hepdata

---

name: hepdata description: Use when the user wants the numerical tables, ROOT files, or YAML data attached to a published HEP measurement, identified by HEPData submission ID (ins1234567, 123/v2), INSPIRE record id, paper DOI, or paper title — typically because they want to re-fit, plot, or run systematics studies on the published measurement. Routes to https://www.hepdata.net/ via WebFetch and the HEPData REST API. Returns table URLs (YAML, ROOT, CSV) suitable for downstream pyhf, pandas, or uproot. Does NOT cover the paper's narrative or the headline measured value (use read-publication), PDG canonical constants (use pdg-lookup), or Open Data primary datasets (use cern-opendata or atlas-opendata). Disambiguator phrase: HEPData published-tables retrieval. data_scope: both experiment: all

hepdata — published HEP tables for re-fitting

HEPData hosts the numerical tables, covariance matrices, and ROOT inputs that LHC papers publish alongside their results. Use this skill when the user wants to consume those tables (re-fit, plot, systematics study) — not when they want the headline measurement value (that's read-publication).

Scope

Load this skill when the user provides any of:

• A HEPData submission id (ins1234567 — INSPIRE-keyed, or 123/v2 — HEPData-internal versioned id)
• An INSPIRE record id (recid:1234567 or inspire-hep:1234567)
• A paper DOI (10.1103/..., 10.1007/JHEP...)
• A paper title (exact match preferred — HEPData search is fuzzy)

…and they want the tabulated data, not the prose result.

Do NOT load this skill for:

• The headline measured value or the paper narrative → read-publication.
• PDG canonical constants → pdg-lookup.
• Open Data primary datasets (CMS / ATLAS / LHCb data releases on opendata.cern.ch) → cern-opendata.
• ATLAS Monte Carlo metadata by DSID → atlas-opendata.

HEPData URL patterns

Pattern	Meaning
https://www.hepdata.net/record/ins<inspire-recid>	INSPIRE-keyed (most common)
https://www.hepdata.net/record/<hepdata-id>	HEPData-internal id
https://www.hepdata.net/record/ins<id>?version=N	Specific version
https://www.hepdata.net/record/ins<id>?format=json	Machine-readable record metadata
https://www.hepdata.net/search/?q=<query>&format=json	Search by title / DOI / keyword

INSPIRE → HEPData mapping

If the user gives you an INSPIRE recid, the HEPData URL is https://www.hepdata.net/record/ins<recid>. This is 1:1 by construction.

If the user gives you a DOI or title, resolve it to an INSPIRE recid first via the read-publication retrieval recipe (or run INSPIRE search yourself), then build the HEPData URL.

Choosing a format

Each HEPData record exposes its tables in multiple formats. Pick by use case:

Format	URL suffix	When to use
YAML	?format=yaml (per-table)	Most portable; human-readable; pyhf-friendly
ROOT	?format=root	When the user works in ROOT / uproot already
CSV	?format=csv	Quick pandas.read_csv integration
JSON	?format=json	Programmatic record introspection

Default recommendation: YAML for new analysis pipelines, ROOT for ATLAS- internal habits.

Retrieval recipe

1. Resolve the input to an INSPIRE recid or HEPData id.
2. WebFetch https://www.hepdata.net/record/ins<recid>?format=json to get the record metadata. Look at:
   • data_tables[].name and .description to pick the right table(s).
   • data_tables[].data for the per-table data URL.
3. Return per-table URLs to the user with table names and descriptions.
4. Suggest the right downstream pipeline (pyhf for likelihoods, pandas.read_csv for tabular extraction, uproot.open for ROOT inputs).

Common downstream pipelines

After retrieving HEPData tables:

• Re-fit a likelihood → use pyhf if the record exposes workspace.json (HistFactory standard; ATLAS publishes these for many measurements).
• Plot a spectrum → pandas.read_csv + matplotlib / mplhep.
• Combine with another measurement → covariance matrices are usually in a separate table; flag this to the user.

Failure recovery

• HEPData has no record for an INSPIRE recid → not every paper publishes to HEPData. Tell the user; offer to extract the value from the paper via read-publication instead (lossy: just the published number, no uncertainties propagated).
• Search returns multiple records → present top 3 hits with title + experiment + year, ask the user to disambiguate.
• A specific format isn't available for a table → fall back to YAML (always available).