Iryna Biziukova

Major: Physics

Mentor: Inigo Valenzuela Lombera, Nathaniel Craig

The Fermiacc Validator Module: Agents For BSM Theory Extraction

The Standard Model (SM) of particle physics stands as one of the most predictive frameworks in modern science, successfully describing fundamental interactions across high-energy regimes explored by the Large Hadron Collider (LHC). Despite its empirical successes, the SM remains fundamentally incomplete, failing to account for critical phenomena such as the hierarchy problem, non-zero neutrino masses, and a quantum description of gravity. While theoretical physics has produced a vast landscape of Beyond the Standard Model (SM) candidates to resolve these open questions, evaluating each proposed framework against experimental collision data remains a labor-intensive computational bottleneck. Large Language Model (LLM) agents present a scalable paradigm to accelerate this iterative discovery process. The FERMIACC framework was recently developed to automate the generation of theoretical BSM hypotheses and evaluate their correspondence with experimental collision results. In this work, we present an extension to the FERMIACC pipeline: the Validator module, designed to address scientific reproducibility by extracting theoretical BSM specifications directly from published research literature. The Validator utilizes an LLM-driven multi-agent extraction engine to parse domain-specific papers, reconstruct theoretical models and simulation parameters into structured benchmark schemas, and route them into external Monte Carlo simulation and analysis tools.