IQ Intel Request access
IQ Intel

What do you need to validate?

SaC is IQ Intel’s Quantum State Reference & Validation Architecture. Choose the application that matches your work.

Quantum circuit validation Materials science
Quantum Validation

Did your quantum computation produce the right state—or just the right number?

For eligible standard gate-model quantum computations, IQ Intel can construct an independent SaC reference, retain direct access to the reference state where tractable, recompute agreed observables, and referee the supplied result against criteria frozen before grading.

Public evidence

  • VQE: 0.56% energy error while the final H4 state contained approximately 8.6×10−13 ground-state fidelity.
  • Variance: the wrong state still passed an energy-variance eigenstate check because it was an excited eigenstate.
  • EMR: 35 preregistered predictions, truth-blind grading, 2 correct; mitigation improved an observable without thereby certifying the state.
Explore quantum validation VQE study EMR study
Materials Science

Do you need snapshots—or a validated record of how a material moves through state space?

SaC can evolve modeled material systems through connected microstates under applied conditions, preserving lineage and a multidimensional observable record at every state. IQ Intel uses those trajectories to build proprietary materials-intelligence datasets for client programs, with validation defined as part of the work rather than added afterward.

Validation evidence

  • NMC811-IQ: six circuit families converge on the same H1→H2 transition; 80/80 cross-checks agree with the seam backbone.
  • MagNet-IQ: order-parameter direction, Curie–Bloch reemergence, cross-measure consistency, null controls, and algebraic integrity are checked independently.
  • 14–19 properties per state: the dataset is validated as a system rather than by asking one plotted property to carry the claim.
Explore materials science Discuss a dataset program
SaC is IQ Intel’s Quantum State Reference & Validation Architecture. Current demonstrated operating envelope includes QASM 2.0 standard gate-model circuits, lattice systems up to 10×10×10, and workloads approaching 3 million gates. Eligibility for reference validation depends on circuit structure, depth, requested state access, observables, and available compute rather than qubit count alone.
Quantum Reference Validation

What endpoint metrics can miss.

The question is not whether an energy, probability, expectation value, or confidence interval looks plausible. The validation question is whether the eligible computation produced the state and result it was intended to produce.

VQE state-verification study
0.56%

Energy error from exact on H4 at 2.5 Å—while the final state contained approximately 8.6×10−13 ground-state fidelity.

Energy-variance check
≈ 0

The wrong state also passed the variance test because it was an excited eigenstate. A strong endpoint check certified the wrong state.

EMR referee study
2 / 35

Preregistered predictions correct under sealed grading. The result also showed that improving an observable does not, by itself, certify the underlying state.

VQE state-verification study

Standard molecular VQE circuits, optimizer-path replay, exact diagonalization, eigenstate overlaps, state fidelity, and energy-variance analysis. The complete reproducibility package is public.

Open VQE study & package →

Error-mitigation referee study

A preregistered 35-cell matrix with published prediction hash, truth-blind grading, selected ZNE/PEC/readout-mitigation workflows, fit diagnostics, state-vs-observable analysis, and a reproducible referee package.

Open EMR study & package →
Initial service envelope

What IQ Intel can referee today.

Eligibility is determined before an engagement is accepted. The initial service is designed for standard gate-model circuits and workflows that fall within a tractable, independently reproducible reference envelope.

  • Circuit compatibility: QASM 2.0 standard gate-model circuits within the supported SaC instruction subset, parameterized gate sequences, hardware-efficient ansätze, Trotter-style circuits, and tested Pauli-operator/Hamiltonian workflows.
  • State validation: exact state-vector comparison where tractable, fidelity, overlaps, target/eigenstate content, and comparison of nominally equivalent outputs.
  • Observable validation: energy, probabilities, populations, variances, operator expectations, and other agreed state-derived quantities.
  • Optimization-path analysis: starting-state vs final-state comparison, restart behavior, local traps, divergence, and recorded optimizer evaluations.
  • Error-mitigation referee: unmitigated vs mitigated bias, confidence-interval coverage, selected ZNE/PEC/readout workflows, fit behavior, overhead, and observable-vs-state comparison.
  • Hardware-output comparison: compare supplied hardware measurements or observables against the independent SaC computational reference under an agreed protocol.

This is computational reference validation, not universal QPU certification. Pulse-level device characterization, arbitrary circuit sizes, unsupported semantics, and hardware properties not observable from the computation remain outside the initial service unless separately demonstrated and contracted.

