15–18 Dec 2026
Duke University
America/New_York timezone
Artificial Intelligence for the Electron Ion Collider (2026)

TPTNP: Transferable Particle Tracking in Nuclear Physics

Not scheduled
15m
LSRC B101 Love Auditorium (Duke University)

LSRC B101 Love Auditorium

Duke University

Durham, NC 27710
AI/ML for Data Analysis and Theory AI/ML for Data Analysis and Theory

Speaker

Chenliang Jin (Stony Brook University)

Description

The Transferable Particle Tracking in Nuclear Physics (TPTNP) project,
supported by the DOE Genesis Mission, aims to develop a foundation model
for tracking across different detector technologies and geometries.
Building on FM4NPP, we investigate how learned representations of raw
detector measurements can support track reconstruction across experimental
settings. We will present ongoing studies using data from sPHENIX, MUSE,
and DarkLight, exploring their complementary challenges within a common
framework.

Our proposed strategy combines pretraining a shared Mamba-2 backbone
across diverse detector configurations with subsequent fine-tuning for
each detector. Explicit descriptions of geometry and measurement response
provide context for interpreting different detector signals. We will
discuss how this approach can support adaptation to unfamiliar experimental
settings while retaining transferable tracking representations. The broader
goal is to reduce experiment-specific development and establish a reusable
framework for particle reconstruction, with potential applications to the
detector systems of the Electron-Ion Collider.

Author

Chenliang Jin (Stony Brook University)

Presentation materials

There are no materials yet.