Thomas Fel
@Napoolar
Followers
1K
Following
10K
Media
74
Statuses
1K
Explainability, Computer Vision, Neuro-AI @Harvard. Research Fellow @KempnerInst. Prev. @tserre lab, @Google, @GoPro. Crêpe lover.
Boston, MA
Joined February 2017
Train your vision SAE on Monday, then again on Tuesday, and you'll find only about 30% of the learned concepts match. ⚓ We propose Archetypal SAE which anchors concepts in the real data’s convex hull, delivering stable and consistent dictionaries.
6
79
356
RT @livgorton: What if adversarial examples aren't a bug, but a direct consequence of how neural networks process information?. We've found….
0
41
0
Check out @fenildoshi009's work! .Love the 20° tuning finding: models finetuned at this curvature generalize to unseen curvatures, rediscovering the old Geisler results from visual neuroscience! 👀.
🧵 Can a purely feedforward network — with no recurrence or lateral connections — capture human-like perceptual organization?. 🤯 Yes! Especially for contour integration, and we pinpoint the key inductive biases. New paper in @PLOSCompBiol with @talia_konkle & @grez72!. 1/24
0
2
18
RT @_onionesque: Looking at the thread. The common frame to look at the more general phenomenon involves an eigenproblem of the form Oƒ = λ….
0
25
0
RT @davidbau: This Friday NEMI 2025 is at Northeastern in Boston, 8 talks, 24 roundtables, 90 posters; 200+ attendees. Thanks to @GoodfireA….
0
12
0
RT @raphaelmilliere: Transformer-based neural networks achieve impressive performance on coding, math & reasoning tasks that require keepin….
0
100
0
RT @sukrutrao: Are you an XAI researcher attending #ICCV2025? Submit your recently published work — at CVPR, ICCV, ECCV, NeurIPS, ICML, IC….
0
10
0
RT @SabineMuzellec: Very proud of our new preprint introducing reverse predictivity — a two-way test of AI–brain alignment. We find a strik….
0
4
0
RT @KempnerInst: #CCN2025 begins tomorrow! Check out the presentations from the #kempnerinstitute. #NeuroAI #neuroscience .
0
4
0
RT @jeffreycider: nn layers align their singular vectors. each matrix syncs to its neighbor, its rotation neatly clicking into the basis di….
0
21
0
RT @GretaTuckute: Can't wait for #CCN2025! Drop by to say hi to me / collaborators!. @bkhmsi @yingtian80536 @NeuroTaha @ABosselut @martin_….
0
10
0
RT @KempnerInst: New in the #DeeperLearningBlog: @shunnnli describes the discovery of a type of synapse that switches between more excitato….
kempnerinstitute.harvard.edu
Synaptic plasticity rules in the brain are normally thought of as changing synaptic weights but not signs, unlike artificial neural networks. We show that a type of synapse in the […]
0
2
0
RT @CSProfKGD: Introducing Dr. Kowal! It was truly a family affair. Matt had his parents, fiancée, in-laws, aunts/uncles. I had my own PhD….
0
1
0
RT @AndrewLampinen: In neuroscience, we often try to understand systems by analyzing their representations — using tools like regression or….
0
61
0
RT @WenjieWLi: (1/3) Presenting at #CogSci2025 TODAY (Aug 2; Poster P3-W-211): "Expanded Information Capacity in Neural Networks Predicts t….
0
11
0
RT @AFornito: Interested in cortical wave dynamics? Check out our new preprint led by J Cruddas with.@jchrispang where we review the physic….
0
50
0
RT @livgorton: Come work with me on an interp project this fall via SPAR! I had a really fun time mentoring in the spring and am really exc….
sparai.org
SPAR connects rising talent with AI safety expertise through structured mentorship and impactful research projects. Apply to work on cutting-edge AI safety research.
0
6
0
Chatted with Le Monde about interpretability and sparse autoencoders. (Yes, SAE made it into mainstream news 😅) . . Merci à Nicolas Six pour l’échange !.
lemonde.fr
Les rouages des robots conversationnels demeurent très opaques, mais des chercheurs commencent à localiser les « neurones » qui stockent les informations et prennent des décisions-clés.
4
8
57