{"id":"0ae5c3f0-555c-428e-8457-570dc5f15935","arxiv_id":"2504.17621","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"N-fold parallel repetition of routed Bell tests lowers the detection-efficiency threshold for certifying non-jointly-measurable measurements on the distant device to eta*=1/2^N, matching the tight lower bound.","lead":"Long-distance quantum cryptography fails when photons are lost before being detected. By repeating a routed Bell test N times in parallel, this paper proves the required detection efficiency of the distant detector drops to 1/2^N, the best possible value for N repeated copies.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 2's key bound (B17) is false for a valid pair of operators, so the rCHSH_N proof is invalid as written.","rationale":"The reader's weakest assumption was the perfectness of A/B0 and the lack of a concrete robust self-testing function. That is a valid concern about the advertised robustness. However, I found a more immediate, internal issue that the reader did not flag: the analytical proof of Proposition 2, which is one half of the central exponential-threshold claim (rCHSH_N), relies on a numerically false bound. Because the central claim explicitly includes rCHSH_N for all N, an invalid proof step is load-bearing even in the ideal case. The rest of the ideal-case argument for rBB84_N (Proposition 1) appears coherent, and the rCHSH_N theorem is likely repairable by replacing the constant β with the true maximum (~0.683), which still lies below α and preserves the q-range. But as written, the proof of Proposition 2 does not go through, and the exhaustive search covers only N ≤ 3. I therefore retain the reader's CONDITIONAL verdict, but for a different reason: the paper should be accepted only after the bound in Eq. (B17) is corrected and the argument re-verified, or the rCHSH_N claim is removed or moderated to N ≤ 3. This is not an ad hominem or a disagreement with accepted consensus; it is a concrete internal error in a central proof.","tokens_in":22156,"tokens_out":38116,"duration_ms":340201,"concrete_test":"Compute the operator norm of √A√B for A = (|0><0|+|+><+|)/2 and B = (|0><0|+|−><−|)/2. Its square is the largest eigenvalue of A^{1/2} B A^{1/2}, equal to (2+√3)/8, giving ||√A√B|| ≈ 0.6830, which exceeds β ≈ 0.6658. Then rerun the proof of Proposition 2 with the corrected maximum β_true = max_{π'≠π''} ||√π'√π''|| over the four operators, and verify that the q-range [α^{N−1}β_true, α^N) is nonempty and still yields the JM bound (α^N−q)/2^N; if it does, the result is repairable, but the paper must state the corrected bound explicitly.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The proof of Proposition 2 (rCHSH_N) reduces the JM upper bound to bounding ||Σ S_l|| via the Popovici–Sebestyén bound. In Appendix B, Eq. (B17) asserts that for any two distinct operators π' in the set {1/2(|0><0|+|+><+|), 1/2(|0><0|+|−><−|), 1/2(|1><1|+|+><+|), 1/2(|1><1|+|−><−|)}, ||√π'√π''|| ≤ β = (2+√2+√(4√2−2))/8 ≈ 0.6658. This is false. Take π' = (|0><0|+|+><+|)/2 and π'' = (|0><0|+|−><−|)/2, both in the allowed set. A direct calculation gives ||√π'√π''||² = λ_max(π'^{1/2}π''π'^{1/2}) = (2+√3)/8, hence ||√π'√π''|| ≈ 0.6830 > β. Therefore the elementwise inequality Γ ≤ G_k used after Eq. (B18) does not hold, and the subsequent norm bound ||Γ|| ≤ ||G_k|| is not justified. Since this step supplies the q-range and the JM bound in Proposition 2, the analytic proof of the rCHSH_N result is incomplete as written. The threshold may survive with a corrected constant, but that corrected proof is not in the paper; the exhaustive search covers only N ≤ 3 and is conjectural beyond.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies N-fold parallel repetitions of two routed Bell strategies, rBB84_N and rCHSH_N, and claims that when the near-source devices A and B0 maximally violate the N-product CHSH inequality, non-jointly-measurable (NJM) measurements in the distant device B1 can be certified for every detection efficiency η > 1/2^N. The threshold is argued to be tight because B1 has 2^N settings. The rBB84_N result is proven in Proposition 1 via a penalized BB84 inequality and a Popovici–Sebestyén bound on sums of tensor-product projectors. The rCHSH_N result is stated as Proposition 2 with a proof in Appendix B. The paper also presents numerical NPO results for N=2 showing improved robustness against source visibility, and a conditional analytic robustness statement in Proposition 3 based on an assumed robust self-testing bound. The central ideal-case result, if correct, would give an optimal exponential reduction of the critical detection efficiency for a DIQKD-relevant task.","tokens_in":22402,"tokens_out":22153,"duration_ms":212178,"significance":"If the claims are established, the paper represents a significant step for device-independent quantum cryptography: it constructs explicit routed Bell strategies for which the critical detection efficiency for certifying NJM measurements decreases as 2^{-N}, which is optimal given the number of settings of the distant device. The rBB84_N part is proven analytically with a clean argument, and the paper provides reproducible code and numerical evidence for the N=2 robustness comparison. The main caveats are that the analytic proof of the rCHSH_N result currently relies on a false bound in Appendix B, and the 