{"id":"c662395e-2ea6-4f0b-9d46-653d91cf45ca","arxiv_id":"2501.09339","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":8.0,"correctness_risk":"low","formal_verification":"none","parameter_count":2,"one_line_summary":"Every POVM on C^d becomes projectively simulable after depolarizing with dimension-independent visibility c = 0.02, and can be simulated with postselection probability 1/8 using only a single auxiliary qubit.","lead":"The paper proves that any quantum measurement on a finite-dimensional system can be simulated by simple projective measurements after adding a fixed amount of noise, with the noise level independent of the system size. It also shows a constant-probability simulation using only one helper qubit, which caps the practical advantage of general measurements across many quantum information tasks.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: the non-constructive Kadison–Singer dependence is a disclosed scope limitation, not a correctness threat.","rationale":"The reader's weakest assumption—non-constructivity inherited from Kadison–Singer—is real but is a scope limitation, not a correctness defect. The paper's Results 1 and 2 are existential, and the authors openly distinguish the existential partition from the efficient random-partition variant in Appendix C. My own walk-through of the proof chain found no fatal gap: Lemma 1's nearly-flat fine-graining is valid once the rational case is read correctly; Theorem 4 follows from MSS by summing the norm bound and tracing the size bound; Lemma 5's subpartition preserves the norm bound and gives the claimed q lower bound; Lemma 7's convex decomposition Φ_τ(N)=τF+(1−τ)C is algebraically consistent, with both F and C projectively simulable; and the final constant product q·t_NP ≥ 0.0204/(1+δ)^2 exceeds 0.02 for sufficiently small δ. The paper also correctly limits its own claims about efficient circuit knitting. No load-bearing objection survives scrutiny, so the ACCEPT verdict should stand unchanged.","tokens_in":30518,"tokens_out":36288,"duration_ms":372104,"concrete_test":"Independently re-derive the algebraic step in Lemma 7: with a_i and b_i as in Eq. (A24), verify that τ F_i + (1−τ) C_i = Φ_τ(N_i) for i∈[l], that the normalization conditions Σ a_i ≤ 1 and Σ b_i ≤ 1 hold for τ ≤ t_N, and that the leftover effect automatically matches Φ_τ(N_{l+1}); this settles the one place where a hidden sign or normalization error could enter.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claims are existence statements: for every POVM M there exists a convex decomposition of Φ_0.02(M) into projective measurements, and a one-qubit-ancilla simulation with postselection q=1/8. The only genuinely load-bearing external input is the Marcus–Spielman–Srivastava partition theorem, which is non-constructive. This limits algorithmic applicability but does not threaten the existence claims, and the paper explicitly discloses this in Section III.G and Appendix C, where an efficient random partition gives q=Θ(1/log d). I re-checked the main chain—Lemma 1's fine-graining, Theorem 4's norm and size bounds, Lemma 5's subpartition step, and Lemma 7's depolarized simulation via the dimension-deficient Naimark theorem—and found no internal inconsistency. Lemma 4's rational-case wording is compressed (α_i is 0 rather than in [1−Δ,1]), but the intended construction of equal parts of size 1/k is valid and the lemma is sound.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves two structural results about generalized quantum measurements. Result 1 states that every POVM on C^d can be simulated, up to a constant postselection probability q = 1/8, by a convex combination of POVMs each implementable with a single auxiliary qubit and classical post-processing. Result 2 states that for every POVM M, the depolarized measurement Φ_{0.02}(M) belongs to SP(d), i.e., it is a convex combination of ancilla-free projective measurements with classical post-processing. The proof proceeds by fine-graining an arbitrary POVM into a nearly flat rank-one POVM (Lemma 1), using the Marcus–Spielman–Srivastava solution of Kadison–Singer to partition the outcomes into small groups with constant success probability (Theorem 4), and then using a new dimension-deficient Naimark theorem to implement the resulting nearly projective POVMs by projective measurements under depolarizing noise (Lemma 3 / Theorem 6). The paper derives applications to state discrimination, shadow tomography, circuit knitting, hidden-variable models for noisy two-qudit states, and joint measurability of noisy POVMs.","tokens_in":30694,"tokens_out":23356,"duration_ms":221204,"significance":"The significance is high: the paper settles in the affirmative a conjecture from the authors' earlier work [28] and establishes that, asymptotically in dimension, general POVMs offer only a constant-factor advantage over projective measurements (with or without a single-qubit ancilla) for the tasks considered. The proof is detailed and the main logical chain is internally consistent; the constants c = 0.02 and q = 1/8 are derived from the construction rather than fitted. The reliance on the non-constructive Kadison–Singer theorem is explicitly disclosed, and Appendix C provides an efficiently generatable random partition achieving q = Θ(1/log d), so the algorithmic