Explicit or Implicit? Encoding Physics at the Precision Frontier

TL;DR

This paper compares explicit symmetry encoding and implicit data learning in machine learning models for particle physics, finding both approaches yield similar performance across challenging tasks, indicating efficiency gains are largely method-independent.

Contribution

It provides a comparative analysis of explicit versus implicit physics encoding methods, demonstrating their similar effectiveness in high-precision particle physics tasks.

Findings

Both models achieve comparable performance across tasks.

Efficiency gains from encoding physics are largely method-independent.

Performance is limited by dataset statistical precision.

Abstract

High-performance machine learning tools in particle physics rest on two complementary directions: encoding symmetries explicitly in the architecture, and implicitly learning the structure of the data through large-scale (pre-) training. We compare the performance of the representative L-GATr and OmniLearn models on three especially challenging tasks: reweighting-based unfolding, likelihood-ratio estimation, and weakly supervised anomaly detection. Across all benchmarks, both methods achieve comparable performance given the statistical precision of the finetuning datasets, suggesting that the significant efficiency gains from encoding known particle physics structures are largely method-independent.