Quantum Reference Validation Pilot

Give us a computation you already trust.

We will jointly define the claim and acceptance criteria before grading, construct the independent SaC reference, inspect the state where tractable, and return a reproducible PASS / FAIL / INCONCLUSIVE referee package.

Request a validation pilot Review public evidence
One architecture · two applications

Reference the state. Then validate what the result actually supports.

Quantum computing and materials engineering are different markets, but they share the same validation problem: important information can be lost when a complex state is reduced to an endpoint number or an isolated snapshot.

Quantum Validation

Referee a supplied computation.

Standard circuit → independent SaC reference → retained state and observables → predeclared comparison → referee finding.

Materials Science

Generate and validate connected material states.

Material model → applied conditions → connected SaC microstates → lineage + multidimensional observables → dataset-specific physical validation.

Define
claim & criteria
Freeze
inputs & protocol
Reference
state + observables
Compare
against target
Publish
what survived
Materials Science

The same validation architecture applied to matter.

Instead of refereeing a supplied quantum computation, SaC can evolve a modeled material system through connected microstates under applied conditions, retaining state lineage and a common multidimensional observable record throughout the process.

For IQ Intel, validation is not a layer added after a dataset is produced. It is part of how the work is designed: define what the output must demonstrate, preserve enough information to test it, and let the result stand or fail on the data.

NMC811-IQ is cross-checked across six quantum circuit families on the H1→H2 transition. MagNet-IQ is tested through order-parameter direction, critical-scaling behavior, cross-measure consistency, null controls, and algebraic integrity. Both datasets preserve the state-level record rather than asking a single plotted property to carry the validation burden.

6 from 1 NMC811 circuit families from one architecture
80 / 80 H1→H2 cross-checks pass vs seam backbone
14–19 numeric properties / observables moving with each state

NMC811-IQ — cross-family validation

Connected charge and discharge trajectories are validated through observable reproducibility, structural-transition positioning, ion-mobility behavior, charge/discharge symmetry, continuity, and six-family convergence around the H1→H2 transition.

Claimed: structural turnover on both directions at literature-associated lithium content; mobility and structural analyses co-locate; observables reproduce from source states; charge/discharge mirror within 0.004 Li fraction.

Not claimed: blanket “matches experiment” or universal physical equivalence outside the validated targets.

NMC811-IQ Validation Reference (PDF) →

MagNet-IQ — NdFeB process validation

The public manufacturing seam is evaluated through order-parameter direction, Curie–Bloch reemergence, cross-measure consistency, null controls for antiferromagnetic and transverse components, algebraic integrity, and process-stage continuity.

Claimed: primary order-parameter direction through each stage vs Nd2Fe14B Tc; critical-scaling reemergence shape; cross-measure consistency; and null-control behavior.

Not claimed: extrinsic coercivity/remanence/BHmax, low-temperature spin reorientation, or unvalidated finishing-stage physics.

MagNet-IQ Validation Protocol (PDF) →
In plain English: when more than a dozen properties are moving collectively across the same region of state space, that is likely the strictest validation we can face outside of putting the material in a laboratory.
Materials datasets are IQ Intel assets. IQ Intel no longer publishes the underlying NMC811-IQ or MagNet-IQ datasets as open data. They are retained as proprietary reference assets and examples of the dataset class available through client and partner engagements.
Public evidence

Read the studies. Review the validation. See what the architecture has already survived.

This site is IQ Intel’s public evidence base. The VQE and error-mitigation referee studies, including their reproducibility packages, are part of the validated asset. IQ Intel’s materials datasets are proprietary assets used in client and partner programs; the validation methods and selected benchmark results are presented here to show what the architecture has already demonstrated.

VQE state-verification

Endpoint energy and variance can certify the wrong eigenstate. Full public package and checksums.

Open VQE study →

Error-mitigation referee

Preregistered 35-cell study, truth-blind grading, state-vs-observable analysis, and reproducible referee package.

Open EMR study →

NMC811-IQ

Connected cathode microstates across charge and discharge, up to 19 observables per state, with six-family H1→H2 cross-validation.

Review NMC811-IQ validation →

MagNet-IQ

Connected NdFeB manufacturing microstates with approximately 14 numeric measures plus classifier labels and a tiered validation protocol.

Review MagNet-IQ validation →

Flat benchmarks ask you to infer motion between points. These datasets show motion: each microstate is the direct downstream evolution of the last. Load a seam, scrub frame by frame, and read the observables at every step — lithiation fraction, manufacturing stage, charge direction, and lineage included.

