Wow, a new blogpost from the GOAT Sebastian Raschka! Recent Developments in LLM Architectures: KV Sharing, mHC, and Compressed Attention
MACHINE LEARNING
-
Representing AI agents as state for improved persistence
By
–
maybe you can’t just tack statefulness onto an agent, you have to figure out how to represent the agent as a state
-
Google researchers: can’t imagine publishing transformer paper now, return to academia?
By
–
Google researchers say they can’t imagine being allowed to publish the transformer paper now. Time to return to academia?
-
Yoshua Bengio’s Large-Scale Graduate Student Descent Breakthrough
By
–
The key AI breakthrough of the last 20 years, due to Yoshua Bengio, was large-scale graduate student descent.
-
RSI and continual learning as barriers to AI takeoff
By
–
So the two most obvious barriers to some sort of true AI takeoff are robust RSI (AI acting as an independent AI researcher, rather than “merely” a multiplier of human effort) and continual learning. Either would represent a major change in trajectory for AI development.
-
Technical challenges in stateful AI agents and memory management
By
–
yeah they do cool stuff around memory. starting to think stateful goes beyond that to tracking changes in the agents capability, etc which I’m not sure if many memory tools do
-
LLM traces masquent un planificateur myope
By
–
“Extracting Search Trees from LLM Reasoning Traces Reveals Myopic Planning” Reasoning models can write traces that look like real tree search, but this paper shows their decisions are mostly driven by shallow one-step evaluation. They extract search trees from LLM CoT in
-
Complexity of state tracking in AI agents
By
–
one thing that makes agents more complex is that they generally add capability over time which should also be tracked as part of the state
-
Critical Perspective on GPT Development and AI Capabilities
By
–
What I am about to describe ain’t AGI; it’s a sign of a trillion dollar trainwreck. If I had told you in 2022 that the 2026 version of GPT (which by the way would only be GPT 5.5 and not GPT-6 or 7 like many people fantasized about) would still have strange quirks like inserting
-
China’s fully automated AI port moves cargo without humans
By
–
Fully #Automated: China’s #AI-Powered Port Moves Cargo Without Humans
— Ronald van Loon (@Ronald_vanLoon) 17 mai 2026
by @_fluxfeeds#ArtificialIntelligence #MachineLearning #ML #MI pic.twitter.com/pNEu6UX9EnFully #Automated: China’s #AI-Powered Port Moves Cargo Without Humans
by @_fluxfeeds #ArtificialIntelligence #MachineLearning #ML #MI