limitation is clearly separated from the existential claims. The applications (improved POVM-locality ranges, tight incompatibility robustness up to constant factors, and the circuit-knitting scheme) are nontrivial and correctly derived from the main results.","major_comments":[],"minor_comments":[{"comment":"The displayed bound t_NP ≥ 0.3/(1+δ) is not correct for all δ ∈ (0,1]. Using A_i ≥ 0.47/(1+δ) and the worst case |W| = |W⊥| = d/2 in Lemma 3 gives t_NP ≥ 0.47/(2(1+δ)−0.47), which is below 0.3/(1+δ) for δ ≳ 0.085. Since the final constant c = 0.02 only requires sufficiently small δ, the main result is unaffected, but the bound should be restated with its validity range or replaced by the exact expression.","section":"Section IV, Step 3"},{"comment":"When x_i is rational and k_i is chosen so that k_i x_i is an integer, the remainder α_i in Eq. (A2) equals 0, contradicting the stated range 1−Δ ≤ α_i ≤ 1. The intended construction still works if the zero-size remainder is discarded (or if one instead uses a common-denominator construction of equal parts of size 1/k), but the proof as written is internally inconsistent.","section":"Appendix A.1, Lemma 4"},{"comment":"The statement says Φ_t(N) ∈ SP(d) for t equal to the displayed minimum; it should say 'for all t ≤ ...' to match Lemma 7 and to justify the later use of mixing with Φ_0(N).","section":"Section IV, Lemma 3"},{"comment":"The passage treating C as a free parameter via 'C can be effectively regarded as an unconstrained real parameter' is hand-wavy; a rigorous argument should fix C first and then choose ε sufficiently small and an integer r with rε close to C, using continuity of the bounds in C.","section":"Section V, proof of Lemma 2"},{"comment":"There are several typos, including a missing period after 'Let ρ be a N-qubit state' and duplicated words ('for for', 'by by'); also 'transforming classical states on N+1 bits into states on 2N bits' should read 'transforming probability distributions on N+1 bits into probability distributions on 2N bits'.","section":"Section III.D, Proposition 3"}],"recommendation":"minor_revision","confidential_remarks":"To the editor: the paper is technically sound and the central claims are significant. The non-constructive dependence on Kadison–Singer is a genuine limitation for algorithmic applications, but it is clearly disclosed and does not affect the validity of the existence theorems. The requested revision is minor polish; I do not see any need for additional external validation. The paper should be acceptable after the minor corrections."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper delivers what the title promises: for every POVM on C^d, a single-qubit-ancilla simulation with postselection probability q=1/8, and a depolarized version with constant visibility c=0.02 that is projectively simulable without ancillas. Both constants are dimension-independent, and the second result was not something I expected to be true. The proof chain looks sound to me: fine-graining to nearly flat rank-one effects, Kadison-Singer partition, subpartition to groups of size at most d/2, then the dimension-deficient Naimark theorem to convert nearly projective measurements into convex combinations of projective ones under depolarization. I checked the key inequalities and did not find a load-bearing gap. The applications—bounded POVM advantage in state discrimination, shadow tomography, and metrology, plus improved local-model thresholds for two-qudit states and asymptotically tight joint-measurability bounds—follow cleanly once the two main results are accepted.\n\nWhat is genuinely new: the resolution of the conjecture from [28] with a single ancilla qubit, and the new constant-visibility simulability result. The dimension-deficient Naimark theorem is a useful tool in its own right. The paper is also honest about what it does not do: the Kadison-Singer partition is non-constructive, and no efficient algorithm for finding it is provided. Appendix C gives a random partition achieving q = Theta(1/log d), which is efficiently generatable, so the protocol is not purely existential. That is a disclosed limitation, not a hidden flaw.\n\nThe soft spots are minor. The rational-case handling in Lemma 4's proof is compressed and a bit sloppy—alpha_i is effectively 0 rather than in the stated range—but the intended construction works. The bound t_NP >= 0.3/(1+delta) is slightly optimistic for larger delta, but the authors only need delta small, so it does not affect the final constant. The circuit-knitting application is information-theoretic and likely not practically efficient for general circuits; the authors say so explicitly.\n\nThis is a serious paper with rigorous proofs and real consequences. It deserves a careful referee and publication. My own verdict matches the reader's ACCEPT.","headline":"Constant-factor simulation of arbitrary POVMs by projective measurements, resolving a conjecture with q=1/8 and c=0.02; the main caveat is the non-constructive Kadison-Singer dependence.","tokens_in":31272,"tokens_out":1098,"would_cite":true,"duration_ms":13018,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P15","81P45"],"pacs":["03.65.Ta","03.67.