The author of FlashAttention, Tri Dao, just dropped a new paper called SonicMoE With 1.86x higher MoE kernel throughput and 45% lower activation memory per layer on H100s, by introduceing tile-aware routing that cuts padding waste for sparse MoEs Trending on alphaXiv
MACHINE LEARNING
-
Stanford HAI Researchers Predict AI Trends for 2026
By
–
What’s next in AI in 2026? Stanford HAI's leading researchers just dropped their predictions: @landay https://t.co/0eRDUypDCZ pic.twitter.com/laz63RarMR
— Stanford HAI (@StanfordHAI) 19 décembre 2025What’s next in AI in 2026? Stanford HAI's leading researchers just dropped their predictions: @landay https://
hai.stanford.edu/news/stanford-
ai-experts-predict-what-will-happen-in-2026
… -
Hands-On Product Testing Remains Critical in AI Age
By
–
In The AI Age, Hands-On Product Testing Still Counts
#AI #AIio #AIInnovation #ML #DataScience #Futureofwork @fabiomoioli @pascal_bornet @alliekmiller @mattshumer_ @OfficialLoganK @jeremyphoward @GaryMarcus http://
ow.ly/9leB30sRS5H -
AI Reveals Amyloid-Beta Feedback Loop in Alzheimer’s Pathogenesis
By
–
The amyloid-β feedback loop and pathogenesis of Alzheimer's disease
when aggregates form, the shift from protecting/building synapses to destroying them @jclinicalinvest https://
jci.org/articles/view/
200393
… -
LLMs as Brainstorming Partners for Scientific Discovery and Innovation
By
–
Evaluating the potential of LLMs to help with scientific discovery. In short: new ideas are direly needed to move AI towards invention. LLMs can be useful as brainstorming partners though.
-
Emotion-Based AI Transforms Computer Vision and Robotics Interactions
By
–
Emotion-based AI is a key differentiator in the new wave of computer vision and robotics, making interactions more natural and personalized while still protecting privacy and ethics. Source @Gartner_inc Link https://
gtnr.it/3KasVBM via @antgrasso #AI #ArtificialIntelligence -
Deep Reinforcement Learning for Multi-Agent Robot Coordination
By
–
Deep Reinforcement Learning for Multi-Agent Coordination Multi-agent reinforcement learning doesn't scale. The default approach to coordinating robot teams remains explicit communication or centralized training. This is inefficient. On the other hand, social insects solve this
-
Stanford’s Coordination Layer Theory Challenges LLM AGI Bottleneck
By
–
Are LLMs a dead end for AGI? Stanford's Edward Y. Chang argues the bottleneck isn't the model, but a missing "coordination layer." The new theory (UCCT) sees reasoning as a phase transition, moving from simple pattern-matching to goal-directed thinking. He built MACI, a
-
MathWorks Showcases AI Embedded Systems for Smart Vehicles
By
–
Join MathWorks at #CES2026! See how we’re driving smarter, safer, more connected vehicles at Booth 4477: AI for Embedded Model-Based SDV Design Simulink Virtualization Digital Twins w/ Databricks Meet our experts: https://
spr.ly/6014CuYuK -
Fine-tuning LLMs: The Continued Focus on Reinforcement Learning
By
–
Their hit was fine-tuning LLMs (“RL”), and this is still that. (Even CoT came from Google.)