiv.org/abs/2502.03714 (1/9)

Phenomenology → principle → method. From observed phenomena in representations (conditional orthogonality) we derive a natural instantiation. And it turns out to be an old friend: Matching Pursuit! 📄 arxiv.org/abs/2506.03093 See you in San Diego, NeurIPS Conference 🎉

🕳️🐇Into the Rabbit Hull – Part I (Part II tomorrow) An interpretability deep dive into DINOv2, one of vision’s most important foundation models. And today is Part I, buckle up, we're exploring some of its most charming features.

I am starting a venture on top of LeRobot! We’re at a pivotal time. AI is moving beyond digital to the physical world. Embodied AI will change our surroundings in ways we can barely imagine. This technology holds the potential to empower everyone. It must not be controlled by

🕳️🐇Into the Rabbit Hull – Part II Continuing our interpretation of DINOv2, the second part of our study concerns the geometry of concepts and the synthesis of our findings toward a new representational phenomenology: the Minkowski Representation Hypothesis