Model Cards for Quantum Technologies Reporting

Mark J Everitt; Siyuan Ji

REVIEW 4 major objections 9 minor 14 references

Model Cards for Quantum Technologies Reporting

T0 review · 4 major / 9 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper argues that every quantum technology should be documented with a standardized Model Card stating intended uses, quantum specifications, errors, performance metrics, and assurance evidence.

desk verdict A decent position paper adapting AI Model Cards to quantum technologies; the template is thoughtful but untested, and the missing worked example is its main weakness. read the letter →

arxiv 2412.13151 v1 pith:XNZIF247 submitted 2024-12-17 quant-ph

Mark J Everitt , Siyuan Ji This is my paper · ORCID

classification quant-ph PACS 03.67.-a

keywords modelcardsquantumtechnologiesdocumentationstandardtransparencyassurabilityperformancemetricstechnologyselectionbenchmarking

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper proposes that quantum technologies—computing, communication, and sensing—be documented with structured Model Cards analogous to those used for machine-learning models. The cards are meant to disclose what an entity is, what it is for, what quantum phenomena underpin it, what can go wrong, how its performance is measured, and what evidence supports those claims. The authors argue this transparency is needed for technology selection, systems integration, risk management, regulatory compliance, and roadmapping. They provide a detailed ten-section template and recommend that cards be encoded with FAIR metadata so they can be searched and mined automatically. The paper is explicitly a starting proposal: the final form would need community agreement and standardisation.

What carries the argument

The central object is the quantum Model Card template, a ten-section structured document: Entity Details; Intended Use; Factors; Quantum Technology Specifications; Errors; Performance Metrics; Ethical Considerations; Evaluation Criteria; Assurability; and Supplementary Materials. Its mechanism is forced disclosure: each section obliges a developer to state not only what a quantum entity is claimed to do, but also its limitations, residual errors, evaluation conditions, and supporting evidence. The framework's extensibility is carried by a recommendation that cards be created with metadata conforming to FAIR principles, with 18-word statements for purpose and use cases, and with supporting documents such as FMEA analyses and assurance cases.

What would settle it

Take any existing quantum device—say, a quantum key distribution link or a small superconducting processor—fill in the proposed card from the developer's documentation, and have a second team independently reproduce the performance metrics by following the card's measurement and evaluation sections. If the card cannot be completed without recourse to undocumented assumptions, or if independent replication yields materially different numbers, the template fails as a transparency device. A simpler observational test: if model cards see widespread use and yet almost none report residual errors, limitations, or failed evaluations, the voluntary-honesty premise has not held.

Watch

Extended reading notes

Core claim

The paper's central claim is that the documentation framework known as Model Cards—originally developed for reporting machine-learning models—can be extended into a general, extensible reporting standard for quantum technologies. It defines the quantum 'entity' broadly enough to cover computers, communication systems, sensors, and sub-components, and it insists the cards must capture quantum-specific information that classical datasheets do not: coherence times, entanglement structure, error sources with non-classical failure modes, error budgets, benchmarks such as quantum volume and QBER, interface and control requirements, and assurance evidence such as certifications, audit reports, and security proofs. On the paper's account, such documentation would let end users evaluate a device against its stated use cases, let systems engineers judge fitness for integration, and let policy makers track the technology ecosystem. The claim is not that any current device meets a quality bar, but that a standardised reporting structure is a necessary step toward transparency and assurability.

Load-bearing premise

The framework's usefulness depends on stakeholders voluntarily providing accurate, complete, and honest information in model cards; the paper itself states that usefulness and accuracy depend on those engaging with it to act with transparency and include sufficient data on assurability, verification, and validation.

