From LEGEND's Neutrons to LIGO's Binary Black Holes: A Rare-Event Journey Across Physics

Some of the most important questions in modern physics, from whether neutrinoless double-beta decay exists to how black holes form and evolve, live in regimes where the signals that matter are almost never observed. Yet it is precisely these rare events that define the edges of our knowledge, from the behavior of neutrinos deep underground to the evolution history of massive black holes across cosmic time.

This talk introduces a research direction that seeks to bridge those extremes through a unified framework for rare-event surrogate modeling. The ambition is straightforward: replace brute-force simulation with probabilistic surrogates that are built to learn from scarcity rather than be constrained by it. By integrating tools such as conditional neural processes, multi-fidelity and Gaussian processes, this framework constructs statistically consistent approximations of complex likelihoods even when only a handful of informative samples exist.

The impact becomes clear when applied across two seemingly unrelated domains. In neutrino physics, rare-event surrogates open up detector design spaces that would otherwise be locked behind prohibitively expensive Monte Carlo simulations. In gravitational-wave astrophysics, the same principles stabilize population-inference pipelines where truly merging binaries are vanishingly rare within population-synthesis models. Despite the enormous difference in scale and context, the core problem and the methodological solution remains the same.