BEACON

Discovery without a resonance

The history of particle physics shows that new phenomena often announce themselves indirectly, well before they can be produced directly: precision measurements anticipated the W and Z bosons, the charm quark, and the mass of the top quark. The SMEFT places this strategy on a systematic, model-independent footing — and for the first time, the measurements, the theory pipeline, and the inference technology exist to pursue it across every frontier simultaneously.

If new physics is heavy and decouples, its imprint at accessible energies — whatever the underlying theory — is a specific point in the space of SMEFT Wilson coefficients. An electric-dipole-moment measurement in a tabletop experiment and a Higgs cross section at the LHC then probe two projections of the same underlying coefficient vector. A statistically significant, internally consistent pattern of nonzero coefficients across energy scales would constitute evidence for new physics even in the absence of a resonance in any spectrum — and the structure of that pattern would point to the mass scale and quantum numbers of the states responsible.

One parameter space, probed by everything

Three properties elevate the SMEFT from bookkeeping to a discovery tool:

Universality

Renormalization-group running of the full 2499-dimensional coefficient space, and complete tree-level and one-loop matching onto the low-energy EFT below the electroweak scale, connect a hypothesized new-physics scale Λ to observables in atomic, nuclear, flavor, collider, and cosmological systems. The pipeline is public: WCxf and wilson for running and matching, flavio/EOS/smelli for low-energy likelihoods, SMEFTsim/SMEFT@NLO for collider simulation, SMEFiT and fitmaker for global fits.

Correlated predictions

A genuine UV completion populates operators in patterns fixed by its symmetries. Example: a 30 TeV Z′ coupled to left-handed fermions is invisible to direct searches, but coherently shifts parity-violating electron scattering (MOLLER), atomic parity violation (Cs), Drell–Yan tails (ATLAS/CMS), coherent neutrino–nucleus scattering (COHERENT), and — through running — flavor observables. Each deviation alone might be 2σ; the joint posterior excludes zero decisively. Correlation of anomalies, not anomaly chasing.

Falsifiability

Truncation uncertainties are quantified, validity criteria are explicit, and corrections are systematically improvable in both couplings and 1/Λ. The analysis sharpens indefinitely as data and theory improve.

What each frontier contributes

Who probes which coefficient

The division of labor across the catalog, made explicit. Left: the 59 Warsaw-basis operator structures versus the regimes that probe them (schematic sensitivities, flavor-diagonal coefficients). Right: the same map at full flavor resolution — at 2499 coefficients, roughly 86% of all directions carry flavor violation, and ownership of the space shifts decisively to the flavor, CLFV, and low-energy frontiers. These assignments are illustrative placeholders; replacing them with fit-derived sensitivities is itself a BEACON deliverable.

Low-energy precision — the deepest probes

The JILA electron-EDM bound |d_e| < 4.1×10⁻³⁰ e·cm reaches hundreds of TeV for tree-level CP violation and multi-TeV even for CP violation first arising at two loops. Neutron EDMs cover quark and gluonic CP violation. Beta-decay correlations are linearly sensitive to scalar/tensor currents; μ→e processes probe flavor violation to 10³–10⁴ TeV — the deepest indirect reach in physics.

Flavor & intensity — per-mille interferometry

LHCb, Belle II, NA62 and KOTO turn rare decays into precision probes of four-fermion operators, and are the key discriminant among flavor structures (MFV, U(2)⁵, …) that any global analysis must confront.

Neutrinos — the blind-spot coverage

DUNE, Hyper-K, IceCube, KATRIN and COHERENT constrain neutrino-sector operators (non-standard interactions, right-handed currents) that no other frontier touches.

Colliders — energy-growing sensitivity

Dimension-six amplitudes grow as E²/Λ²: tails convert luminosity into reach. Global fits already constrain ~50 operator directions with ~445 datapoints; HL-LHC and FCC-ee tighten many by 10–100×. ATLAS and CMS increasingly publish full likelihoods and unbinned measurements built for reuse.

Cosmology — the highest scales, the earliest times

CMB and LSS surveys bound Σm_ν, N_eff, and dark-sector energy injection; dimension-six modifications of the Higgs potential change the electroweak phase transition, linking baryogenesis, gravitational-wave backgrounds (LISA-band), EDMs, and Higgs couplings in one over-constrained system.

From global fits to a SMEFT foundation model

Today's global fits are hand-built: each new experiment costs bespoke expert effort, so the program scales linearly while the data grows faster. Simulation-based inference (MadMiner; ATLAS's neural-SBI off-shell Higgs measurement) and cross-domain foundation-model pretraining (OmniJet-α, RS3L, and the lightweight NEXUS model — arXiv:2607.27501 — which transfers from LHC data to gravitational waves and beyond) break that regime. If unsupervised pretraining on one frontier transfers to another, the shared structure is the question. For BEACON the answer is by construction: the SMEFT is the shared latent space.

The Genesis Mission fit

The Genesis Mission (Executive Order, Nov 24 2025) directs DOE to fuse national-lab computing and federal scientific data into an AI platform for discovery, with scientific foundation models as a named instrument. Its Open Models Initiative solicits pretraining datasets, fine-tuning data/evaluations, and open-weight models on rolling quarterly deadlines; a first $293M funding opportunity ran in spring 2026.

Proposal skeleton: Phase I (9–12 months) = corpus + living fit; Phase II (3 years) = event-level pretraining and unification, with this site as the public progress tracker.

Register, extend, stress-test

Experimental collaborations: register your data products in the catalog, or upgrade an entry's readiness grade by publishing likelihoods/statistical models (HEPData and Zenodo make this straightforward).

Theorists: extend the prediction pipeline's operator and observable coverage — dimension-8, flavor hypotheses, improved matrix elements for low-energy systems.

AI-for-science community: engage with the evaluation suite; the SMEFT corpus is a rare cross-domain benchmark with exact, physics-grade labels.

Contact: saptaparnab@smu.edu · White paper draft v0.1 (August 2026) accompanies this site.