{"id":"64be8d87-dc76-4f98-862d-edc9795b768d","arxiv_id":"2506.20658","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A framework defining quantum advantage as verifiable plus classically superior, with a conclusion that random circuit sampling is not yet a satisfactory path.","lead":"Quantum advantage, according to this position paper, is only real when a quantum computer produces answers that can be checked and that clearly beat the best classical methods in speed, cost, or accuracy. The authors argue that recent random circuit sampling demonstrations fail this test, and that the most promising early wins will come from error-mitigated chemistry simulations and sampling problems.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The RCS rejection applies a 'rigorous validation' bar that the paper's own preferred error-mitigation path also fails to meet at scale; the asymmetry is asserted, not derived.","rationale":"The reader's weakest_assumption correctly identifies that the rejection of RCS rests on a normative choice: statistical certification is declared insufficient for validation. My stress-test agrees with this but sharpens it into an internal asymmetry claim: the paper's own preferred near-term methods inherit the same epistemic weakness, because their 'rigorous error bars' are conditional on noise models that cannot be verified at scale. This makes the load-bearing assumption more concrete than a pure matter of taste: it is a double standard unless the paper can show why a characterized noise model is easier to certify than a sampled output distribution. I did not find a separate fatal flaw in the definitional framework; as a position paper, its proposal is coherent and its taxonomy is useful. The concrete test would settle whether the asymmetry is real or merely apparent by comparing the two verification burdens under adversarial noise-model mismatch. Since the reader's verdict is already CONDITIONAL and the condition is essentially the need to defend this epistemic standard, my analysis does not move the verdict. I would keep CONDITIONAL and add the symmetry check as an explicit condition to be satisfied before applying the framework to dismiss RCS claims.","tokens_in":13010,"tokens_out":6023,"duration_ms":84704,"concrete_test":"Take a published near-term error-mitigation result with rigorous error bars, e.g., the expectation-value experiment of ref. [8] using PEC or zero-noise extrapolation. Corrupt the assumed noise model with an adversarial but physically plausible term that is invisible to the characterization used, such as two-qubit crosstalk below the tomography resolution or slowly drifting gate parameters. Recompute the mitigated expectation value and its stated confidence interval. If the ideal value exits the reported interval for a noise perturbation within current characterization uncertainty, then the same verification standard used to exclude RCS also excludes error-mitigated expectation values, and the paper needs an explicit argument for the asymmetry. If the bound survives for a realistic adversarial perturbation, the asymmetry is supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's key applied conclusion—that RCS 'does not yet constitute a fully satisfactory pathway to quantum advantage'—depends on an epistemic standard that is applied unevenly across the methods it compares. Section III rejects RCS because XEB-style statistical certification is not 'rigorous validation.' But the near-term alternatives the paper embraces, especially error-mitigated expectation values in Section IV A2, are rigorous only up to an assumed and partially characterized noise model (e.g., sparse Pauli-Lindblad noise, Markovianity, stationarity). The paper itself concedes in Section II A that 'formally proven results always rely on a set of initial assumptions... that must themselves be verified.' At scales beyond classical simulation, those assumptions cannot be fully verified independently; they are supported by the same kind of heuristic evidence—randomized benchmarking, small-system tomography, cross-checks—that the paper deems insufficient for RCS. Conversely, the RCS experiments of refs. [21–23] pass strong statistical tests consistent with the ideal circuit, yet the paper treats those tests as inadmissible because they can in principle be spoofed. A symmetric critique applies to a noise model that can be spoofed by correlated or drifted errors not captured by the characterization. The paper does not supply a principled boundary between 'heuristic but acceptable' and 'heuristic but unacceptable.' Therefore the central conclusion rests on a normative validation hierarchy rather than on a demonstrated technical asymmetry. If the community accepts XEB as validation, the RCS conclusion is reversed; if the paper requires rigorous certification, error-mitigated expectation values at scale also fail unless their noise-model assumptions are certified at scale, which is not shown.