Assistant professor @ Ben-Gurion University. Studies and tweets about human and machine vision.
Israel
Joined July 2016
- Tweets777
- Following896
- Followers1.1K
- Likes1.6K
Pinned Tweet
How can we design experiments that make models disagree? One section of our new @NatRevNeurosci Review with @KriegeskorteLab and @SchuettHeiko examines studies that used stimulus sets designed to elicit distinct predictions from competing models. Full text link at the end. 1/14
Tal Golan retweeted
Super excited to share that I'm co-founding Read/Write Labs for #NeuroAI applications to human brain health and human-aligned AI. We could not ask for a more wonderful team: prn.to/4yK9vqo
Tal Golan retweeted
1/ Can brain inspired self-supervised objectives lead to more robust visual representations in DNNs?
Excited to share our preprint with @adeli_hossein, Tian Zeng, @KriegeskorteLab, Ning Qian
biorxiv.org/content/10.64898…
Tal Golan retweeted
Meet kornia==0.9.0 - more correct, much faster, more documented, better half-precision, CUDA and MPS support.
More than 800 issues have been raised and fixed.
Some ops are now 2000x faster.
Others not crash :)
Way better docs. Check the release post
github.com/kornia/kornia/rel…
Tal Golan retweeted
Our review of Bayesian efficient coding as a theory of perception is out now in TICS – with Ivan Tomić and Máté Lengyel. cell.com/trends/cognitive-sc…
Tal Golan retweeted
Please vote for the Linton Color Constancy Illusion in this year's Illusion of the Year Competition!
Link: vote.illusionoftheyear.com/
From a NeuroAI perspective, it's really annoying that feedforward neural nets rule the world. Unrolling equivalence notwithstanding, what inductive biases does recurrence buy us? Looking forward to reading this.
1/5
New preprint: what does lateral recurrence add to visual representations?
Recurrence reuses the same weights over time: more computation, no new parameters.
Does that extra computation refine representations? If so, which ones?
doi.org/10.64898/2026.10.02.…
Tal Golan retweeted
Sami Mollard @MollardSami from Pieter Roelfsema's lab @Pieters_Tweet visited us at Brown @BrownCoPsy @CarneyInstitute, and the project brought together their work on incremental grouping in visual cortex with our recurrent vision models. New preprint:
arxiv.org/abs/2610.05419
Tal Golan retweeted
New article in the Chose You Stimulus Wisely issue bit.ly/47LzJxE by Thomas S. A. Wallis from the Technical University of Darmstadt describes a Python package to create ambiguous two-tone images.
doi.org/10.1167/jov.26.10.8
Tal Golan retweeted
Leading AI models appeared to predict brain activity equally well…
…until we asked them to synthesize images that precisely control neural firing, like these 👇
Across 25 neural sites, closed-loop experiments revealed striking hidden differences in model-brain alignment.
🧵
Tal Golan retweeted
Another update from our topography work from (high-schooler!) Navneet Prakash. Navneet trained path-integrating RNNs with TopoLoss and found that they recover a much more brain-like center-surround RF connectivity structure than baseline models (w/o any post-hoc sorting)!
Tal Golan retweeted
🚨Can a model recognise an object when its usual neighbours disappear?
Our #NeurIPS2026 paper, “Stranger Things: When Objects Appear Without Their Typical Neighbours” 👻, examines typicality bias across modern vision models. 1/8🧵
Tal Golan retweeted
JEP General finally published our paper about impressive recognition asymmetries. Ingredients: true and false "memorability" cooked with a good amount of signal-detection theory
psycnet.apa.org/record/2028-…
It was great to work w/ Nick Chiang and @WilmaBainbridge to make this happen
In a new preprint with Nick Chiang and @WilmaBainbridge, we show that our visual recognition memory is often asymmetric, and sometimes even hugely asymmetric. These asymmetries are consistent across people, and we know how to predict them
osf.io/preprints/psyarxiv/6n…
1/4
Tal Golan retweeted
Today I want to share an update from our topographic ANN work led by @haidertahan. What does adding topography actually do to a model’s internal representations? Does it make is more brain like? Introducing TopoAudio
Tal Golan retweeted
Check out @climatepaige latest podcast where we discuss exciting questions like how billions of neurons (in AIs and brains) learn to make sense of the world 🌎
Thanks so much for having me this was great fun, I learned a lot 🙏
We’ve built computers to do our thinking while we live our lives, but do they actually understand anything?
In this episode, @klindt_david explains to me how brains interpret the world — and whether AI can.
open.spotify.com/episode/3gr…
Tal Golan retweeted
I've always really hated this illustration of byte-pair encoding from my book, but can't figure out how to make it better. Anyone out there got some design skills and want to have a crack at it? DM me if you are interested.
Tal Golan retweeted
Why do neural networks become modular? Our paper is now out in Nature Machine Intelligence! @NaturePortfolio We show that multi-task learning promotes modularity under limited capacity and task demands help explain features of brain architecture in addition to the spatial constraints.
Great thanks to all the co-authors. @WHYWYH007 @yifannnwu @DaniSBassett @marcelomattar
rdcu.be/YJ12NBxvrHsa