Editorial extensions

If this is right

If quantum Model Cards became standard, a systems engineer could compare two candidate devices on the same reported metrics, including residual errors and fundamental limits, before integrating either.
End users could track a deployed entity's performance against the card's stated use cases over time, surfacing drift or degradation.
Regulators and policy makers would gain a common format for assessing risk and compliance across different quantum technology types.
Machine-readable FAIR-compliant cards could feed automated roadmapping and trade-space exploration, linking device-level capability to ecosystem-level targets.
The reporting structure could also help the community build standardised taxonomies and layer models for quantum systems, since card fields would expose where current categories break down.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

One extension the paper leaves implicit is a direct test: fill in the proposed template for an existing quantum device and ask independent evaluators whether the card's claims are reproducible from the supplied measurement details.
If adoption is the goal, the template may work better as a minimal core card plus optional annexes; a too-long mandatory form could push vendors toward vague entries, which would defeat the transparency purpose.
The same card structure could in principle be extended to hybrid classical-quantum systems, since the Interface and Evaluation sections already expect classical components to be specified.
The paper's honesty premise could be tested empirically: if cards become common, check whether reported residual errors and limitations correlate with third-party audit results; if not, incentives beyond voluntary disclosure are needed.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Quick take: this is a reasonable position paper adapting AI Model Cards to quantum technologies. The template is detailed and the writing is honest about its own gaps, but the absence of a single worked example leaves the central claim of completeness and usability unproven. Worth sending to peer review as a position piece, expecting the authors to test the template on a real device.

What's new: the quantum-specific sections—coherence, entanglement, error budgets, assurability—go well beyond Mitchell et al., and the FAIR metadata recommendation is a sensible addition. The paper correctly identifies that quantum systems have failure modes and assurance needs that differ from classical ML. The focus on technology selection, risk management, and policy is a useful framing that the AI model-card literature has only partially addressed.

What it does well: the structure is clear, the template is laid out in enough detail to be actionable, and the paper explicitly acknowledges its main omission (data provenance) in the concluding remarks. It also resists overclaiming: it calls itself a proposal to stimulate discussion, and notes that a final form needs community agreement. Citations are appropriate, with the AI model card and datasheets papers correctly identified as the basis.

Soft spots: the stress-test critique is on target. No worked example means we don't know whether fields like 'Non-local Quantum Coherence' (Section III.D.2), 'Impact Estimates' (III.E), or 'Fundamental limit' (III.F) have well-defined values for a real magnetometer, QKD link, or annealing processor. The admitted data-provenance gap is real and does undercut the FAIR-compliance recommendation. And the paper's own caveat about voluntary transparency is a structural issue: model cards only work if stakeholders fill them in accurately, and no incentive mechanism is proposed. These are not fatal for a position paper, but they are the difference between a proposal and a standard.

I'd treat the 'UNVERDICTED' label as fine — this isn't a testable research claim. But if I were an editor, I'd still send it out: the quantum-standards community needs proposals from which to argue. My recommendation: accept for peer review, but the referee should ask for at least one worked example and a more concrete handling of provenance. It's a useful contribution for people working on benchmarking, tech evaluation, and roadmapping.

Referee Report

4 major / 9 minor

Summary. This manuscript proposes extending the Model Cards documentation framework, originally developed for machine learning models in ref. [2], to quantum technologies. The authors argue that emerging quantum technologies in computing, communication, and sensing need standardized, transparent documentation covering intended use, performance characteristics, evaluation conditions, and assurance information. The paper presents a detailed template with ten sections: Entity Details, Intended Use, Factors, Quantum Technology Specifications, Errors, Performance Metrics, Ethical Considerations, Evaluation Criteria, Assurability, and Supplementary Materials. The proposal additionally recommends FAIR-compliant metadata and emphasizes extensibility to accommodate future standards. The manuscript is explicitly positioned as a proposal to stimulate community discussion rather than a validated or finalized standard.

