🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

With co-authors Thomas Serre, Lakshmi Govindarajan, Rex, and Alekh Karkada Ashok (an exceptionally talented RA who is applying for grad school this cycle, in case you have any interest in recruiting PhD students who are brilliant-beyond-belief). Check out the work and get in touch!

Imagine trying to track a single fly in a swarm. How would you do it? Can machine vision do it? Does it need to? Introducing a new computational challenge + solution: Tracking without re-recognition in humans and machines. arxiv.org/abs/2105.13351

Shamelessly asking for your likes and retweets to help me get the word out for Audrey, who can’t yet do it herself. She is in need of a part of someone’s liver 🙏

Excited to present this work next week at NeurIPS 2021! In it, we ask how humans are able to track objects when they look identical, whether neural networks can do the same, and why this matters in the real world. papers.nips.cc/paper/2021/has… #NeurIPS2021

#VSS2022 I'll be presenting my work titled "Toward modeling visual routines of object segmentation with biologically inspired recurrent vision models" at 6.30 PM today, Talk Room 1. Come check it out :)

Every year I read a lot of grad school applications from accomplished people that don't give me the info I'm looking for. It feels like a major hidden curriculum thing. So here's (my opinion on) how to write a great Statement of Purpose/Research for a PhD program. 🧵 1/

Check out our new paper, to appear at NeurIPS. We show that DNNs are becoming progressively *less* aligned with human perception as their ImageNet accuracy increases. Ignore the elections, Elon, and FTX for a moment — this is important! serre-lab.github.io/Harmonization/

[1/5] Much of the progress on attribution methods has been driven by theoretical metrics --without much consideration for human end-users Our #NeurIPS22 paper investigates whether progress has translated to explanations more useful in real-world scenarios serre-lab.github.io/Meta-predictor

Cem Anil Cool! You should also check out our paper (Alekh Karkada Ashok Lakshmi Govindarajan Thomas Serre) from NeurIPS 2020: arxiv.org/abs/2005.11362 Similar insights with recurrent backprop: more processing time helps systematic generalization on a line-tracing task and extends to panoptic seg on MS-COCO.

Hi 👋 ! Interested in explainability or planning to use XAI in your future projects? Then you might be interested in this 🧵, let me introduce you to Xplique! github.com/deel-ai/xplique with Thomas Serre, Remi Cadene, Mathieu Chalvidal, Julien Colin, Louis Béthune, Paul Novello, ANITI Toulouse

Our article "Diffusion Models as Artist: Are we Closing the Gap between Humans and Machine" (arxiv.org/abs/2301.11722) has been accepted at #icml2023 and selected as an oral 🎉🎊! 🤖 = 👨🏻‍🎨 ??  (1/5) 🧵

The sensitivity of DNNs to adversarial attacks has long been thought to be an "Achilles heel" that will ultimately make them unsafe in real-world applications. Has this sensitivity changed as DNNs have scaled-up and rivaled or beat human-level accuracy? serre-lab.github.io/Adversarial-Al…

✨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. 🧵

👋 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. 🧶

Excited about this collaborative work between our group and the Sheinberg lab! Carney Institute for Brain Science

🎉 Come join us at some exciting events! TOMORROW: Pawtucket Brain Fair on March 9 (Hope Artiste Village, 1005 Main Street, 10AM-2PM) facebook.com/events/7503784… Brown Brain Fair on March 16 (Brown Engineering Research Center, 10AM-2PM)! facebook.com/events/2568766…

I'll be presenting this work Conference on Language Modeling on Monday afternoon, and I'll be hanging out in Philly until Wednesday! Feel free to reach out if you want to chat about mechanistic interpretability and/or cognitive science.

Our work on mental simulation in monkeys and RNNs just got published! This was a fantastic collaboration between the Sheinberg lab and Theresa on the experimental side and our group, with Alekh Karkada Ashok leading the computational work. Funded in part by a Zimmerman Innovation