New Categorical Framework Unifies Causal Model Abstractions

A new categorical framework treats causal abstractions as transformations between Markov functors, linking low‑ and high‑level models. Submitted 6 Oct 2025 (stat.ML; cs.LG). getnews.me/new-categorical-framewor... #causalabstraction #categorytheory

Distributionally Robust Causal Abstractions Improve AI Model Reliability

Researchers introduced robust causal abstractions with a Wasserstein ambiguity set in a min‑max formulation, maintaining performance when test environments shift. Read more: getnews.me/distributionally-robust-... #causalabstraction #robustai

New Projected Abstraction Framework Tackles Lossy Causal Representations

A new projected abstraction method handles lossy causal representations and was validated on image tasks with thousands of variables. It was presented at ICML 2025. getnews.me/new-projected-abstractio... #causalabstraction #projectedabstraction