A common Rivet validation layer across POWHEG, GENEVA, MCFM and JetPhox
POWHEG ↔ Pythia/Herwig ↔ Rivet. POWHEG generates events as Fortran
common blocks (HEPEUP); Rivet expects a C++ AnalysisHandler fed HepMC
events. The interface sits directly on that boundary, converting POWHEG’s
native event record into HepMC for Pythia 6, Pythia 8.2/8.3 or Herwig 7
indiscriminately through one abstraction layer, with explicit staging for
POWHEG-BOX-RES’s multi-stage radiation generation (separate hooks for the
Born-level, first-radiation and shower-matched stages) so a Rivet analysis
can run at whichever stage is actually needed. It also taps directly into
POWHEG’s native reweighting infrastructure (pwhg_weights), so scale- and
PDF-variation weights registered by POWHEG propagate through as proper
multi-weight Rivet/YODA histograms rather than single central-value
curves — the uncertainty bands are real, not bolted on afterwards.
Upstream ALICE analyses. Several of the analyses these interfaces run are not just used but authored and merged directly into Rivet itself: ALICE_2025_I2895564 (isolated prompt photon production in pp and p-Pb), ALICE_2024_I2803487 and ALICE_2019_I1738300 (inclusive isolated photon production at 13 and 7 TeV), and ALICE_2020_I1797621 and ALICE_2018_I1672792 (di-electron production in pp and p-Pb) — plus assorted validation fixes to several other ALICE analyses already in the library. These are exactly the observables (isolated photons, low-mass dielectrons) the research on this site is about, now available to anyone running Rivet, not just internally.
GENEVA validation. Validating GENEVA’s NNLO-QCD + NLO-EW Drell-Yan predictions means checking several perturbative combinations (NLO and NNLO QCD, with and without NLO electroweak and photon-induced contributions) at two LHC energies against each other and against published measurements. The validation analysis books every observable twice — once at the nominal order and once as an explicit NNLO-QCD variant with matching binning — and several histograms are booked with bin edges taken directly from specific figures in published papers, so the generated distributions overlay the original plots exactly rather than needing to be rebinned by eye afterwards.
MCFM and JetPhox. MCFM books histograms through its own internal “nplotter” system, not Rivet; a small converter exposes the same differential information as a proper Rivet analysis instead, which is what made it possible to cross-check the photon-fragmentation extraction against MCFM on equal footing. JetPhox has a similar gap on the output side: it writes events to ROOT ntuples rather than any event-record format Rivet understands, so a converter turns those ntuples into HepMC3 events, letting unmodified, standard Rivet analyses run on JetPhox predictions directly.