Significance. Strengths: the proposal is internally consistent, well-referenced, and builds on an established concept (AI Model Cards) with known traction for transparency reporting; it adds quantum-specific content (coherence, entanglement, QBER, FMEA, assurance cases, FAIR metadata) that goes beyond naive adaptation; and it is honest about its own limitations, explicitly flagging the data-provenance gap in Section IV. Weaknesses: the paper provides no worked example, no data, and no case study, so the claim that the template is complete and general enough for all quantum technology types is asserted rather than demonstrated; and the stated usefulness condition in Section IV depends on voluntary, accurate participation with no proposed mechanism to secure it. If validated by application to real entities, the framework could genuinely support technology selection, risk management, and regulatory assurance for quantum technologies; as it stands, its value is programmatic rather than demonstrated.

major comments (4)

[Section III.A–I and Section IV] The central claim that the proposed template is complete, general, and extensible for all quantum technologies is never demonstrated. The template is not applied to a single concrete entity——not a quantum computer, a QKD system, or a quantum sensor——so there is no evidence that every field is well-defined and completable with meaningful, non-vacuous entries. This is load-bearing because the paper's purpose is to propose a documentation standard for stakeholders to adopt; a field that cannot be filled for real entities would undermine the uses claimed in the abstract (technology selection, risk management, compliance). The concluding remarks defer validation to future stakeholder feedback (Section IV), confirming that the completeness claim is currently unsubstantiated. I recommend adding at least one fully worked example, together with a per-section analysis of which fields are mandatory, conditional, or not applicable for each of the three main technology families (computing, communication, sensing), or explicitly re-scoping the claim as a draft template pending such a study.
[Section III (FAIR recommendation) vs. Section IV] The paper recommends that model cards conform to the FAIR Guiding Principles (ref. [5]), stating 'we make a further recommendation beyond that of [2] and assert that model cards be created using appropriate metadata conforming to FAIR Guiding Principles.' However, Section IV concedes that data provenance is 'one strong omission in our current proposal.' Provenance is not an optional refinement: under FAIR, reuse requires rich provenance metadata (R1.2), and traceability of measurement conditions is arguably central to a documentation standard whose stated aims include transparency and risk management. As written, the FAIR recommendation and the admitted omission are in direct tension: a card whose metrics cannot be traced to their measurement conditions cannot be FAIR-compliant, nor can it support the assurance claims made in Section I. The authors should either integrate a provenance mechanism (for example, a required measurement-conditions and data-lineage block in Sections E and F, or explicit linkage to a provenance standard) or explicitly delimit the FAIR recommendation.
[Sections III.D.2, III.E, III.F] Several fields presuppose that well-defined, obtainable quantitative values exist for entities of every type. 'Non-local Quantum Coherence' (III.D.2) requires an enumeration of all entanglement resources; 'Impact Estimates' (III.E) requires quantitative or qualitative impact for each error source; and 'Fundamental limit' (III.F) requires a 'theoretical best possible performance' for each metric. For many real systems, particularly sensing and communication devices and noisy intermediate-scale processors, such values may be unknown, contested, or not well-defined at the level the template implies. The paper asserts that these fields should be filled but gives no account of how to determine them, and no example. Without a demonstrated method for populating these fields, the template risks producing empty or invented entries, which would defeat the transparency goal. A worked example or an explicit field-level feasibility analysis is needed.
[Section IV (Concluding Remarks)] The manuscript's own success condition is behavioral: 'The usefulness and accuracy of this approach will depend on those engaging with it to act with transparency and include sufficient data on assurability, verification, and validation.' The paper proposes no mechanism, incentive, or independent validation step to secure accurate self-reporting, yet the abstract claims the approach will enable stakeholders to 'manage risk and assure compliance with regulatory frameworks.' For a documentation standard aimed at risk management and regulatory assurance, reliance on unverified self-reporting is a load-bearing concern. This is not by itself a reason to reject a proposal paper, but it should be addressed in the text——for example, by discussing what role independent audits, standardized benchmarks, or a registration body could play, or by referencing how the AI Model Cards and Datasheets literature has handled the analogous problem.

