Pinned post
Johannes Oswald
Verified account This account is verified. Learn more@oswaldjoh
Research Scientist, Paradigms of Intelligence Team, Google Zurich
732 Following 1.8K Followers 332 Posts
Latest posts 5
Great work led by @KaplFer analyzing how looped and grown models are biased towards better in-context learning vs reducing loss by in-weight learning.
Our hypothesis: looping and growing only need to learn complicated circuits once - which are then naturally replicated throughout depth!
Our hypothesis: looping and growing only need to learn complicated circuits once - which are then naturally replicated throughout depth!
Looped LLMs are everywhere right now! But why are they interesting? Do they model language differently than regular LLMs?
Grown and looped transformers are better sequence models!
With @kaitlinmaile & @oswaldjoh (@Google, Paradigms of Intelligence Team) and @stefanAbauer. 🧵View quoted post on X
In-context learning powered exploration and data collection is super intriguing and important … Lots more to come in this space!
Excited to introduce our new paper exploring the use of in-context learning to support intrinsic curiosity! This was with an amazing team at Google including @sangnie, @g_lajoie_, @derifatives, @neurosacramento, Rajaj Nasser, @oswaldjoh, and @blaiseaguera. arxiv.org/abs/2606.19476View quoted post on X
Isn’t running MesaNets?
Introducing MSI EdgeMesa N AI+.
Powered by @nvidia RTX Spark , this ultra-compact powerhouse is built for AI devs, data scientists & creators.
Local LLMs. Real-time inference. Zero cloud dependency—high-performance AI at the edge!
📍 Visit us tomorrow at #COMPUTEX2026.View quoted post on X
Love this work!
[1/5] Intelligent behavior relies on maintaining an evolving, dynamic model of the environment. But even frontier models sometimes fail at tracking state in dialogs, e.g., actual transcript from 4/20/2026:View quoted post on X