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This position paper proposes a functional definition of quantum advantage with two criteria: (i) the correctness of the output can be rigorously validated, and (ii) the computation is performed with a quantum separation that demonstrably offers superior efficiency, cost-effectiveness, or accuracy over classical computation. The authors apply this framework to three algorithmic families: sampling problems, variational/diagonalization methods, and expectation-value estimation. They conclude that random circuit sampling (RCS) does not yet constitute a satisfactory pathway to quantum advantage because its outputs cannot be rigorously validated at scale without fault tolerance, whereas error-mitigated expectation values and quantum diagonalization methods (SQD/SKQD) are more likely to yield early, verifiable advantage. The paper also reviews error correction, error mitigation, error detection, quantum-centric supercomputing, and current hardware platforms.","tokens_in":13248,"tokens_out":6494,"duration_ms":78177,"significance":"If accepted, the proposed definition would provide a much-needed common vocabulary for evaluating near-term quantum advantage claims, and the paper's emphasis on verifiability, falsifiability, and open benchmarking is a constructive contribution to the field. The manuscript is clearly written and well-referenced, and it gives a detailed, honest discussion of error detection as an intermediate path between error mitigation and fault tolerance. The paper's concrete roadmap—prioritizing error-mitigated expectation values and diagonalization-based methods over sampling—is a useful hypothesis that can stimulate further research. However, the paper is a perspective rather than a proof-based contribution, and its central applied conclusion about RCS rests on a normative epistemic standard that is applied asymmetrically to the methods it favors. The definition itself also lacks an operational specification of what counts as 'demonstrably superior.' These issues do not destroy the paper's value but do require substantial clarification before the conclusions can be considered fully supported.","major_comments":[{"comment":"The validation standard is applied unevenly. Section III rejects RCS because XEB-style statistical certification is treated as insufficient, asserting that 'the only universally accepted method for achieving this is fault-tolerant quantum computing.' Yet Section IV A2 claims that several quantum error mitigation methods 'have demonstrated the ability to yield accurate expectation values from short-depth circuits, with rigorous error bounds [60,61].' Those rigorous error bounds are conditional on noise-model assumptions (e.g., sparse Pauli-Lindblad models) that are themselves verified only through heuristic evidence such as randomized benchmarking and small-system tomography. Section II A concedes that 'formally proven results always rely on a set of initial assumptions... that must themselves be verified.' At scales beyond classical simulation, those noise-model assumptions cannot be fully verified independently, leaving a symmetric vulnerability to unmodeled errors. The paper does not supply a principled boundary between heuristic validation that is admissible and heuristic validation that is not; this asymmetry is load-bearing because it drives the central conclusion that RCS is not a satisfactory pathway while error-mitigated expectation values are.","section":"Section III vs. Section IV A2"},{"comment":"The assertion that fault-tolerant quantum computing is the 'only universally accepted method' for certifying error-free sampling is a contestable empirical claim about community consensus, not a technical result. The paper does not engage with the substantial literature on verification of random circuit sampling, including linear cross-entropy benchmarking and its known limitations, nor does it explain why statistical evidence of sampling correctness is categorically inadmissible while the statistical evidence supporting noise-model accuracy is admissible. Because the RCS conclusion depends entirely on this premise, the authors should either justify the consensus claim with evidence or reframe it explicitly as a normative choice rather than a universal standard.","section":"Section III, RCS paragraph"},{"comment":"The second criterion—'quantum separation that demonstrably offers superior efficiency, cost-effectiveness, or accuracy'—is not operational as stated. The paper does not specify the baseline (e.g., best known classical algorithm at the time of the claim, including future algorithmic improvements), the metric (wall-clock time, financial cost, energy, accuracy measure), or the required confidence level. The abstract calls the definition 'empirically verifiable,' but without these specifications it is difficult to falsify any particular claim of advantage. This weakens the central contribution of the paper, which purports to provide a functional framework for evaluating advantage claims.","section":"Section II, Definition of quantum advantage"}],"minor_comments":[{"comment":"The paper uses 'quantum advantage' without clarifying its relationship to the earlier term 'quantum supremacy'; a sentence distinguishing the two would help