Pith. sign in

REVIEW 2 major objections 2 minor 25 references

Q-DASC wraps variational quantum circuit policies with a certified classical safety layer to keep HVAC comfort violations near zero even under local model misspecification and quantum readout noise.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.3

2026-06-30 09:04 UTC pith:J2RZFTL3

load-bearing objection Q-DASC adds a classical FDR-plus-shrinkage wrapper around VQC policies for HVAC that cuts comfort violations to near zero on BOPTEST emulators even under local misspecification and NISQ noise. the 2 major comments →

arxiv 2606.28834 v1 pith:J2RZFTL3 submitted 2026-06-27 eess.SY cs.SY

Q-DASC: State-of-the-Art Safe Quantum Control for HVAC under Local Model Misspecification

Yifan Wang This is my paper
classification eess.SY cs.SY
keywords quantum reinforcement learningsafe controlHVACmodel misspecificationbuilding energy managementvariational quantum circuitscomfort violationdiscrepancy attribution
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper establishes that a hybrid approach can deliver safe deployment of compact quantum policies in building energy control by adding a classical wrapper that detects and repairs local model errors before acting. It identifies misspecified thermal regimes, shrinks their gains, projects the quantum schedule onto a feasible comfort set, and attributes any remaining issues. A sympathetic reader would care because quantum controllers promise efficient representations yet have lacked guarantees against the model inaccuracies common in real heating and cooling systems. The reported experiments on building emulators show the wrapper reduces violations dramatically while preserving invariance to NISQ noise.

Core claim

Q-DASC discovers misspecified operating regimes with false-discovery-rate control, repairs their local thermal gains with shrinkage, projects the proposed quantum schedule onto the repaired comfort-feasible set, and attributes residual violations to policy error, model error, or physical limits. Because the final certificate is produced by classical projection, comfort feasibility is invariant to finite-shot and depolarizing read-out noise. On real BOPTEST building emulators across three buildings, two localized misspecifications, and three seeds, Q-DASC reduces average comfort violation from 26.0% for the raw VQC controller and 55.3% for a model-trusting scheduler to 0.02%, matching a clair

What carries the argument

The certified classical safety layer that discovers misspecified regimes with FDR control, repairs local thermal gains with shrinkage, and projects quantum schedules onto comfort-feasible sets.

Load-bearing premise

The classical safety layer can reliably discover misspecified regimes with false-discovery-rate control and produce feasible projections after local gain repairs.

What would settle it

If the safety layer is applied to a new building emulator with an undetected local misspecification and comfort violations rise well above the oracle level, the invariance claim would fail.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Comfort violation drops to 0.02 percent on BOPTEST emulators across multiple buildings and misspecifications while matching oracle performance.
  • Violation stays at 0.24 percent even when NISQ readout noise is added.
  • The wrapper transfers directly to EnergyPlus heating and cooling benchmarks and to real hospital air-handling-unit data.
  • A repair-aware variant of the quantum policy reaches zero violation and reduces how often the projection step intervenes.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same wrapper structure could be tested on quantum policies for other physics-constrained tasks such as robotic manipulation or grid frequency control.
  • Partial local repairs may prove sufficient for safety in many domains, reducing the need for globally accurate models before deployment.
  • Observational data from real buildings could be used to further tune the shrinkage step without requiring additional simulator runs.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 2 minor

Summary. The manuscript proposes Q-DASC, a wrapper around variational quantum circuit (VQC) policies for HVAC control. The wrapper uses false-discovery-rate (FDR) controlled discovery of misspecified operating regimes, shrinkage repair of local thermal gains, and classical projection of the proposed schedule onto a repaired comfort-feasible set. The central claims are that the resulting certificate makes comfort feasibility invariant to finite-shot and depolarizing readout noise, and that on BOPTEST emulators across three buildings, two localized misspecifications, and three seeds the method reduces average comfort violation from 26.0% (raw VQC) and 55.3% (model-trusting scheduler) to 0.02% (matching an oracle) while remaining at 0.24% under NISQ noise. A repair-aware VQC variant reaches 0.00% violation; the approach is also shown to transfer to EnergyPlus benchmarks and real hospital AHU data.

