Uncovering Physical Drivers of Dark Matter Halo Structures with Auxiliary-Variable-Guided Generative Models

Collection of new ML research (arXiv 2026-03-02) introducing algorithmic and theoretical advances across generative modeling, inference, causal discovery, fairness, spatio-temporal/event modeling, Bayesian policy learning, calibration, active ranking, interpretability, symbolic regression, distributed optimization, and partition-function estimation. Highlights include a disentangled latent conditional flow-matching model for tSZ maps linking latent coordinates to halo mass/concentration; an information-theoretic characterization of partition-function estimation via integrated coverage and f‑d