Vinicius Mikuni · Physicist

Learning from data.
Exploring the Universe.

I'm an Associate Professor at the Kobayashi–Maskawa Institute, Nagoya University. My group develops machine learning methods to understand the physical world, from particle collisions to the cosmos.

Vinicius Mikuni speaking at a lectern
Associate Professor · KMINagoya, Japan

Research directions

AI for the physical sciences

01 / Representations

Foundation models

Learning transferable representations of scientific data, and investigating how knowledge from one experiment can support another.

02 / Simulation

Generative models

Building fast, faithful simulations of particles, detectors, and physical systems using diffusion models and scientific structure.

03 / Inference

Measurements & discovery

Developing methods for unfolding, uncertainty quantification, and anomaly detection to extract more information from experimental data.

Selected work

Research highlights

All publications
Top-quark tagging receiver operating characteristic curves comparing compact models and distillation; Figure 1.Foundation models · 2026

Foundation models for real-time physics

We distill the knowledge of OmniLearned into compact, attention-free networks for top-quark jet tagging. Message passing and pretrained teacher predictions improve background rejection, especially at the low signal efficiencies relevant for real-time triggers.

Read the paper
OmniLearned forward-pass architecture with classification, flow-matching, and masked-particle training objectives; Figure 2.Foundation models · 2026

Choosing what foundation models learn

We compare classification, flow matching, and masked-particle pretraining within OmniLearned. The best objective depends on the downstream task and available labels: masking complements classification when labels are scarce, while generative transfer requires generative pretraining.

Read the paper
Transfer learning from collider jet data to neutrino interactions in MINERvA; Figure 1.Cross-domain learning · 2026

From collider jets to neutrino interactions

Knowledge learned from collider data transfers to few-GeV neutrino interactions in MINERvA. Pretrained OmniLearned and ParticleViT models improve energy regression and event classification at matched compute budgets, despite large differences in detectors, energy scales, and physics processes.

Read the paper
Azimuthal particle-pair correlations and trigger transverse momentum for OmniLearn and Angantyr heavy-ion events; Figure 6.Fast simulation · 2026

Generating entire heavy-ion events

A point-cloud diffusion model learns lower-multiplicity oxygen collisions before adapting to lead collisions. We assess generated high-multiplicity events through particle observables, collective flow, reconstructed jets, and classifier-based fidelity tests.

Read the paper
Simulated proton shower images from Monte Carlo, WGAN, and score-based diffusion models for H.E.S.S.; Figure 10.Fast simulation · 2026

Diffusion models for gamma-ray astronomy

We generate gamma-ray and proton shower images for H.E.S.S. with score-based diffusion models. Tests of image structure, correlations, and gamma–hadron separation show improved fidelity over GANs, including the more challenging proton showers.

Read the paper
Mass-distribution scan for the OmniLearned anomaly-search analysisAnomaly detection · 2026

Searching for anomalies with foundation models

Using OmniLearned to select events in CMS data, we test a complete background estimate. The estimate describes validation regions but does not accurately model the signal region, motivating further scrutiny of both the events and the method.

Read the paper
FLEX super-resolution results on turbulent flow fieldsGenerative models · 2025

FLEX: learning physical dynamics

A diffusion architecture combining convolutional layers and a latent transformer performs super-resolution and forecasting for turbulent flows, with uncertainty estimates from sampling.

Read the paper

Background

From experiments
to algorithms

My research connects high-energy physics, astrophysics, nuclear physics, and fluid dynamics through shared challenges in learning from complex data.

Before joining Nagoya University, I was a postdoctoral research fellow at Lawrence Berkeley National Laboratory's NERSC, working on anomaly detection, unfolding, and surrogate models for detector simulation.

My PhD research with the CMS experiment used machine learning to distinguish rare signals in particle collisions, reconstruct objects, and mitigate pileup.

View my CV

Let's work together.

Students, postdocs, and researchers interested in AI for the physical sciences are welcome to get in touch. Send me an introduction and tell me about your interests.

Say hello