Reproduce “Prompting Particle Physics: Tokenized Multi-modal Foundation Models for Combinatorially Many Tasks” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Nilotpal Kakati, Daniel Murnane, Baran Hashemi, Samuel Klein, Jeffrey Krupa, Eilam Gross, Lukas Heinrich, Michael Kagan
The authors develop a shared token interface that maps among detector- and truth-level objects in simulated collider events. Tasks are specified by choosing input and output modalities, with the autoregressive model predicting outputs iteratively. On particle flow, it produces a narrower jet-response distribution than HGPflow, but its median is more biased low. The study tests an unseen four-input combination, but only in single-jet simulation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.