What if your federated learning system could intelligently guide new tasks to the best client, rather than just aggregating data? Researchers from Renmin University of China, Zhongnan University of Economics and Law, Meta, and East China Normal University present a novel
RESEARCH
-
AI adoption lag in scientific publishing systems
By
–
The fact that every scientific paper in 2026 is still uploaded only as fully formatted PDFs to academic archive sites that often limit downloads tells you everything you need to know about how quickly the scientific system is adjusting to the potential of AI to accelerate science
-
BaBot: DIY Ball-Balancing Robot Teaching Real Control Systems
By
–
BaBot: The DIY Ball-Balancing #Robot That Teaches Real Control Systems
— Ronald van Loon (@Ronald_vanLoon) 31 mars 2026
via @WevolverApp
#Robotics #Engineering #ArtificialIntelligence #Technology pic.twitter.com/tGizRJRcmJBaBot: The DIY Ball-Balancing #Robot That Teaches Real Control Systems
via @WevolverApp #Robotics #Engineering #ArtificialIntelligence #Technology -
Future AGI Development and Platform Feature Limitations
By
–
*here. in some decade we will be able to edit replies. Maybe AGI will come first.
-
Humans Still Excel at 3D Understanding and Autonomous Driving
By
–
depends on what information. humans are still better at comprehending the three-dimensional world, and much better at driving on unfamiliar roads, etc.
-
AI Diagnoses ER Patients More Accurately Than Doctors in New Study
By
–
Early days, nobody wants to over-invest in side-quests and non-standardized concepts that might end up being swallowed by the models themselves. The focus is now on building a strong foundation rather than anything else.
-
3D Gaussian Splat of Maritime Plaza San Francisco
By
–
captured this 3d gaussian splat of the maritime plaza in san francisco
— Bilawal Sidhu (@bilawalsidhu) 31 mars 2026
came out pretty damn crispy for a handheld 3d scan!
– 334 stills taken with a sony a7iii + 18mm prime lens
– imagery posed w/ epic's reality scan
– radiance field trained w/ litchfeld studio pic.twitter.com/MQN7RzPslJcaptured this 3d gaussian splat of the maritime plaza in san francisco came out pretty damn crispy for a handheld 3d scan! – 334 stills taken with a sony a7iii + 18mm prime lens – imagery posed w/ epic's reality scan – radiance field trained w/ litchfeld studio
→ View original post on X — @bilawalsidhu, 2026-03-31 01:55 UTC
-
Sarashina2.2-OCR Released: High-Precision Document Image Analysis Model
By
–
📢 Sarashina2.2-OCR Now Available Releasing an OCR model specialized in document image analysis 🚀 ✅ Markdown conversion while preserving layout
✅ Strong performance on vertical Japanese documents
✅ Detects diagrams and tables, outputs their positions Convert complex document images into high-precision, user-friendly formats for both humans and AI ✨ Learn more here
https://huggingface.co/sbintuitions/sarashina2.2-ocr [Translated from EN to English]→ View original post on X — @shiqi_yang_147, 2026-03-31 01:49 UTC
-
CIFAR’s Role in Sustaining Deep Learning Research Development
By
–
It's thanks to groups like CIFAR and its Neural Computation and Adaptive Perception group that DL work was maintained. It's not just a case of loving the work – a *lot* of toil went to getting to this point.
-
LeCun’s Commercial Success Despite Academic Neglect
By
–
LeCun's work was hugely commercially successful for decades even whilst being largely ignored academically.