Particle physics, astroparticle physics, and astrophysics increasingly rely on sophisticated statistical methods for searches and measurements. These methods have been transformed by advances in artificial intelligence, opening new opportunities for discovery while raising fundamental questions about uncertainty quantification, model misspecification, and the propagation of uncertainties to final results.
Experiments searching for rare events, such as dark matter or neutrino interactions, face the common challenge of extracting physics knowledge from small-sample data, where standard asymptotic approximations often break down. Here too, recent advances in AI-enabled statistics are opening promising paths toward extracting more scientific information without relying on large-sample assumptions. We therefore invite researchers working on these challenges, particularly at the interface of physics, statistics, and AI, to participate in this workshop.
PHYSTAT is a long-running workshop series that started at CERN, bringing together statisticians, AI researchers, and physicists and astrophysicists to explore shared interests and foster collaboration across these communities. This workshop will focus on the challenges of high-dimensional physics with small-sample data, and the new opportunities AI offers for problems in dark matter physics, neutrino physics and astrophysics, and beyond.
Registration details:
Registration will open soon. Early bird registration will be offered at a reduced rate until March 3, 2027, followed by regular registration until April 15, 2027. Funds may be available to cover registration fees for some participants, to facilitate broader participation, especially among early career scientists. Further details will be announced on this page.
Program overview:
The first day and a half of the workshop will provide introductory lectures bridging the communities: an introduction to statistics for physicists, and an introduction to neutrino physics and dark matter physics for statisticians. These sessions will include hands-on tutorials with AI models using Nexus, the NSF-funded supercomputer at Georgia Tech. The rest of the workshop will feature invited and contributed talks.
Call for Abstracts:
Contributions at the interface of physics, statistics, and AI are welcome. Talks may review the current state of the art in experiments or discuss emerging techniques. They may present novel solutions or highlight open challenges in existing methods. Since the goal of this workshop is cross-fertilisation, we especially welcome contributions that spark discussion of current challenges. Topics of interest include, but are not limited to:
- AI and statistical methods for rare interactions, such as in dark matter and neutrino physics
- High-dimensional inference techniques (e.g., parameter inference with NSBI and unfolding)
- Uncertainty quantification and robustness measures, both for AI and with AI
- Studies of asymptotics and methods that avoid them
- Current methods and emerging techniques for rare events, small samples, and sparse data
- New developments in AI / statistics that could be relevant to particle physics / astronomy
The Call for Abstracts will open soon. The abstract submission deadline is February 1, 2027.
Contributions from early career researchers are encouraged, and limited funds may be available to support their participation.
Organizing Committee:
- Sean Benevedes (Georgia Tech)
- Lydia Brenner (Nikhef)
- Aishik Ghosh (Georgia Tech), Chair
- Scott Kravitz (UT Austin)
- Ann Lee (CMU)
- Roberto Mandujano (Georgia Tech)
- Callum Wilkinson (LBNL)
- Elizabeth Worcester (BNL)
Scientific Committee:
- Olaf Behnke (DESY)
- Mikael Kuusela (CMU)
- Louis Lyons (Imperial)
- Jesse Thaler (MIT)
- Yao Xie (Georgia Tech)
Our Partners:
We gratefully acknowledge the Heising-Simons Foundation, our partner and the primary sponsor of this event.