Calling all #SIGGRAPH2026 film editing buffs! 🎬 Come check out our FilmGPT talk! 🕝 2:40 PM PDT 📍 Room 408B

Shiry Ginosar@shiryginosar.bsky.social · 2w ago

Introducing FilmGPT 🎬 (#SIGGRAPH2026) Can we learn the grammar of film directly from movies? We train an autoregressive transformer on feature films to edit raw footage into coherent cinematic sequences, without generating new pixels. Project page: mudtriangle.com/filmgpt Demo below👇

Introducing FilmGPT 🎬 (#SIGGRAPH2026) Can we learn the grammar of film directly from movies? We train an autoregressive transformer on feature films to edit raw footage into coherent cinematic sequences, without generating new pixels. Project page: mudtriangle.com/filmgpt Demo below👇

Fans of the Platonic Representation Hypothesis argue that all representations converge. However, we find that the current experimental evidence for the PRH is surprisingly fragile. Instead, models may develop their own “Umwelten” (distinct, model-specific views of the world).

A. Sophia Koepke@askoepke.bsky.social · 4mo ago

New paper: Back into Plato’s Cave Are vision and language models converging to the same representation of reality? The Platonic Representation Hypothesis says yes. BUT we find the evidence for this is more fragile than it looks. Project page: akoepke.github.io/cave_umwelten/ 1/9

Great to see corroborating evidence to our motion-forecasting.github.io work from other groups! Check out this concurrent great work from @stefanabaumann.bsky.social, @jannik-w.bsky.social, @tommymarto.bsky.social, Mahdi M. Kalayeh, and Björn Ommer (@compvis.bsky.social)

Stefan Baumann@stefanabaumann.bsky.social · 4mo ago

You don't imagine the future by mentally rendering a movie. You trace how things move -- abstractly, sparsely, step by step. We built a model that does exactly this. It predicts motion, not pixels -- and it's 3,000× faster than video world models. Myriad, accepted at @cvprconference.bsky.social

riddle for kids that combines first/third person perspectives: where are the firefighters on the map?

We take a step toward behavior modeling in the wild: general non-rigid motion forecasting! The coolest part: training on diverse animal behavior is enough to generalize to completely OOD motion. Amazing work from @neerjathakkar.bsky.social, with @carldoersch.bsky.social's technical know-how!

Neerja Thakkar@neerjathakkar.bsky.social · 4mo ago

What’s the right representation for a world model? 3D, pixels, or something else? Excited to release our new paper “Forecasting Motion in the Wild” where we propose point tracks as tokens for generating complex non-rigid motion and behavior From Google Deepmind, UC Berkeley BAIR, and TTIC

🚨 Can real-world human experiences shape the future of generative AI and computer vision systems? Find out at Humans of Generative AI (HuG) workshop at #CVPR2026 🚨 Accepting 1-page extended abstracts for lightning talks and posters 🗓 Deadline: April 10, 2026 See: humansofgenerativeai.github.io

💫 I am recruiting exceptional PhD students & postdocs for my lab @tticconnect.bsky.social this year! Application details: www.ttic.edu/studentappli...

TTIC is hiring tenure track faculty! Come join us in Chicago! I am sadly not at ICCV but Greg Shakhnarovich (Vision) home.ttic.edu/~gregory/ and Matt Walter (Robotics) home.ttic.edu/~mwalter/ are around and would be happy to chat.

TTIC@tticconnect.bsky.social · 10mo ago

TTIC is hiring tenure-track faculty! Join a top #CS research environment with exceptional PhD students, strong research support, and a uniquely light teaching load. Apply by Dec 1: buff.ly/wv1sA0i #CSJobs #FacultyHiring

@alisongopnik.bsky.social's talk happening now!! Ballroom B @iccv.bsky.social

Shiry Ginosar@shiryginosar.bsky.social · 10mo ago

Guest track 2: Our one and only KiVA Challenge!!! kiva-challenge.github.io Veo can (almost...) do it!!! can you?? video-zero-shot.github.io With @euniceyiu.bsky.social, Anisa Noor Majhi, Maan Qraitem, Kate Saenko, @alisongopnik.bsky.social

Join us TODAY for the 3rd Perception Test Challenge perception-test-challenge.github.io @iccv.bsky.social Ballroom B, Full day Amazing lineup of speakers: Ali Farhadi, @alisongopnik.bsky.social, Phlipp Krahenbul, @phillipisola.bsky.social

TODAY! Artificial Social Intelligence Workshop @iccv.bsky.social Room 317B, Full day Social reasoning, multimodality, and embodiment! Speakers: Evonne Ng, @tianminshu.bsky.social @hyunwoo-kim.bsky.social, @diyiyang.bsky.social ang.bsky.social, @hokulabs.bsky.social, @michael-j-black.bsky.social

KiVA (Kid-inspired Visual Analogies) Challenge Test Phase is NOW LIVE (Sep 1–Oct 6)! Can your model reason like a child? Can it beat adults? 🥇 $1,000 🥈 $500 for 2 runner-ups Join/submit: t.co/zQwA1Nmohy And join us at @iccv.bsky.social in Hawaii!! 🌴

I am giving a talk this morning at 10:40AM PST as part of the #ICML2025 Workshop on Assessing World Models. Title: "What Do Vision and Vision-Language Models Really Know About the World?" Come join us! www.worldmodelworkshop.org