minor comments (9)

[Section I] The phrase 'unital proposal' appears to be a typo for 'initial proposal'; if the word is intentional, it should be clarified.
[Section II] The unresolved citation placeholder '[ ? ]' after 'as either the ITRS or the IDRS have benefitted from [ ? ]' must be replaced with a reference or removed.
[Author affiliations] The affiliation 'Loughborough Universi ty' contains an extra space between 'Universi' and 'ty'.
[Section III.A] In the Purpose example, 'a magnetic flux sensor for non-destructive testing in of defects in aircraft composite structures' contains 'in of'; the intended phrasing is presumably 'testing of defects' or 'testing for defects'.
[Section III.D.2] The text 'For composite carries such as logical qubits' should read 'carriers', and the phrase 'with ζ having configurable φ and r within the region of 4' is unclear about the intended parameter range and units.
[Section III.E] The sentence 'Information covered by the above points is inherently connected through casual chains' should read 'causal chains'.
[Section IV] The sentence 'The next is to this reporting structure based on feedback obtained from a range of stakeholders' is garbled; the intended meaning appears to be 'The next step is to refine this reporting structure based on feedback.'
[References] The URL for reference [8] is truncated in the bibliography ('https://incose.onlinelibrary.wiley.com/doi/pdf/10.1002/j.2334-5837.20'); a complete DOI or stable URL should be provided.
[Section II and Section III] The paper would benefit from engaging with known critiques of the original AI Model Cards framework as experienced in the AI community, such as the burden of completing many fields and the difficulty of verifying self-reported information, since these experiences directly inform design choices in Sections F and I.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper is an openly normative documentation proposal, not a derivation that reduces to its own inputs.

full rationale

This manuscript contains no mathematical derivation chain, no fitted parameters, and no equations whose outputs are equivalent to their inputs by construction. Its central claim is a normative proposal to extend the externally established AI Model Cards framework [2] to quantum technologies, explicitly framed as a 'unital proposal to stimulate discussion' (Section I) whose final form would need community agreement and standardization. The template sections in Section III are presented as suggested content areas, not as predictions or first-principles results. The paper's own admitted limitation, 'One strong omission in our current proposal is that of data provenance' (Section IV), and its acknowledgment that usefulness depends on stakeholders acting transparently, are openly stated dependency conditions rather than hidden circular assumptions. The skeptical concern that template completeness is asserted without a worked example is a validity or completeness risk, not a circularity defect. No self-citation is load-bearing: reference [7] is a Loughborough PhD thesis, but the proposal does not rest on it, and the core inspiration is the independent, externally published Model Cards work of Mitchell et al. Therefore no circular step can be exhibited, and the appropriate score is 0.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

No free parameters or invented entities are introduced. The paper's assumptions are qualitative premises about the value of transparency, the analogy to AI model cards, the generality of the template, and stakeholder honesty. These are acknowledged in the text, not hidden.

assumptions (5)

domain assumption Transparency through standardized documentation improves stakeholder decision-making and trust.
This is the stated motivation in Sections I and II; no empirical evidence is provided that model cards cause these outcomes.
domain assumption Quantum technologies share enough assurance challenges with AI that Model Cards are a suitable template.
Section II draws a direct analogy to Mitchell et al. [2] and asserts substantive modifications are needed; the suitability is asserted, not demonstrated.
domain assumption A single generalized and extensible card can accommodate quantum computing, communication, and sensing.
Section III claims a general framework, but no case studies are given to show the template fits all three modalities.
domain assumption Stakeholders will act with transparency and provide sufficient data.
Section IV explicitly makes the usefulness dependent on this, saying it 'will depend on those engaging with it to act with transparency'.
domain assumption The proposed sections are a minimal set covering functional, structural, and interface definitions.
Section III.D.4 asserts this; no external standard is used to justify the minimal set.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Model Cards for Quantum Technologies Reporting." pith.science (2026). https://pith.science/paper/XNZIF247

