RooAgent: An LLM Agent for Root-Based High Energy Physics Analysis

TL;DR

RooAgent is a natural-language interface that enables high energy physics data analysis using LLMs, supporting multiple modes and tasks like histogram inspection, event selection, and visualization.

Contribution

It introduces a versatile LLM-based tool that simplifies complex physics data analysis workflows through natural language commands and multi-mode operation.

Findings

Supports various analysis tasks including histogram inspection and fitting.

Demonstrated on Monte Carlo simulations and ATLAS open data.

Available as open-source on GitHub.

Abstract

We present RooAgent as a natural-language interface for Root-based high energy physics data analysis. The package provides physics analysis functions as tools that an LLM agent invokes in response to plain-language prompts. Two operating modes are supported: a LangGraph-based agent compatible with OpenAI's GPT-4.1 via GitHub Copilot and with DeepSeek-V3 via Ollama, and a Model Context Protocol server for use with the Anthropic Claude CLI (Sonnet~4.6). In both modes the analysis logic is implemented in PyRoot and the LLM selects tools and supplies the required arguments. The package supports histogram inspection, event selection, visualisation of kinematic distributions, fitting, and significance estimation, among other tasks. We illustrate RooAgent with tests based on Monte Carlo simulations of $pp\to ZH$ ($Z\to\ell^+\ell^-$, $H\to b\bar{b}$), a multi-task signal-background workflow, a toy statistical analysis, and an application to ATLAS open data for $H\to ZZ^*\to 4\ell$. The package is available on PyPI and the source code is hosted at https://github.com/amanmdesai/RooAgent.