Significance. If the safety-layer guarantees hold beyond the two tested misspecifications, the work would provide a concrete route to certified deployment of quantum policies in physics-constrained building control, where model error is routine. The reported near-oracle performance, explicit noise invariance via classical projection, and cross-benchmark transfer are strengths that would advance the intersection of variational quantum RL and safety-critical control.

major comments (2)
  1. [Abstract, §4] Abstract and §4 (empirical evaluation): the claim that comfort feasibility is invariant to quantum readout noise rests on the classical projection always producing a non-empty feasible set after FDR-controlled discovery and shrinkage repair. The reported 0.02% and 0.24% figures are obtained only for the two specific localized misspecifications tested on BOPTEST; no additional experiments or analysis are provided for misspecifications lying outside the local-gain family, leaving the general invariance claim load-bearing but incompletely supported.
  2. [§3.2] §3.2 (safety layer): the FDR control and shrinkage procedure are described as guaranteeing discovery and repair of misspecified regimes, yet the manuscript supplies no explicit statement or proof that the resulting feasible set is guaranteed to be non-empty for every VQC-proposed schedule. The empirical success on the tested cases does not substitute for this guarantee when the central safety claim is that the certificate is produced by classical projection.
minor comments (2)
  1. [§3] Notation for the shrinkage estimator and the projection operator should be introduced once with a clear equation reference rather than inline in multiple places.
  2. [Table 2] Table 2 (BOPTEST results) would benefit from an additional column reporting the fraction of time steps in which the projection operator was active, to quantify how often the safety layer intervenes.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on the safety claims. We address each major point below and will revise the manuscript to make the scope and assumptions of the guarantees more explicit.

read point-by-point responses

  1. Referee: [Abstract, §4] Abstract and §4 (empirical evaluation): the claim that comfort feasibility is invariant to quantum readout noise rests on the classical projection always producing a non-empty feasible set after FDR-controlled discovery and shrinkage repair. The reported 0.02% and 0.24% figures are obtained only for the two specific localized misspecifications tested on BOPTEST; no additional experiments or analysis are provided for misspecifications lying outside the local-gain family, leaving the general invariance claim load-bearing but incompletely supported.

    Authors: We agree that the reported performance is specific to the two localized misspecifications within the local-gain family on BOPTEST. The invariance to finite-shot and depolarizing readout noise follows directly from the fact that the final certified action is produced by a purely classical projection step whose output does not depend on quantum measurements. The non-emptiness of the feasible set after FDR-controlled discovery and shrinkage repair is observed empirically for the tested misspecifications. We will revise the abstract and §4 to qualify the invariance claim as holding for the local-gain misspecification class and the evaluated regimes, rather than presenting it as a general guarantee across all possible misspecifications. This removes the load-bearing aspect of an unsupported general claim while preserving the core technical point about classical certification. revision: yes

  2. Referee: [§3.2] §3.2 (safety layer): the FDR control and shrinkage procedure are described as guaranteeing discovery and repair of misspecified regimes, yet the manuscript supplies no explicit statement or proof that the resulting feasible set is guaranteed to be non-empty for every VQC-proposed schedule. The empirical success on the tested cases does not substitute for this guarantee when the central safety claim is that the certificate is produced by classical projection.

    Authors: We acknowledge that §3.2 currently lacks an explicit statement on non-emptiness. The FDR procedure identifies candidate misspecified regimes, after which shrinkage repair adjusts the local thermal gains within bounded intervals chosen to restore satisfaction of the comfort constraints; the subsequent projection is then performed on this repaired model. In all reported experiments the resulting set is non-empty. We will add a clarifying paragraph in §3.2 stating that, under the local-gain misspecification model and the chosen FDR threshold, the shrinkage bounds are constructed so that the repaired constraint set remains non-empty for the VQC schedules encountered in the evaluated regimes, thereby making the classical projection produce a certified feasible action independent of quantum noise. A general proof that this holds for every conceivable VQC output distribution would require additional analysis of the policy class and is beyond the present scope; the revision will make this modeling assumption explicit. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical performance claims rest on external benchmarks, not self-referential definitions or fitted inputs

