Our ICML spotlight paper discovers universal redundancies in time series foundation models: the middle layers of many models can be removed without sacrificing performance 1/
Our group discovered that reasoning models produce fractals when asked to solve hard problems. We can use nonlinear dynamics to probe the thinking processes of recurrent depth models on Sudoku, mathematics, and even ARC-AGI (1/N)
arxiv.org/abs/2609.04963
Today, we announced the 2026 Fellows—we hope you will read their incredible stories and learn about their work! Selected from 3,000+ applications, it was the most competitive year in our history.
pdsoros.org/meet-the-class…
How do time series foundation models forecast unseen dynamical systems? In new experiments, we find that small transformers learn to approximate transfer operators in-context. (1/N)
arxiv.org/abs/2602.18679