-a"],"model":"deepseek-v4-flash","headline":"The paper proves that adding a fixed amount of depolarizing noise, with visibility $c = 0.02$, makes every generalized quantum measurement (POVM) simulable by randomized projective measurements in any finite dimension, with no ancillary…","keywords":["generalized quantum measurements","projective simulation","depolarizing noise","Kadison-Singer problem","Naimark extension","postselection","measurement incompatibility","circuit knitting"],"falsifier":"Compute $t_{\\mathrm{SP}}(M) = \\max\\{t : \\Phi_t(M) \\in \\mathrm{SP}(\\mathbb{C}^d)\\}$ for a fixed sequence of POVMs, for example the symmetric informationally complete POVMs in increasing dimension; the claim predicts a uniform lower bound of $0.02$ for every $d$, so finding any $M_d$ with $t_{\\mathrm{SP}}(M_d) < 0.02$, or a numerical violation of $q = 1/8$ for the single-qubit simulation, would refute Result 2.","tokens_in":30296,"feed_emoji":"⚛️","tokens_out":11406,"duration_ms":102326,"temperature":0.7,"pith_summary":"The paper asks how much the most general quantum measurements (POVMs) can outperform the simple projective measurements of textbook quantum mechanics. It answers that the advantage is bounded by a constant factor: after applying depolarizing noise with a fixed visibility $c = 0.02$, every POVM on a $d$-dimensional system becomes a convex combination of projective measurements, with no ancilla needed. It also shows that every POVM can be simulated with constant postselection probability $q = 1/8$ by measurements that use just one auxiliary qubit. Because these constants do not grow with $d$, general measurements cannot offer a dimension-dependent asymptotic edge in state discrimination, shadow tomography, or other linear tasks; the paper also derives new bounds on local hidden variable models and measurement incompatibility.","feed_headline":"Constant noise makes any quantum measurement projectively simulable","feed_subtitle":"Depolarizing with visibility 0.02 makes arbitrary POVMs a random mixture of projective measurements in every dimension.","key_machinery":"The argument runs through three pieces. (1) A fine-graining lemma converts any POVM into one with rank-one effects of nearly equal magnitude. (2) The solution of the Kadison-Singer problem is invoked as a partition theorem: for any such nearly-flat POVM, the outcome set can be partitioned into blocks of size at most $d/2$ so that the simulation protocol of Theorem 2 succeeds with probability $q \\ge 0.068/(1+\\delta)$; this is the step that makes $q$ independent of $d$. (3) The dimension-deficient Naimark theorem shows that a nearly projective measurement with at most $d/2$ rank-one effects can be twirled, by random phases and unitaries on the orthogonal complement of the span of its effects, into a genuine convex combination of projective measurements, and an auxiliary lemma quantifies the amount of depolarizing noise tolerated by such measurements. Pasting the post-processing through these steps yields the constant $c = 0.02$.","core_discovery":"The paper establishes two dimension-independent simulation theorems. First, every $d$-dimensional POVM can be realized, with postselection probability $q = 1/8$, as a convex combination of measurements that each require only a single ancillary qubit plus classical post-processing. Second, the depolarized measurement $\\Phi_c(M)$ with $c = 0.02$ belongs to $\\mathrm{SP}(\\mathbb{C}^d)$ for every POVM $M$: it is exactly a convex combination of projective measurements on $\\mathbb{C}^d$ itself, needing no ancilla. The constants $q$ and $c$ do not depend on $d$, so the asymptotic power of general measurements over projective ones is bounded by a constant factor in every linear information-processing task considered, and the paper derives consequences for state discrimination, shadow tomography, circuit knitting, local hidden variable models, and joint measurability of noisy POVMs.","pith_inferences":["If the result is correct, any task whose figure of merit is a linear functional of measurement statistics can separate POVMs from projective measurements by at most a constant factor in any dimension, ruling out exponential POVM advantages in principle.","The partition guaranteed by the Kadison-Singer solution is not known to be efficiently findable; the paper shows random partitions achieve $q \\approx \\Theta(1/\\log d)$, so determining whether constant-$q$ partitions can be found in polynomial time would decide whether the circuit-knitting scheme becomes a practical algorithm.","The dimension-deficient technique may extend from measurements to quantum channels, instruments, and combs, suggesting that constant-factor simulation of general processes by low-ancilla devices is a general structural phenomenon rather than a special property of measurements.","The joint-measurability bound is tight in $d$ up to the constant $c$, making it natural to test numerically whether $c$ can be pushed to 1 for specific families; if it can, POVMs would share exactly the same compatibility noise threshold as projective measurements."],"forward_implications":["In minimal-error state discrimination, the best POVM can beat the best projective measurement by at most the factor $1/c = 50$ in success probability, for any ensemble in any dimension.","In classical shadow tomography with traceless observables, a general POVM can reduce the worst-case variance bound by at most the