📣 Upcoming Workshop! Join the ELIAS, ELLIOT & ENFIELD TDW on Trustworthy AI to explore the technical, ethical, & environmental challenges of deploying reliable, fair, and sustainable AI systems. 📅 Mar 6 📍 Paris/Hybrid 🔗 https://bit.ly/4bMMiME @elliot-eu.bsky.social · @horizonenfield.bsky.social

@KL_Div repost after @hanshengchen acknowledge a connection 1/4 🧵 "Thanks for pointing out the similarity between drifting and Implicit Maximum Likelihood Estimation! I worked out the mathematical connection - the crux is that drifting fields are similar to the gradient of a soft version of the

@isskoro 1/5 🧵 "The recent Drifting Models paper from Kaiming's group got very hyped as a new generative modeling paradigm, but it's basically a scaled-up/generalized version of 2015 GMMN (authors acknowledge this in App. C.2 — GMMN = Drifting for particular kernel). Very skeptical about ...

Today I woke up with a sentiment particularly from far right/left & against institutions 🇪🇺 😡🤬🙃

Sergio Charles @eigentopology 1/10 "Thread: Drifting Models - The Math Behind Training-Time Evolution 🧵 New paradigm for generative modeling from Kaiming He's team. Let's break down the mathematics. Paper: arxiv.org/abs/2602.04770 ...

(according to you) What's the most enticing driver behind humanity pursuing AGI? 📽️ Sci-fi 🧞‍♂️ Spiritual 🧪 Science & Eng ~= Maker ❌ None ~= nihilism/realism 💬 ...

A bit late due to the flu 😅 but still very worth sharing: China open source highlights for January 2026🔥 huggingface.co/collections/...

📢 𝑪𝒂𝒍𝒍 𝒇𝒐𝒓 𝑷𝒂𝒑𝒆𝒓𝒔: @cvprconference.bsky.social 2026 Workshop on AI for Creative Visual Content Generation, Editing, and Understanding (CVEU)! We are thrilled to invite submissions to the 9th installment of the CVEU workshop at CVPR 2026 in Denver!

Bubble Crasher (S05E19)

I made a map of 3.4 million Bluesky users - see if you can find yourself! bluesky-map.theo.io I've seen some similar projects, but IMO this seems to better capture some of the fine-grained detail

I'm thinking of building a logged in version of the For You playground: linklonk.com/bluesky It could have some personal stats and settings. What would you like to see/do on that page?

@yaroslavvb.bsky.social 1/3 🧵 "People ask me what I think about AGI. I think modern AGI discourse defies the voice of reason. Thankful to 5-hour "walks" with Ian Goodfellow and Shubho Sengupta that motivated me to put this down, had a lot of fun writing this essay... speakerdeck.com/yaroslavvb/g...

@JitendraMalikCV: "It's great to see the excitement these days about world models and their applications in robotics. World models predate deep learning, and if you're curious, here's my history talk starting from Craik (1943) & Kalman (1960)" www.youtube.com/watch?v=9B4k...

@mariaa.bsky.social and I have been building a custom client that we'd like to make public in the near-term BUT we want to make sure we've gotten the security right. Anyone with some expertise open to helping us out a bit?

In the attention economy, learning whether -> when -> how to engage is a crucial human skill

The #ECCV2026 Call for Workshop Proposals is now posted! Workshop call: eccv.ecva.net/Conferences/...

Check out what’s new at #ECCV2026 Malmo 🇸🇪 Full details: eccv.ecva.net/Conferences/...

Considering to submit a workshop proposal for #ECCV2026, you? Guess & help me prioritizing 🅰️ topic of a fellowship, if awarded. 🅱️ topic that researchers complaint more often than they praise. 📽️ Easy guess. 2nd Hint: we got half-day in #CVPR2026 🇺🇸. Hence mostly for nurturing our 🇪🇺 community.

This! The dirtiest part of the peer review process in top-tier Computer Vision conf is perhaps the discussion phase. Is it still not transparent? The 2nd one, the meta-review. Authors often complain about Rs, but given the lack of transparency, Isn’t it fair to split that difference with ACs?

I'm reviewing four papers. Three of them have a rebuttal. I wrote down my thoughts regarding the rebuttal for each. Across the three papers, one single (out of six) other reviewer has responded to the rebuttal.

Did you learn differential privacy (in part or in whole) from my course? Either the videos, lecture notes, or some combination? Please send me a DM or an email, I'm trying to gather some info. (In case you missed it, here's the course: www.gautamkamath.com/CS860-fa2020..., ft full notes & videos)

📢 Fully funded PhD - 🌍 Dense Detection of Rare Events in Remote Sensing using Generative Models Leverage generative models, unsupervised segmentation and explainability techniques to map disasters w/ @javi-castillo.bsky.social and Flora Weissgerber Apply ⤵️ recrutement.cnes.fr/fr/annonce/4...

@chanwoopark.bsky.social 1/4 🧵 "Unpopular opinion: If any researchers are going to be replaced by AI, AI researchers will be the first. We’re already watching software engineering get automated at a rapid pace. What comes next seems to split into two directions: (1) replacing human labor in ...