@misc{pith2026241213151,
  author       = {Pith},
  title        = {Pith review of: Model Cards for Quantum Technologies Reporting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XNZIF247}},
  note         = {Machine review of arXiv:2412.13151}
}

read the original abstract

There are a number of emerging quantum technologies that have the potential to be disruptive in application areas such as computation, communication and sensing. In such a rapidly emerging field, there is a need for: transparency and accountability pertaining to devices, their performance, and limitations; the ability to assess new entities for integration into existing systems; sufficient information to undertake technology selection; share knowledge within and across institutions and discipline domains; manage risk and assure compliance with regulatory frameworks; drive innovation. Here we propose Model Cards for documentation detailing use-cases and performance characteristics of entities for use in quantum technologies. Purpose of this document is therefore to stimulate discussion and begin to motivate the community to build a sufficient body of knowledge so that the most useful form of Model Cards can be developed and standardised.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

14 extracted references · 13 canonical work pages

[2]

Circuit parame- ters for each component should be given

Hardware Speciﬁcation Carriers of Quantum Information: An enumerated list (and frequency count) of each and every physical component that contains functionalised non-classical in- formation (superconducting device realisation of a tran- som qubit, cavity resonator, photons). Circuit parame- ters for each component should be given. For composite carries su...

work page
[5]

Blume-Kohout and K

R. Blume-Kohout and K. C. Young, A volumet- ric framework for quantum computer benchmarks, Quantum 4, 362 (2020)

work page 2020
[1]

This includes quantum gates, quantum opera- tions, and quantum algorithm’s tailored to the system’s capabilities

System Architecture Quantum Process/Algorithms: Deﬁne the quan- tum processes or algorithms used to achieve the deﬁned use cases. This includes quantum gates, quantum opera- tions, and quantum algorithm’s tailored to the system’s capabilities. Circuit Design: High-level description or schematics of the quantum circuit design (which might be a layer model ...

work page
[3]

Data Handling: Formats and protocols for how the in- terfacing data processing systems handle the transmitted data

Interface Speciﬁcation Data Type: Description of the physical or logical data that is being transmitted, internally within the entity and externally with other systems. Data Handling: Formats and protocols for how the in- terfacing data processing systems handle the transmitted data. Potential Issues: May include back action, vulnerabil- ity to electrical...

work page
[4]

Other Approaches There are diﬀerent frameworks for how a system spec- iﬁcation can be captured and represented. The above strategy is considered a minimal set that covers the three key elements for any (quantum or classical) system archi- tecture: namely, functional, structural, and interface def- inition. Other research, [7], has proposed a layer model f...

work page
[6]

Mitchell, S

M. Mitchell, S. Wu, A. Zaldivar, P. Barnes, L. Vasser- man, B. Hutchinson, E. Spitzer, I. D. Raji, and T. Gebru, Model cards for model reporting, in Proceedings of the Conference on Fairness, Accountability , and Transparency , F AT* ’19 (Association for Computing Machinery, New York, NY, USA, 2019) p. 220–229

work page 2019
[7]

Gebru, J

T. Gebru, J. Morgenstern, B. Vecchione, J. W. Vaughan, H. Wallach, H. D. I. au2, and K. Crawford, Datasheets for datasets (2021), arXiv:1803.09010 [cs.DB]

arXiv 2021
[8]

P. A. Gargini, Overcoming semiconductor and elec- tronics crises with irds: Planning for the future, IEEE Electron Devices Magazine 1, 32 (2023)

work page 2023

Show all 14 references

[9]

M. D. Wilkinson, M. Dumontier, I. J. Aalbersberg, G. Appleton, M. Axton, A. Baak, N. Blomberg, J.-W. Boiten, L. B. da Silva Santos, P. E. Bourne, and et al., The fair guiding principles for scientiﬁc data management and stewardship, Nature News (2016)