constant factor $1/c^2 = 2500$ compared with a projectively simulable measurement.","Noisy isotropic states of two qudits admit local hidden variable models for all POVMs up to visibility $c\\log(d)/d$, and general noisy pure states up to $c\\log(d)/d^2$, improving over previous results.","All POVMs on $\\mathbb{C}^d$ become jointly measurable after depolarizing with visibility $t \\le c\\log(d)/d$, a bound that matches the known projective-measurement scaling up to the constant $c$.","Sampling from the output of any $2N$-qubit unitary can be emulated by randomization over $N+1$-qubit subcircuits with success probability $q = 1/8$, giving a circuit-knitting scheme with constant probabilistic overhead."],"supporting_citations":[{"why":"It introduces the class $\\mathrm{SP}(d)$ of projectively simulable measurements and gives the baseline bound $t_{\\mathrm{SP}}(d) \\ge 1/d$ that the new result improves.","marker":"[2]"},{"why":"It proves that exact simulation of a general POVM by projective measurements with postselection cannot exceed success probability $q = 1/d$, the limitation that forces the paper to consider noisy simulations.","marker":"[27]"},{"why":"It supplies the probabilistic simulation protocol used as Theorem 2, which simulates a POVM by smaller-outcome POVMs, and conjectures a constant success probability that the paper proves.","marker":"[28]"},{"why":"It provides the solution of the Kadison-Singer conjecture in the form of the partition theorem (Theorem 3) that guarantees constant success probability with block sizes at most $d/2$.","marker":"[38]"},{"why":"It gives the projective-measurement local models for isotropic states whose visibility range the paper extends to general POVMs.","marker":"[29]"},{"why":"It establishes the reference bound on joint measurability of noisy projective measurements that the paper's compatibility result for POVMs matches up to the constant $c$.","marker":"[30]"},{"why":"It supplies the matrix concentration inequality used to show that random partitions achieve success probability $\\Theta(1/\\log d)$ and can be found efficiently.","marker":"[90]"},{"why":"It shows that continuous-outcome POVMs on $\\mathbb{C}^d$ can be reduced to measurements with at most $d^2$ outcomes, letting the simulation results cover general POVMs.","marker":"[49]"}],"fun_headline_variants":["Constant noise tames all POVMs into projective ones","All POVMs become projective after tiny depolarization","Universal projective simulation via constant noise","One constant noise level projectivizes all measurements"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument stands on the existence, asserted by the Kadison-Singer solution, of a partition of the outcomes of any nearly-flat rank-one measurement into blocks of size at most $d/2$ whose simulation success probability stays constant as the dimension grows, and nothing in the paper provides an efficient way to find that partition.","fun_headline_variants_meta":{"raw":{"variants":["Constant noise tames all POVMs into projective ones","All POVMs become projective after tiny depolarization","Universal projective simulation via constant noise","One constant noise level projectivizes all measurements"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00033,"raw_usage":{"total_tokens":1862,"prompt_tokens":990,"completion_tokens":872,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":606,"completion_tokens_details":{"reasoning_tokens":813}},"tokens_in":606,"tokens_out":872,"duration_ms":6152,"temperature":1.0,"reasoning_tokens":813,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T20:08:38.458666+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute $t_{\\mathrm{SP}}(M) = \\max\\{t : \\Phi_t(M) \\in \\mathrm{SP}(\\mathbb{C}^d)\\}$ for a fixed sequence of POVMs, for example the symmetric informationally complete POVMs in increasing dimension; the claim predicts a uniform lower bound of $0.02$ for every $d$, so finding any $M_d$ with $t_{\\mathrm{SP}}(M_d) < 0.02$, or a numerical violation of $q = 1/8$ for the single-qubit simulation, would refute Result 2.","supporting_citations":[{"cited_title":"Mozrzymas, M","cited_arxiv_id":null,"evidence_quote":"It proves that exact simulation of a general POVM by projective measurements with postselection cannot exceed success probability $q = 1/d$, the limitation that forces the paper to consider noisy simulations."},{"cited_title":"Oszmaniec and T","cited_arxiv_id":null,"evidence_quote":"It supplies the probabilistic simulation protocol used as Theorem 2, which simulates a POVM by smaller-outcome POVMs, and conjectures a constant success probability that the paper proves."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It gives the projective-measurement local models for isotropic states whose visibility range the paper extends to general POVMs."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It establishes the reference bound on joint measurability of noisy projective measurements that the paper's compatibility result for POVMs matches up to the constant $c$."},{"cited_title":"Chiribella, G","cited_arxiv_id":null,"evidence_quote":"It shows that continuous-outcome POVMs on $\\mathbb{C}^d$ can be reduced to measurements with at most $d^2$ outcomes, letting the simulation results cover general POVMs."}],"review_version":1}