NMC811-IQ dashboard at 10x10x10 showing a lithiating state near the H1 to H2 transition.
NMC811-IQ · 10×10×10 · lithiating view near the H1→H2 transition.
NMC811-IQ dashboard at 10x10x10 showing a delithiating state near the H1 to H2 transition.
NMC811-IQ · 10×10×10 · delithiating view at the same state region for directional comparison.
MagNet-IQ dashboard at 10x10x10 showing one frame from the step 4 field anneal stage.
MagNet-IQ · 10×10×10 · field-anneal frame 1.
MagNet-IQ dashboard at 10x10x10 showing the consecutive next frame from the step 4 field anneal stage.
MagNet-IQ · 10×10×10 · consecutive field-anneal frame 2.
Representative 10×10×10 dashboard views. IQ Intel materials programs retain state, lineage, and multidimensional observables frame by frame, allowing validation to be examined as the process evolves rather than inferred from isolated snapshots.

Two material systems today; the architecture is domain-agnostic. Browse what is open, read how each dataset was validated, and request access when you need the full range or a dataset built for your stack.

Materials assets

Two demonstrated material systems, one validation discipline.

NMC811-IQ and MagNet-IQ are proprietary IQ Intel data assets produced by the same SaC architecture and validation discipline. They demonstrate the form of connected, state-resolved materials intelligence IQ Intel can build for client programs.

IQ Intel asset validated charge/discharge trajectories · 1–99% program coverage

NMC811-IQ

Ballistic transport · cathode material

Four complete connected seams (two lithiation, two delithiation) spanning 1–99% lithiation. Public windows at 10–25% and 60–75% SOC across all four seams — compare charge direction and independent runs across the validated program.

Proprietary IQ Intel materials asset
IQ Intel asset 711-state manufacturing trajectory · validated process stages

MagNet-IQ

Manufacturing process · permanent magnets

Single connected manufacturing trajectory of 711 microstates from green-compact powder through field anneal. 444 frames (P0–B reemergence) are part of the validated asset; used to demonstrate process-stage continuity, observable evolution, and validation behavior.

Proprietary IQ Intel materials asset
Can you verify it yourself?

What is demonstrated today.

Validation is only as useful as what someone can check themselves. Here is exactly what is demonstrated today, and what sits behind engagement, for both datasets.

NMC811-IQ

Four complete connected seams — two lithiation, two delithiation — each spanning the cathode operating range. The same two windows (10–25% and 60–75%) are part of the validated assetly exposed across all four seams. Compare the same SOC window across independent runs and charge directions in the dashboard viewer today. Complete seams are available to approved entities.

MagNet-IQ

Full manufacturing run: 711 microstates from green-compact powder through field anneal. 444 frames spanning powder through reemergence (P0–B) are part of the validated asset in the dashboard viewer today. Finishing stages and the remainder of the run are available to approved entities.

Access

Two tiers, with verification on the second.

Public

Benchmark range

The benchmark range of each dataset is available without authentication. Browse the fractions and seams, inspect microstate detail pages, load any microstate or seam into the dashboard viewer. No account required.

  • NMC811-IQ: 10–25% & 60–75% lithiation/delithiation, sanitized JSON and plots
  • MagNet-IQ: P0 through B reemergence (444 microstates, 1 public manufacturing seam)
  • Full viewer access for any open file
Browse the benchmark range →
Private

Client dataset program

Complete IQ Intel materials datasets are proprietary client and partner assets. Access is provided only under an executed engagement or other approved agreement.

  • NMC811-IQ: full 1–99% range
  • MagNet-IQ: full manufacturing run (711 frames through field anneal)
  • Manual approval; expect a turnaround of several business days
Request access
Intended uses

What this data is for.

  1. 01

    Reproducibility against the benchmark range

    The benchmark range exists so that anyone can sanity-check the methodology against existing knowledge for both material systems. Run the published microstates through your own analysis — NMC811 spin observables at fixed SOC windows, MagNet order-parameter trajectories through manufacturing stages. Disagreement is the kind of finding we want to know about.

  2. 02

    Training and evaluation data for materials models

    Each microstate carries a full set of observables alongside lattice-resolved bitstrings and per-site fields. Conventional training data for materials ML is thin at this resolution and rarely covers connected trajectories across charge direction or process history. Both benchmark datasets are candidate sources where ground truth is otherwise costly, slow, or unavailable.

