arXiv subjects

Johann Ioannou-Nikolaides

Publications and source records attributed to Johann Ioannou-Nikolaides.

4 recordsLinked to original sources

arXiv2026

Interplay of Lorentz Invariance Violation and Earth's Matter Potential in High-Energy Neutrinos

Searches for Lorentz invariance violation (LIV) in the neutrino sector have traditionally focused on non-standard neutrino oscillations induced by LIV in vacuum. In this work, we study anisotropic LIV in matter. First, we review vacuum LIV phenomenology, explaining the energy and direction dependence of sidereal modulations for anisotropic coefficients in the Standard Model Extension. We then demonstrate that for high-energy neutrinos, the interplay between anisotropic LIV operators and the Earth's matter potential produces distinct, observable signatures absent in the vacuum case. We identify a crossover regime where the energy-dependent LIV Hamiltonian becomes comparable to the matter potential, leading to strong interference effects. By analyzing the propagation of neutrinos through a realistic Earth model, we establish three key phenomenological consequences: (1) direction-dependent resonant enhancements of oscillation probabilities, (2) a macroscopic breakdown of neutrino-antineutrino symmetry for CPT-even operators, and (3) a significant increase of the $ν_τ$ flux due to LIV-driven injection of high-energy neutrinos into the $τ$ regeneration cycle. These results highlight that accounting for the interplay between LIV and matter is essential for future LIV searches at large-scale neutrino telescopes.

hep-ph↗

arXiv2026

Finding and using interpretable latents in a neutrino foundation model with sparse autoencoders

We present a first application of sparse-autoencoder-based mechanistic interpretability to particle physics. Studying a neutrino foundation model pretrained on IceCube data and fine-tuned for direction reconstruction, we identify a validated atlas of physical concepts in the model representation, using a strict validation protocol consisting of held-out tests, matched nuisance controls, and replication across independent dictionary trainings. Causal interventions show that the direction head barely draws on this atlas. Motivated by this underused information, we train an uncertainty head on the same event-level representation to predict the model's angular reconstruction error. Unlike the direction head, it depends causally on quality and brightness features from the atlas. At $20\%$ selection efficiency, this interpretable estimator improves the median angular resolution from $20.2^\circ$ to $3.2^\circ$. These results suggest that mechanistic interpretability can reveal learned latent physics encoded within a model's internal representation and help design downstream tasks that exploit it.

astro-ph.HE↗

arXiv2026

Multiclass Classification without Labels via Posterior Simplex Geometry

In many classification problems, reliable instance-level labels are unavailable. However, it is often possible to construct weakly enriched unlabeled samples: datasets selected by different cuts, sources, populations, or experimental conditions that change latent class proportions without revealing them. Classification without Labels (CWoLa) shows that, in the binary case ($K=2$), a classifier trained to distinguish two impure mixtures with different class proportions can recover an optimal class discriminator without knowing the mixture proportions. We extend this principle to multiclass learning from several unlabeled mixtures ($K>2$), where the learner observes only mixture identity and neither latent class labels nor class-prior matrices. We prove that, for a multiclass mixture model, the Bayes-optimal mixture classifier $g^\star$ maps data points into a $(K-1)$-simplex embedded in mixture-posterior space. The $K$ vertices of this simplex are induced by the latent classes through the unknown mixing matrix. Leveraging this geometry, we propose prior-free procedures that train a standard classifier to distinguish mixture identities and then extract latent class structure using either post-hoc simplex fitting or a bottleneck architecture. Experiments on MNIST, CIFAR-10, and Galaxy10 DECaLS show that mixture identity alone can recover latent classes and their fractions in the mixture. By narrowing the gap between weakly supervised and fully supervised performance, we provide a mathematically grounded, scalable tool for multiclass discovery in label-scarce domains.

cs.LG↗

arXiv2026

Neutrino Oscillations as a Probe of Macrorealism

The correlations between successive measurements of a quantum system can violate a family of Leggett-Garg Inequalities (LGIs) that are analogous to the violation of Bell's inequalities of measurements performed on spatially separated quantum systems. These LGIs follow from a macrorealistic point of view, imposing that a classical system is at all times in a definite state and that a measurement can, at least in principle, leave this state undisturbed. Violations of LGIs can be probed by neutrino flavour oscillations if the correlators of consecutive flavour measurements are approximately stationary. We discuss here several improvements of the methodology used in previous analyses based on accelerator and reactor neutrino data. We argue that the strong claims of LGI violations made in previous studies are based on an unsuitable modelling of macrorealistic systems in statistical hypothesis tests. We illustrate our improved methodology via the example of the MINOS muon-neutrino survival data, where we find revised statistical evidence for violations of LGIs at the $(2-3)σ$ level, depending on macrorealistic background models.

hep-ph↗