full rationale

The paper reports observed comfort-violation percentages (0.02 % matching oracle, 0.24 % under noise) on BOPTEST emulators across three buildings, two misspecifications, and three seeds. These numbers are presented as direct experimental outcomes compared against raw VQC, model-trusting scheduler, and clairvoyant oracle baselines. The invariance statement (“comfort feasibility is invariant to finite-shot and depolarizing read-out noise” because “the final certificate is produced by classical projection”) is a logical separation between classical and quantum components, not a definitional equivalence or a fitted parameter renamed as prediction. No equations, self-citations, or uniqueness theorems are shown that would reduce the reported results to quantities defined inside the paper itself. The method description (FDR-controlled discovery + shrinkage repair + projection) is treated as an engineering wrapper whose success is validated externally rather than assumed by construction. This satisfies the default expectation of a non-circular empirical contribution.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract supplies no information on free parameters, background axioms, or newly postulated entities; assessment is therefore limited to surface claims.

pith-pipeline@v0.9.1-grok · 5834 in / 1242 out tokens · 38164 ms · 2026-06-30T09:04:43.044023+00:00 · methodology

0 comments
read the original abstract

Variational quantum reinforcement learning offers a compact policy class for building-energy control, but it inherits a deployment weakness shared by learned controllers: when the thermal model is locally wrong, a policy that appears safe on the model can violate occupant comfort in the real building. Guarantees that depend on noisy quantum read-out are also insufficient for safety-critical control. We address this gap with Q-DASC, Discrepancy-Attributed Safe Quantum Control. Q-DASC wraps a variational-quantum-circuit (VQC) policy with a certified classical safety layer that discovers misspecified operating regimes with false-discovery-rate control, repairs their local thermal gains with shrinkage, projects the proposed quantum schedule onto the repaired comfort-feasible set, and attributes residual violations to policy error, model error, or physical limits. Because the final certificate is produced by classical projection, comfort feasibility is invariant to finite-shot and depolarizing read-out noise. On real BOPTEST building emulators across three buildings, two localized misspecifications, and three seeds, Q-DASC reduces average comfort violation from 26.0\% for the raw VQC controller and 55.3\% for a model-trusting scheduler to 0.02\%, matching a clairvoyant oracle, and remains at 0.24\% under NISQ read-out noise. A repair-aware VQC variant reaches 0.00\% violation and reduces projection intervention, while the default Q-DASC keeps lower energy and stronger observational-data behavior. The same wrapper transfers to EnergyPlus heating and cooling benchmarks and to real hospital air-handling-unit data. These results establish a safety-efficiency frontier for deploying quantum policies in physics-constrained control.

Figures

Figures reproduced from arXiv: 2606.28834 by Yifan Wang.

Figure 1
Figure 1. Figure 1: Overview of Q-DASC. The comfort guarantee is [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: State-of-the-art deployment on real BOPTEST. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: (a) Data efficiency: Q-DASC reaches near-oracle comfort with very few training episodes. (b) NISQ robust￾ness: Q-DASC is invariant to depolarizing read-out noise because its guarantee is classical and model-certified, while the raw quantum controller remains unsafe. Proposition 3. Q-DASC is insensitive to the FDR level q ∈ {0.05, 0.10, 0.20} and needs only a 3×2 regime grid, and reaches low violation with … view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

25 extracted references · 25 canonical work pages

  1. [1]

    Building Optimization Testing Framework (

    Blum, David and Arroyo, Javier and Huang, Sen and Drgo. Building Optimization Testing Framework (. Journal of Building Performance Simulation , volume=. 2021 , publisher=

    work page 2021

  2. [2]

    and Lawrie, Linda K

    Crawley, Drury B. and Lawrie, Linda K. and Winkelmann, Frederick C. and Buhl, W. F. and Huang, Y. Joe and Pedersen, Curtis O. and Strand, Richard K. and Liesen, Richard J. and Fisher, Daniel E. and Witte, Michael J. and Glazer, Jason , journal=. 2001 , publisher=

    work page 2001

  3. [3]

    Annual Reviews in Control , volume=

    All You Need to Know about Model Predictive Control for Buildings , author=. Annual Reviews in Control , volume=. 2020 , publisher=

    work page 2020

  4. [4]

    Control Engineering Practice , volume=

    Data-Driven Methods for Building Control: A Review and Promising Future Directions , author=. Control Engineering Practice , volume=. 2020 , publisher=

    work page 2020

  5. [5]