readers.","section":"Section I"},{"comment":"The paragraph on peaked random circuits first presents peakedness as enabling verifiable advantage and then notes that peaked distributions may be simulable in quasi-polynomial time [29]; the presentation would be clearer if this tension were addressed head-on rather than leaving the reader to reconcile the two statements.","section":"Section III, peaked random circuits"},{"comment":"The PEC sampling overhead expression '~ (1+15ε/8)^{nd}' should define n, d, and ε explicitly and state the noise model to which ε refers.","section":"Section IV A2"},{"comment":"The term 'quantum-centric supercomputing (QCSC)' is used as a proprietary label; consider defining it in more neutral language so the framework is accessible to a broad community.","section":"Section IV B"},{"comment":"The prediction that credible evidence of quantum advantage will emerge 'within the next two years' is speculative and lacks supporting analysis; either cite a roadmap study or soften the claim.","section":"Section V"},{"comment":"Several statements are phrased as opinions ('we believe,' 'we anticipate') mixed with technical assertions; marking the distinction would improve clarity in a position paper.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"All authors are affiliated with IBM Quantum or PASQAL, and the manuscript promotes technologies in which those companies have commercial interests, including Qiskit, SQD, and neutral-atom platforms. The proposed 'consensus-driven' definition should therefore not be treated as an independent community consensus without external scrutiny. The asymmetry between the validation standard applied to RCS and that applied to error mitigation is, in my reading, a real and central problem; the authors should be asked to address it explicitly, either by providing a principled epistemic criterion or by presenting the RCS conclusion as a clearly stated normative choice. This is fixable within the scope of a perspective article, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know two things about arXiv:2506.20658. First, it is not a technical paper; it is an operational position statement on what should count as quantum advantage, co-authored by senior people at IBM Quantum and Pasqal. Second, the definition it proposes—rigorous validation plus demonstrable quantum separation—is coherent and useful as an organizing framework, but the paper's punchline that RCS 'does not yet constitute a fully satisfactory pathway' rests on a validation standard that is applied more strictly to sampling than to the error-mitigated methods the authors favor.\n\nWhat is genuinely new here is the synthesis. The two-criteria definition collects ideas that exist in pieces (Aaronson and Chen on verification, Preskill's NISQ perspective), but the paper turns them into a single yardstick and then applies it across three problem families: sampling, variational, and expectation values. The discussion of validation modes—rigorous error bars, efficient classical verification, and variational scoring—is clear and worth stealing for classroom use. The paper also makes a fair point that any quantum advantage claim should be framed as a falsifiable hypothesis, and that classical improvements overtaking a claim should not be treated as a failure.\n\nThe soft spots are real. The stress-test concern lands: error-mitigated expectation values at scale are rigorous only up to an assumed noise model (e.g., sparse Pauli-Lindblad, Markovianity), and those assumptions are supported by the same kind of heuristic evidence—randomized benchmarking, small-system tomography—that the paper dismisses as inadmissible for RCS. The paper concedes in Section II A that formally proven results rely on assumptions that must themselves be verified, but it never gives a principled boundary between acceptable heuristic evidence and unacceptable heuristic evidence. The result is that the central applied conclusion is a normative hierarchy, not a demonstrated technical asymmetry. The two-year prediction in the conclusion is unsupported by anything in the framework. And there is an unavoidable promotional thread: the infrastructure sections read like product announcements for IBM and Pasqal platforms.\n\nStill, the paper is honest about the limits of heuristics, cites the relevant literature well, and is carefully argued. The definition, even if qualitative, is a step toward a consensus vocabulary. I would send it to peer review; a good referee could push the authors to either soften the RCS conclusion or supply a principled reason why noise-model certification at scale is more tractable than statistical certification of sampling outputs. The paper deserves engagement, not dismissal.","headline":"A coherent, useful position paper on quantum advantage whose dismissal of RCS depends on a validation standard applied unevenly; worth engaging, but the asymmetry needs confrontation.","tokens_in":13882,"tokens_out":2035,"would_cite":true,"duration_ms":21707,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P68","68Q12"],"pacs":["03.67.Lx"],"model":"deepseek-v4-flash","headline":"The paper argues