2016
[10]

Burge, 18 words statement (the systems thinking tool box) (2015)

S. Burge, 18 words statement (the systems thinking tool box) (2015)

2015
[11]

Lowe, Engineering Design and Quantum Computing (subbmtted version), Ph.D

D. Lowe, Engineering Design and Quantum Computing (subbmtted version), Ph.D. thesis, Loughborough Univer- sity (2022)

2022
[12]

Oakes, R

J. Oakes, R. Botta, and T. Bahill, 11.1.1 technical performance measures, INCOSE International Symposium 16, 1466 (2006) , https://incose.onlinelibrary.wiley.com/doi/pdf/10.1002/j.2334-5837.20

2006 doi
[13]

R. Wei, T. P. Kelly, X. Dai, S. Zhao, and R. Hawkins, Model based system assurance using the structured as- surance case metamodel, Journal of Systems and Soft- ware 154, 211 (2019)

2019
[14]

Lodewyck, T

J. Lodewyck, T. Debuisschert, R. Garc ´ ıa-Patr´ on, R. Tualle-Brouri, N. J. Cerf, and P. Grangier, Ex- perimental implementation of non-gaussian attacks on a continuous-variable quantum-key-distribution system, Phys. Rev. Lett. 98, 030503 (2007)

2007

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.

[1] [2]

Circuit parame- ters for each component should be given

Hardware Speciﬁcation Carriers of Quantum Information: An enumerated list (and frequency count) of each and every physical component that contains functionalised non-classical in- formation (superconducting device realisation of a tran- som qubit, cavity resonator, photons). Circuit parame- ters for each component should be given. For composite carries su...

work page

[2] [5]

Blume-Kohout and K

R. Blume-Kohout and K. C. Young, A volumet- ric framework for quantum computer benchmarks, Quantum 4, 362 (2020)

work page 2020

[3] [1]

This includes quantum gates, quantum opera- tions, and quantum algorithm’s tailored to the system’s capabilities

System Architecture Quantum Process/Algorithms: Deﬁne the quan- tum processes or algorithms used to achieve the deﬁned use cases. This includes quantum gates, quantum opera- tions, and quantum algorithm’s tailored to the system’s capabilities. Circuit Design: High-level description or schematics of the quantum circuit design (which might be a layer model ...

work page

[4] [3]

Data Handling: Formats and protocols for how the in- terfacing data processing systems handle the transmitted data

Interface Speciﬁcation Data Type: Description of the physical or logical data that is being transmitted, internally within the entity and externally with other systems. Data Handling: Formats and protocols for how the in- terfacing data processing systems handle the transmitted data. Potential Issues: May include back action, vulnerabil- ity to electrical...

work page

[5] [4]

Other Approaches There are diﬀerent frameworks for how a system spec- iﬁcation can be captured and represented. The above strategy is considered a minimal set that covers the three key elements for any (quantum or classical) system archi- tecture: namely, functional, structural, and interface def- inition. Other research, [7], has proposed a layer model f...

work page

[6] [6]

Mitchell, S

M. Mitchell, S. Wu, A. Zaldivar, P. Barnes, L. Vasser- man, B. Hutchinson, E. Spitzer, I. D. Raji, and T. Gebru, Model cards for model reporting, in Proceedings of the Conference on Fairness, Accountability , and Transparency , F AT* ’19 (Association for Computing Machinery, New York, NY, USA, 2019) p. 220–229

work page 2019

[7] [7]

Gebru, J

T. Gebru, J. Morgenstern, B. Vecchione, J. W. Vaughan, H. Wallach, H. D. I. au2, and K. Crawford, Datasheets for datasets (2021), arXiv:1803.09010 [cs.DB]

arXiv 2021

[8] [8]

P. A. Gargini, Overcoming semiconductor and elec- tronics crises with irds: Planning for the future, IEEE Electron Devices Magazine 1, 32 (2023)

work page 2023