OpenAI Codex and Claude Co-Author a Real Particle Physics Paper
Two AI coding agents performed a full particle physics measurement on archived ALEPH data, from event selection to unfolding to writing the paper.
- AI agents Codex and Claude performed a full ALEPH thrust measurement under physicist direction, from raw data to published paper.
- Analysis used Iterative Bayesian Unfolding with a Monte Carlo response matrix built by the agents from scratch.
- Corrected thrust spectrum at 91.2 GeV agrees with historical ALEPH results within uncertainties.
- Seven experimental systematic variations covering tracking cuts and neutral energy scale were handled by the agents.
- Agents also drafted the LaTeX note, generated figures, and iterated on referee feedback.
- Full pipeline open-sourced at GitHub repo; published in HiHEP 2 (2026) 3, 10.
Physicists gave OpenAI Codex and Anthropic Claude responsibility for much of an experimental particle-physics analysis, from processing archived detector data to generating figures and drafting the manuscript. The humans defined the measurement, approved methods, and reviewed the output. The resulting arXiv paper, Agentic AI - Physicist Collaboration in Experimental Particle Physics, reports the thrust distribution in electron-positron collisions at a center-of-mass energy of 91.2 GeV using data from CERN’s ALEPH experiment.
A benchmark with an answer key
Thrust compresses the shape of a collision event into a number between 0.5 and 1. For each event, the calculation finds the axis that maximizes the summed particle momenta projected along it. A value near 1 describes two narrow, back-to-back jets; a value near 0.5 describes a more spherical spray of particles. Physicists use thrust to study quantum chromodynamics, including gluon radiation and the strong coupling constant.
ALEPH recorded millions of Z-boson decays at CERN’s Large Electron-Positron Collider during the 1990s. At 91.2 GeV, the collider operated near the Z-boson resonance, where hadronic decays provided a large, well-studied event sample. Archived ALEPH data suit an agentic workflow test because historical measurements provide a reference against which researchers can detect biased selections, incorrect corrections, or faulty uncertainty estimates.
From event records to a corrected spectrum
The agents executed the main stages of a collider analysis while physicists reviewed intermediate results and methodological choices:
- Read ALEPH event records and classify reconstructed objects as charged particles, neutral hadrons, photons, electrons, or muons.
This story is for Pro members
You've reached the end of the free preview. Upgrade to AlphaSignal Pro to read the full article - and everything else behind the paywall.
Comments