that quantum advantage requires both rigorous validation of outputs and a demonstrable separation from classical computation, and that random circuit sampling does not yet meet this bar.","keywords":["quantum advantage","random circuit sampling","error mitigation","error detection","validation","variational quantum eigensolver","quantum diagonalization","quantum-centric supercomputing"],"falsifier":"A random-circuit-sampling experiment whose outputs are certified by a method the community accepts as rigorous—for example, fault-tolerant or post-selected error-detected sampling—and whose distribution is verified classically at a scale beyond classical simulation would falsify the claim that RCS is not a satisfactory pathway to quantum advantage.","tokens_in":12823,"feed_emoji":"⚛️","tokens_out":7578,"duration_ms":79869,"temperature":0.7,"pith_summary":"Quantum advantage, the paper argues, should not be treated as a buzzword but as an operational claim with two mandatory criteria: the output must be rigorously validated, and the quantum computation must show a demonstrable separation from classical alternatives in efficiency, cost, or accuracy. Applying this standard, the paper concludes that random circuit sampling—the basis of several high-profile supremacy demonstrations—does not yet meet the bar, because certifying its outputs at scale is classically intractable without fault tolerance. The constructive side of the framework identifies where early advantage is more plausible: problems whose answers are classically checkable, such as ground-state energies from sample-based quantum diagonalization, or expectation values produced by error mitigation with proven error bounds. The practical upshot is a shift in near-term strategy away from sampling supremacy and toward verifiable hybrid quantum-classical computation in high-performance computing centers.","feed_headline":"Sampling supremacy fails the verification test for quantum advantage","feed_subtitle":"Two-part definition—rigorous validation plus demonstrable separation—points to error-mitigated and diagonalization-based methods.","key_machinery":"The load-bearing object is a two-criterion definition of quantum advantage, coupled with a taxonomy of validation modes. Validation, the first criterion, can be achieved in three ways: rigorous error bars (from fault-tolerant computation, formally proven error mitigation, or post-selected error detection); efficient classical verification of the answer's structure (as in factoring or peaked sampling); or variational scoring, where approximate solutions can be ranked by energy or cost without knowing the exact answer. The second criterion, quantum separation, requires the quantum result to be demonstrably better than the best available classical approach, measured by efficiency, cost, or accuracy. This definitional machinery does the work of classifying algorithms: it elevates sample-based quantum diagonalization and error-mitigated expectation values as verifiable, and demotes random circuit sampling as unverifiable at scale.","core_discovery":"On the paper's own terms, the central discovery is a definition plus a verdict. Quantum advantage is defined as the execution of an information-processing task on quantum hardware that satisfies two criteria: (i) the correctness of the output can be rigorously validated, and (ii) the computation is performed with a quantum separation that demonstrably offers superior efficiency, cost-effectiveness, or accuracy over classical computation alone. The paper then applies this definition to three algorithmic families—sampling, variational ground-state problems, and expectation values of observables—and concludes that random circuit sampling does not yet constitute a fully satisfactory pathway to quantum advantage, because the only universally accepted way to certify that RCS outputs are drawn faithfully at scale is fault-tolerant quantum computing. Experimental supremacy claims based on RCS therefore remain unsubstantiated under this criterion. In contrast, sample-based quantum diagonalization and error-mitigated expectation values with provable error bounds achieve the highest degree of verifiability, because their outputs can be classically ranked and reproduced, making them the most credible candidates for early advantage.","pith_inferences":["Editorial extension: the paper's definition makes community acceptance of statistical certification, such as cross-entropy benchmarking, the decisive judgment; if that judgment flips, so does the verdict on RCS.","Editorial extension: the 'quantum separation' criterion compares against best-known classical algorithms and hardware-specific metrics, so in practice the framework yields sequential, falsifiable benchmarks rather than unconditional separations.","Editorial extension: the constructive path depends on the paper's own caveat that error mitigation has exponential sampling overhead and analog verification is hard; if hardware fidelity plateaus, the expectation-value route weakens.","Editorial extension: a direct testable next step is identifying which local observables in analog