ICML Workshop on Assessing World Models

Date: Friday, July 18 2025 Time: 8:45am - 5:15pm (Pacific Time) Location: West Ballroom B at ICML 2025 in Vancouver, Canada (Same Floor as Registration)

worldmodelworkshop.org

Join us for 3rd Perception Test Workshop &Challenge @iccv.bsky.social #iccv2025 *NEW* this year: - 3 unified tracks - novel interpretability track - guest tracks: KiVA and Physics-IQ - 4 world-class speakers (see pic) Up to 50K in prizes sponsored by Google DeepMind 🧵 for details [1/4]

🧠How “old” is your model? Put it to the test with the KiVA Challenge: a new benchmark for abstract visual reasoning, grounded in real developmental data from children and adults. 🏆 Prizes: 🥇$1K to the top model 🥈🥉$500 📅 Deadline: 10/7/25 🔗 kiva-challenge.github.io @iccv.bsky.social

When it comes to goal-directed work, people prioritize controllable variability (a.k.a. empowerment!). But in undirected play, we shift toward embracing pure variability. Check out our forthcoming Phil. Trans. A (2026) paper!

Eunice Yiu@euniceyiu.bsky.social · last yr.

Children & adults don't just explore—they seek empowerment: controllable and variable interventions. This intrinsic drive supports causal learning & generalization. 📄 Forthcoming in Phil. Trans. A (2026): osf.io/ept4n/download With @alisongopnik.bsky.social, @shiryginosar.bsky.social & Kelsey Allen

Check out our new paper at #ICLR2025, where we show that multi-task neural decoding is both possible and beneficial. As well, the latents of a model trained only on neural activity capture information about brain regions and cell-types. Step-by-step, we're gonna scale up folks! 🧠📈 🧪 #NeuroAI

Mehdi Azabou@mehdiazabou.bsky.social · last yr.

Scaling models across multiple animals was a major step toward building neuro-foundation models; the next frontier is enabling multi-task decoding to expand the scope of training data we can leverage. Excited to share our #ICLR2025 Spotlight paper introducing POYO+ 🧠 poyo-plus.github.io 🧵

🎧 Listen to the podcast! Professor @alisongopnik.bsky.social and @newamerica.org CEO @slaughteram.bsky.social spoke with @alexis-madrigal.bsky.social about how rethinking our approach to caregiving and how we support care providers could lead to a better society. 🔗:

Alison Gopnik and Anne-Marie Slaughter on Why We’re Not Paying Enough Attention to Caregiving

KQED's Forum · Episode

buff.ly

Think LMMs can reason like a 3-year-old? Think again! Our Kid-Inspired Visual Analogies benchmark reveals where young children still win: ey242.github.io/kiva.github.... Catch our #ICLR2025 poster today to see where models still fall short! Thurs. April 24 3-5:30 pm Halls 3 + 2B #312

Neuroscience is finally taking more and more baby steps towards running experiments at scale!

Cole Hurwitz@colehurwitz.bsky.social · last yr.

Another step toward a foundation model of the mouse brain: "Neural Encoding and Decoding at Scale (NEDS)" Trained on neural and behavioral data from 70+ mice, NEDS achieves state-of-the-art prediction of behavior (decoding) and neural responses (encoding) on held-out animals. 🐀

In his new book, published today, Nachum Ulanovsky calls on the field to embrace naturalistic conditions and move away from overcontrolled experiments. #neuroskyence www.thetransmitter.org/systems-neur...

‘Natural Neuroscience,’ an excerpt

In his new book, Nachum Ulanovsky calls on the field to embrace naturalistic conditions and move away from overcontrolled experiments.

thetransmitter.org

Welcome to TTIC!! We are so excited to have you join us!!

Nick Tomlin@nickatomlin.bsky.social · last yr.

Writing my first post here to announce that I've accepted an assistant professor job at TTIC! I'll be starting in Fall 2026, and recruiting students this upcoming cycle. Until then, I'll be wrapping up the PhD at Berkeley, and this summer I'll join NYU as a CDS Faculty Fellow 🏙️

We're very excited to introduce TAPNext: a model that sets a new state-of-art for Tracking Any Point in videos, by formulating the task as Next Token Prediction. For more, see: tap-next.github.io

This shows the importance of expressive (rather than functional) gesture really nicely in a beautiful non-anthropomorphic robot design machinelearning.apple.com/research/ele...

ELEGNT: Expressive and Functional Movement Design for Non-Anthropomorphic Robot

Nonverbal behaviors such as posture, gestures, and gaze are essential for conveying internal states, both consciously and unconsciously, in…

machinelearning.apple.com

Did Italo Calvino discover bag of words and topic models in 1979?

"That's what the people promulgating these horrible policies want - a bored, indifferent public who figures that who cares, nothing matters any more, it's gonna happen no matter what. But it doesn't have to. Never forget that: it doesn't have to happen." www.science.org/content/blog...

What's Happening Inside the NIH and NSF

science.org

Why does Western paleolithic cave art strongly prefer animal side views and often abbreviations? Our new #eSymb preprint (osf.io/preprints/ps... w Pagnotta, Psujek, Mendoza Straffon and Tylén ) challenges long-held assumptions about these artistic choices based on cogsci experiments. 1/

New paper! A self-supervised object-centric 2.1D image representation using 3D Gaussians, extending MAE with a Gaussian bottleneck. While Gaussian splatting has been used for single-scene reconstruction, we’re the first to apply it to image representation learning! brjathu.github.io/gmae/.

Can video MAE scale? Yes. Do you need language to scale video models? No. arxiv.org/abs/2412.15212 Great rigorous benchmarking from my colleagues at Google DeepMind.

Scaling 4D Representations

Scaling has not yet been convincingly demonstrated for pure self-supervised learning from video. However, prior work has focused evaluations on semantic-related tasks $\unicode{x2013}$ action classifi...

arxiv.org