ColliderML Release 1: Simulated HL-LHC Collision Events in the Open Data Detector

Provider
CERN, ColliderML team (Open Data Detector)
License
CC-BY-4.0 (source)
Size
48,002 files
Format
Parquet (in sampled files)
Stored at
Hugging Face (https://huggingface.co/datasets/CERN/ColliderML-Release-1)
Access
Free and public. No account with the data provider is needed.
DOI
10.57967/hf/7269
Category
Physics
Source
https://arxiv.org/abs/2512.15230

About this dataset

Simulated 14 TeV proton-proton collision events in the Open Data Detector (ODD), a generic detector resembling those planned for the High-Luminosity LHC, produced with DD4hep, Geant4 and ACTS (A Common Tracking Software) inside Key4hep. It is built for machine learning in high-energy physics: charged-particle tracking, track-to-particle matching, jet tagging, calorimeter energy reconstruction and signal versus background event classification. Listing data/ (measured 2026-11-13) shows 48 configuration folders that combine a physics process (ttbar, ggF Higgs, di-Higgs, Z to ee, Z to mumu, Z prime, hidden valley, Higgs portal, pileup only), a pileup condition (pu0 with no pileup, pu200 at HL-LHC pileup) and an object table (truth particles, tracker_hits, calo_hits, tracks). ttbar_pu200_tracker_hits alone holds 1,000 Parquet shards of about 0.7 GB; the repository uses about 29.1 TB. Each row is one event, with per-particle and per-hit quantities in Parquet list columns joined on event_id. Open it with the colliderml Python library (pip install colliderml), with Hugging Face datasets using streaming=True, or with polars, pyarrow or pandas. Released under CC-BY-4.0; cite the ColliderML paper (arXiv:2512.15230).

How to load this dataset in Python

from datasets import load_dataset
ds = load_dataset("CERN/ColliderML-Release-1")
print(ds)

Data structure

File types in sampled files: Parquet

Files

Example file: data/dihiggs_pu0_calo_hits/train-00000-of-01000.parquet

Analyze this dataset without downloading it

On Scigantic, ColliderML Release 1: Simulated HL-LHC Collision Events in the Open Data Detector opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.

This page is also available as Markdown and as JSON.

Related Physics datasets