@alphaxiv.org 1/2 🧵 "Following Google's Genie 3, Ant Group released an open source alternative: LingBot-World! This video trained world model can also be steered like a game (WASD + camera), with sub-1s latency at 16fps in its real-time setup, and demonstrates coherent rollouts up to ...

GLM just entered the OCR field🔥 huggingface.co/zai-org/GLM-... ✨ 0.9B ✨ MIT licensed ✨ Multimodal GLM-V architecture ✨ #1 on OmniDocBench v1.5 (94.62)

Google presents PaperBanana Automating Academic Illustration for AI Scientists huggingface.co/papers/2601....

As part of my personal site revamping, I'm considering writing up a blog post. Which topics you wanna hear about? I'm considering: 1. To AI Scientists or perish? My thoughts as enthusiastic content creator & Scientist (researcher & community builder) 2. Anything else? 💬

Help me to reach >= 42 likes ❤️ What's in there for you? A thoughtful/interesting/plain?/🎲 Rant from my mate 😂

@techemails.bsky.social 1/2 🧵 "Microsoft CTO: Why the OpenAI board really fired Sam Altman November 19, 2023

@keenancrane.bsky.social In X/Twitter 1/3 🧵 "Intelligence is powerful, but expensive. Just because you know how to do a task well doesn’t mean efficiency suddenly becomes irrelevant.

Why was old prose verbose? 1. Account for missing gestures and physical communication perhaps 2. Pleasing & not bothering the elite ‼️ For example, formal syntax & expressions in some languages, or even NOT using your town ~language~ dialect. Welcome to the elite?

The February issue of the PAMI-TC newsletter is out: tinyurl.com/4w4d6xjk Inside is a call for motions for #WACV2026, a few open calls for #CVPR2026, and a bunch of other misc. vision items to consider.

Yet another tool as JM & others 'request.' An author's TL;DR in the next tweet

More words on this: realizable.substack.com/p/hayeks-abs...

@srchvrs.bsky.social 1/4 🧵 ""I’m regularly asked if one should leave a Ph.D. to get an actual job, and my decision criteria is fairly simple. If you’re not looking to become a professor and have an offer to do modeling research at a frontier lab (Gemini, Anthropic, OpenAI is my list) then ...

Yi Tay (@YiTayML & @agihippo), 1/17 🧵 "I agree and disagree with many things in this blog post, but as someone that hired a full team recently and had thousands of applications everywhere (that even bled into my instagram DMs 😅), I thought I shed some perspective on this. ...

😂

Baidu just released a new VLM : PaddleOCR-VL-1.5🔥 huggingface.co/PaddlePaddle... ✨ 0.9B - Apache 2.0 ✨ 94.5% OmniDocBench v1.5 ✨ Multilingual OCR: strong on rare characters & ancient texts

Been a while since I did an LLM architecture post. Just stumbled upon the Arcee AI Trinity Large release + technical report released yesterday and couldn't resist :) Also added a new section to my LLM architecture comparison article with more details: magazine.sebastianraschka.com/i/168650848/20

the first two pages of suggested abstracts were very relevant to my interests but it's quite an exhausting exercise

The people want to know: how many submissions made it over the finish line? 🏁 Answer: 🥁 ... ... ... 24371! This is the number of submitted PDFs. There will be more checks (& hopefully withdrawal of half baked submissions 👀), stay tuned to see how many enter review

I'm re-reading various papers surrounding the debate between computationalist and dynamicist approaches to cognition (e.g., Fodor vs. van Gelder), and it's a hot mess. For one, of course Turing machines and finite automata are dynamical systems! But also, the dynamical systems formalism in ... (1/2)

A study presents Order-Token Search, a decoding method for Diffusion Language Models that explores generation order and token values, achieving gains in reasoning and coding tasks, surpassing existing baselines and creating a standard for multi-trajectory potential. https://arxiv.org/abs/2601.20339

A study indicates that AI boosts productivity for novice developers but impairs their skill formation, harming their understanding and debugging skills. This stresses the need for careful AI integration in training, particularly in critical fields. https://arxiv.org/abs/2601.20245

Matthew McConaughey & Simon Sinek Episode hit me 🎧 - Humility takes multiple forms & it's a superpower. - “Journaling” for me doesn’t have to mean writing, but talking it out. - And growth? It lives inside the discomfort you keep avoiding open.spotify.com/episode/0ebF...

For the HERAtlas project at the University of Amsterdam (Netherlands) we are looking for a Postdoc to combine data science, psychology, and digital history to uncover the "invisible" women of art history! More info at: werkenbij.uva.nl/en/vacancies...

CS ArXiv recently banned “review and position” papers, but what are those? Do they include more generated content? Who is most affected by this change? @yanai.bsky.social and I dug into the data to find out! Nearly 50% of Computers & Society papers might be censored, vs 3% of Computer Vision ‼️

Qwen3-ASR is out🚀 huggingface.co/collections/... ✨ 0.6B & 1.7B - Apache2.0 ✨ 30 languages + 22 Chinese dialects, plus English accents across regions ✨ Single model for language ID + ASR (no extra pipeline stitching) ✨ Qwen3-ForcedAligner-0.6B, a strong forced aligner outperforming E2E baselines