    IEEE Access , volume=

    Variational Quantum Circuits for Deep Reinforcement Learning , author=. IEEE Access , volume=. 2020 , doi=

    work page 2020

  6. [6]

    Quantum Agents in the

    Skolik, Andrea and Jerbi, Sofiene and Dunjko, Vedran , journal=. Quantum Agents in the. 2022 , doi=

    work page 2022

  7. [7]

    Advances in Neural Information Processing Systems (NeurIPS) , volume=

    Parametrized Quantum Policies for Reinforcement Learning , author=. Advances in Neural Information Processing Systems (NeurIPS) , volume=. 2021 , url=

    work page 2021

  8. [8]

    Physical Review A , volume=

    Quantum Circuit Learning , author=. Physical Review A , volume=. 2018 , doi=

    work page 2018

  9. [9]

    Physical Review A , volume=

    Effect of Data Encoding on the Expressive Power of Variational Quantum-Machine-Learning Models , author=. Physical Review A , volume=. 2021 , doi=

    work page 2021

  10. [10]

    Quantum , volume=

    Data Re-Uploading for a Universal Quantum Classifier , author=. Quantum , volume=. 2020 , doi=

    work page 2020

  11. [11]

    Quantum Computing in the

    Preskill, John , journal=. Quantum Computing in the. 2018 , doi=

    work page 2018

  12. [12]

    Nature Reviews Physics , volume=

    Variational Quantum Algorithms , author=. Nature Reviews Physics , volume=. 2021 , publisher=

    work page 2021

  13. [13]

    Energy , volume=

    Quantum Computing for Energy Systems Optimization: Challenges and Opportunities , author=. Energy , volume=. 2019 , publisher=

    work page 2019

  14. [14]

    Journal of Machine Learning Research , volume=

    A Comprehensive Survey on Safe Reinforcement Learning , author=. Journal of Machine Learning Research , volume=. 2015 , url=

    work page 2015

  15. [15]

    18th European Control Conference (ECC) , pages=

    Control Barrier Functions: Theory and Applications , author=. 18th European Control Conference (ECC) , pages=. 2019 , doi=

    work page 2019

  16. [16]

    Proceedings of the AAAI Conference on Artificial Intelligence (AAAI) , volume=

    End-to-End Safe Reinforcement Learning through Barrier Functions for Safety-Critical Continuous Control Tasks , author=. Proceedings of the AAAI Conference on Artificial Intelligence (AAAI) , volume=. 2019 , doi=

    work page 2019

  17. [17]

    Automatica , volume=

    A Predictive Safety Filter for Learning-Based Control of Constrained Nonlinear Dynamical Systems , author=. Automatica , volume=. 2021 , publisher=

    work page 2021

  18. [18]

    Journal of the Royal Statistical Society: Series B (Methodological) , volume=

    Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing , author=. Journal of the Royal Statistical Society: Series B (Methodological) , volume=. 1995 , publisher=

    work page 1995

  19. [19]

    The Annals of Statistics , volume=

    Estimation of the Mean of a Multivariate Normal Distribution , author=. The Annals of Statistics , volume=. 1981 , doi=

    work page 1981

  20. [20]

    Proceedings of the Fourth Berkeley Symposium on Mathematical Statistics and Probability , volume=

    Estimation with Quadratic Loss , author=. Proceedings of the Fourth Berkeley Symposium on Mathematical Statistics and Probability , volume=. 1961 , url=

    work page 1961

  21. [21]

    Nature Reviews Physics , volume=

    Physics-Informed Machine Learning , author=. Nature Reviews Physics , volume=. 2021 , publisher=

    work page 2021

  22. [22]

    Maddalena, Emilio T. and M. Experimental Data-Driven Model Predictive Control of a Hospital. Energy and Buildings , volume=. 2022 , publisher=

    work page 2022

  23. [23]

    National Solar Radiation Data Base (

    Wilcox, Stephen and Marion, William , howpublished=. National Solar Radiation Data Base (. 2008 , note=

    work page 2008

  24. [24]

    and others , journal=

    Virtanen, Pauli and Gommers, Ralf and Oliphant, Travis E. and others , journal=. 2020 , doi=

    work page 2020

  25. [25]

    and Millman, K

    Harris, Charles R. and Millman, K. Jarrod and van der Walt, St. Array Programming with. Nature , volume=. 2020 , doi=

    work page 2020