A team led by @songhan_mit just released XAttention, a plug-and-play framework that accelerates attention computation by 13.5×!
RESEARCH
-

Cite Original Paper Explaining AI Tokens Definition
By
–
Cite this OG paper when explaining what a token is in AI.
-
Mel Mitchell’s Essay on Large Reasoning Models
By
–
Great plain English essay, open-access @ScienceMagazine by @MelMitchell1 on large reasoning models
-
Lineage: Family Tree Data Expression Engine for AI
By
–
GitHub – bengarvey/lineage: Family Tree Data Expression Engine https://
buff.ly/XetpXw8
#AI #MachineLearning #DeepLearning #LLMs #DataScience -

NVIDIA EPRI Launch Open Power AI Consortium
By
–
Announced at #GTC25, EPRI, NVIDIA, and collaborators launch the Open Power AI Consortium (#OPAI). The consortium is building open datasets, #AI models, and software for use cases such as interconnection and rate case studies. Learn more: https://
nvda.ws/4kJA8pS -
Newton: Open-Source Physics Engine for Robot Learning Unveiled
By
–
Big news from #GTC25 ! @NVIDIA, @GoogleDeepMind, and Disney Research just introduced Newton—an open-source physics engine designed to supercharge robot learning.#Humanoidrobots #PhysicalAI@NVIDIARobotics @NVIDIAGTC
— Harold Sinnott #MWC26 (@HaroldSinnott) 20 mars 2025
pic.twitter.com/34KTNhmlJHBig news from #GTC25 ! @NVIDIA
, @GoogleDeepMind
, and Disney Research just introduced Newton—an open-source physics engine designed to supercharge robot learning. #Humanoidrobots #PhysicalAI @NVIDIARobotics @NVIDIAGTC -

Improved Load Balancing Bounds for Computational Cluster Scheduling
By
–
Today we examine classical scheduling problems, common in computational cluster management, and present improved theoretical upper and lower bounds for load balancing with random orders of arrival. Learn more at https://
goo.gle/4kKNxxW -
GPU cluster scheduling practices at major AI labs
By
–
Have any of the big AI labs written publicly about how they schedule usage of their GPU clusters between research, frontier training, and production?
-

AlexNet Source Code: Historical Logbook Comments Reveal Neural Network Development
By
–
And maybe the real history here are the logbook comments at the end of each experiment config file in the AlexNet code A seminal neural net in the making… https://
github.com/computerhistor
y/AlexNet-Source-Code
… -
H2O.ai Reclaims Top GAIA Benchmark Score Position
By
–
That’s a big deal.https://t.co/AHCQLzjMBA now holds the highest GAIA Benchmark score – ever. 🏆
— H2O.ai (@h2oai) 20 mars 2025
⚡@deepseek_ai & @ManusAI_HQ pushed past @OpenAI – but https://t.co/AHCQLzjMBA just reclaimed the top spot.
🎥 @srisatish breaks it down @SiliconANGLE & @theCUBE at @NVIDIAGTC. pic.twitter.com/yi2yZGde5CThat’s a big deal. http://
H2O.ai now holds the highest GAIA Benchmark score – ever. @deepseek_ai & @manusai pushed past @OpenAI – but http://
H2O.ai just reclaimed the top spot. @srisatish breaks it down @SiliconANGLE & @theCUBE at @NVIDIAGTC
.