  3. 03

    Methodology validation before engagement

    A program considering a Subatomic Computing engagement can read the published data first. Inspect what the engine produces. Read the observables. View the seam trajectories in the dashboard. The decision to engage should rest on what the methodology does, not on what the methodology is described to do.

  4. 04

    Cross-referencing with the Signals series

    The findings discussed in the Signals blog series — multi-variant microstates at fixed state of charge, connected state evolution as a methodology, what defect characterization surfaces at quantum resolution — are illustrated against this dataset. Read the posts alongside the data; the data is the ground truth the posts describe.

  5. 05

    Citation in published work

    The dataset is intended to be citable in publications that build on it. Citation guidance, DOI assignment, and the licensing terms for the benchmark range are being finalized and will be posted here before formal release. Researchers preparing manuscripts that reference the benchmark data are welcome to contact us in the interim.

More about NMC811-IQ →  ·  More about MagNet-IQ →

Research areas

NMC811-IQ: path dependence in ballistic transport

An emerging pattern from stitched seam analysis on the battery benchmark. MagNet-IQ process trajectories raise parallel questions about stage-to-stage memory; the validation seam is available for independent analysis.

One of the notable patterns emerging from our stitched seam analysis is the apparent path dependence of microscopic spin configurations at equivalent macroscopic lithium fractions.

Across multiple independently generated seams, we observe both:

  • Convergent microstates — identical or near-identical spin configurations (as measured by bitstring patterns, magnetization components, coherence proxy, and ordering strength) appearing in seams of both lithiation and delithiation direction.
  • Divergent microstates — distinct configurations reachable only through specific charge histories or directions.

These observations are consistent with theoretical expectations of history-dependent behavior in strongly correlated systems. However, at this stage we present them as a possible finding rather than a definitive conclusion.

Our validation has been limited to confirming that the generated microstates remain within the physically plausible domain of the underlying Hamiltonian and align with expected regimes (e.g., high-SOC, mid-SOC, and deep delithiation) according to established classification criteria used in the main results dashboard.

Detailed definitions of the key observables are documented at iqintel.io/#observables.

Data Availability & Invitation for Further Research

The complete seam datasets, master seams, and raw state files are made openly available in the repository. We encourage researchers and engineers in battery materials, quantum simulation, statistical mechanics, and related fields to examine these trajectories.

In particular, we welcome further analysis that could:

  • Quantify the extent of path dependence and hysteresis
  • Test for ergodicity breaking or multiple metastable manifolds
  • Explore implications for voltage hysteresis and kinetic limitations in NMC811-IQ and related cathodes
  • Develop improved reduced-order models that incorporate path memory

Any confirmation, extension, or alternative interpretation of these observations would be of significant interest to the community.

Methodology

Subatomic Computing, in brief.

The benchmark datasets are produced by an engine that emerges microstates from a modeled Hamiltonian rather than computing properties for configurations the user specified. Each microstate carries a standardized observable set (up to 19 measures on NMC811-IQ; roughly 14 numeric measures plus classifier labels on MagNet-IQ). Sequences of quantum microstates are validated against physical continuity criteria and assembled into master seams — complete charge or discharge cycles in the battery case, coupled manufacturing trajectories in the magnet case.

The output is connected, trajectory-resolved, multi-observable material data. Validation protocols and tier tables are on this page; the full methodology narrative lives at iqintel.io/about/methodology, with the observable set at iqintel.io/#observables and cross-family validation at iqintel.io/#validation.

Two ways to work with IQ Intel

Choose what you need to validate.

Quantum Reference Validation

Give us a computation you already trust.

Use SaC as an independent state-resolved reference for an eligible circuit, algorithm, mitigation workflow, or hardware output.

  • Client-selected circuit or workflow
  • Predeclared validation criteria
  • State-level analysis where tractable
  • Reproducible referee package
Request validation pilot
Materials Dataset Program

Give us a material system or process you need resolved.

IQ Intel generates connected, lineage-preserving material states under applied conditions and validates the resulting dataset against agreed physical targets.

  • Operating-cycle trajectories
  • Manufacturing process trajectories
  • Transition / seam studies
  • Custom multidimensional observable sets
Discuss materials program
Access

Work with IQ Intel

Select the type of work you want to discuss. Quantum-validation pilots, materials dataset engagements, and research collaborations are reviewed manually.

Validation first

Bring us the quantum computation or material system you need independently resolved.

Review the public evidence first. Then define the validation target with us before the work begins.

Quantum validation Materials science Work with IQ Intel