AgentRivet: an automated system for producing Rivet routines from journal publications

AgentRivet is an automated system designed to generate Rivet routines from particle physics journal publications, addressing the issue of incomplete analysis coverage in Rivet, where only 39% of measurements have available routines. Utilizing commercial Large Language Models from OpenAI, Anthropic, and Google, AgentRivet implements a multi-step workflow that extracts analysis information and produces Rivet routines, achieving competent outputs with minimal syntax errors. This development is significant for practitioners as it enhances the accessibility of model-independent measurements and aids in the tuning of Monte Carlo event generators, although it also highlights challenges in accurately implementing complex observables due to ambiguities in published definitions.