simulators keep size-independent error bounds, using the cited robustness results as a map."],"forward_implications":["Random-circuit-sampling claims will not count as quantum advantage under this standard until sampling is certified by fault tolerance, error detection with post-selection, or an equally rigorous method.","Early advantage claims will most plausibly come from classically verifiable ground-state problems, such as sample-based quantum diagonalization, where the final answer is stored and checked classically.","Error mitigation with proven error bounds, augmented by classical tensor-network and light-cone methods, extends the reach of expectation-value computations beyond brute-force classical simulation.","The benchmark for advantage shifts from quantum-hardware-versus-classical to hybrid quantum-classical systems integrated into HPC, so advantage becomes a property of the combined workflow.","Peaked random circuits are the remaining open avenue for sampling-based advantage, pending a rigorous hardness analysis; quasi-polynomial classical simulation of peakedness threatens them."],"supporting_citations":[{"why":"Introduces random circuit sampling as a candidate for quantum supremacy based on average-case hardness; the paper's central negative verdict targets this candidate.","marker":"[16]"},{"why":"The first high-profile RCS supremacy demonstration; the paper lists it among claims that remain unsubstantiated without rigorous validation.","marker":"[21]"},{"why":"A later RCS-based experimental claim the paper likewise treats as failing the validation criterion.","marker":"[22]"},{"why":"Argues that classical algorithms can match or challenge RCS outputs, precluding the fair comparison the framework requires.","marker":"[28]"},{"why":"Proposes peaked random circuits whose output structure allows classical verification, the paper's main open avenue for sampling-based advantage.","marker":"[10]"},{"why":"Shows peakedness may be simulable in quasi-polynomial time, weakening the hardness case for peaked RCS.","marker":"[29]"},{"why":"Demonstrates sample-based quantum diagonalization on iron-sulfur clusters beyond exact diagonalization; supports the claim that classically certifiable ground-state energies are achievable now.","marker":"[9]"},{"why":"Runs error-mitigated expectation-value computations at scales beyond brute-force classical simulation, underpinning the expectation-value path to advantage.","marker":"[8]"},{"why":"Provides probabilistic error cancellation with rigorous error bounds, the formal basis for treating error-mitigated expectation values as validated.","marker":"[61]"},{"why":"Demonstrates low-overhead error detection in large graph states, supporting post-selected error detection as a middle path to certified samples.","marker":"[25]"}],"fun_headline_variants":["Sampling supremacy fails new quantum advantage test","Quantum advantage defined: verifiable output plus real separation","Error-mitigated and diagonalization methods lead in verifiability","Verifiability key to credible quantum advantage"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The paper's verdict depends on the normative judgment that statistical certification of sampling outputs is not validation; if the community instead accepts cross-entropy benchmarking or similar statistical evidence as sufficient, the conclusion that RCS claims are unsubstantiated collapses.","fun_headline_variants_meta":{"raw":{"variants":["Sampling supremacy fails new quantum advantage test","Quantum advantage defined: verifiable output plus real separation","Error-mitigated and diagonalization methods lead in verifiability","Verifiability key to credible quantum advantage"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001284,"raw_usage":{"total_tokens":5197,"prompt_tokens":847,"completion_tokens":4350,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":463,"completion_tokens_details":{"reasoning_tokens":4288}},"tokens_in":463,"tokens_out":4350,"duration_ms":30666,"temperature":1.0,"reasoning_tokens":4288,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T22:42:50.815948+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A random-circuit-sampling experiment whose outputs are certified by a method the community accepts as rigorous—for example, fault-tolerant or post-selected error-detected sampling—and whose distribution is verified classically at a scale beyond classical simulation would falsify the claim that RCS is not a satisfactory pathway to quantum advantage.","supporting_citations":[{"cited_title":"Bravyi, D","cited_arxiv_id":null,"evidence_quote":"The first high-profile RCS supremacy demonstration; the paper lists it among claims that remain unsubstantiated without rigorous validation."},{"cited_title":"Villalonga, X","cited_arxiv_id":null,"evidence_quote":"Argues that classical algorithms can match or challenge RCS outputs, precluding the fair comparison the framework requires."},{"cited_title":"DeCross, R","cited_arxiv_id":null,"evidence_quote":"Shows peakedness may be simulable in quasi-polynomial time, weakening the hardness case for peaked RCS."}],"review_version":1}