'robust' claim in the abstract is not backed by an explicit robust self-testing function. These issues are local and likely repairable, but they need to be addressed before the full claims can be accepted.","major_comments":[{"comment":"Eq. (B17) claims that for any two distinct operators π' in the set {½(|0⟩⟨0|+|+⟩⟨+|), ½(|0⟩⟨0|+|−⟩⟨−|), ½(|1⟩⟨1|+|+⟩⟨+|), ½(|1⟩⟨1|+|−⟩⟨−|)} one has ||√π'√π''|| ≤ β = (2+√2+√(4√2−2))/8 ≈ 0.6658. This is false. Taking π' = ½(|0⟩⟨0|+|+⟩⟨+|) and π'' = ½(|0⟩⟨0|+|−⟩⟨−|), both of which are in the stated set, a direct calculation gives ||√π'√π''||² = λ_max(π'π'') = (2+√3)/8, hence ||√π'√π''|| ≈ 0.6830 > β. Consequently the elementwise bound Γ ≤ G_k in Eq. (B18) does not follow, and the subsequent bound ||Γ|| ≤ ||G_k||, which is the step that yields the q-range q ≥ α^{N−1}β, is not justified. As written, the analytic proof of Proposition 2 is incomplete; the exhaustive search for N ≤ 3 and the conjecture for N > 3 do not repair the proof for general N. The claimed threshold η* = 1/2^N for rCHSH_N may still be true with a corrected constant, but that corrected proof is not in the manuscript.","section":null},{"comment":"The abstract states that the exponential decrease is achieved 'robustly', but Proposition 3 is conditional on an unquantified robust self-testing function f(N,ε) in Eq. (40). No explicit robust self-testing bound for the N-product CHSH inequality is provided, and the analogous robustness statement for rCHSH_N is explicitly deferred to future work. In addition, the proposition does not ensure that the chosen penalty q = 1/√2 + √δ remains in the admissible range [1/√2, 1] of Proposition 1, since δ is only bounded by an expression involving f(N,ε). As a result, the 'robustly' and 'arbitrary distances' claims in the abstract and conclusions are stronger than what the analytic results support. The robustness statement should be qualified, or the missing robust self-testing input should be supplied with explicit bounds.","section":null}],"minor_comments":[{"comment":"The sentence 'We assume that the detection efficiency is the same for all measurements in B1 and therefore p(b = ∅|y) = η' is inconsistent with η being the detection efficiency (click probability); it should read p(b = ∅|y) = 1 − η. The strategy value in Eq. (16) is consistent with the latter convention.","section":null},{"comment":"The notation ||M_k|| is used for the maximal eigenvalue of M_k, whereas elsewhere in the paper ||·|| denotes the operator norm. Since M_k can have negative eigenvalues, these two quantities differ; please introduce a distinct symbol such as λ_max(M_k).","section":null},{"comment":"There is a typographical error: 'f_k ≡ ... +−kq' should read 'f_k ≡ ... − kq'.","section":null},{"comment":"The statement that an exhaustive search for N=1,2,3 yields β' and that the bound is conjectured for N>3 should be clearly separated from the analytic claim of Proposition 2; currently the text reads as if Proposition 2 itself relies on the conjecture, which is not the case.","section":null},{"comment":"The normalization of the operators S_j in Appendix C should be reconciled with the conditional states in Eq. (14): in the ideal case the conditional states have trace one, whereas the S_j defined in Eq. (C8) appear to be sub-normalized. Please clarify this so that the constants in the threshold (41) can be verified.","section":null}],"recommendation":"major_revision","confidential_remarks":"The paper contains an interesting and likely correct construction for rBB84_N, but the rCHSH_N analytic proof currently hinges on a false constant in Eq. (B17). This is easily checkable with the provided code and should be fixable; the result may survive with a corrected β or a modified argument. The robustness claims in the abstract also need to be toned down or substantiated. I would encourage the editor to seek a revised version that addresses these points."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nHere's my read of arXiv:2504.17621. The headline result—that parallel repetition of the routed rBB84 strategy certifies non-jointly-measurable measurements in B1 at every detection efficiency η > 1/2^N—is real and the proof in Proposition 1 is coherent. It uses the self-testing implications of maximal N-product CHSH violation to pin down Alice's marginals and Bob's remote states, then reduces the JM bound to an operator norm problem. The tightness argument from B1 having 2^N settings is simple but correct, and the NPO numerics for N=2 are a useful bonus. The code is available, and the note-added discussion of the competing qubit strategy [38] is honest.\n\nThe soft spot is load-bearing: the rCHSH_N half does not hold up as written. In Appendix B, Eq. (B17) claims that for any two distinct π' in the set {1/2(|0><0|+|+><+|), 1/2(|0><0|+|−><−|), 1/2(|1><1|+|+><+|), 1/2(|1><1|+|−><−|)}, ||√π'√π''|| ≤ β ≈ 0.6658. That is false. Take π' = (|0><0|+|+><+|)/2 and π'' = (|0><0|+|−><−|)/2. Then ||√π'√π''||² = λ_max(π'^{1/2}π''π'^{1/2}) = (2+√3)/8, so the norm is ≈ 0.6830 > β. Consequently the elementwise bound Γ ≤ G_k in (B18)–(B19) does not follow, and neither does ||Γ|| ≤ ||G_k||. Since that step supplies the q-range and the JM bound in Proposition 2, the analytic claim for rCHSH_N is unproven. The threshold may survive with a corrected constant—the exhaustive search in the paper suggests the constant is not far off—but the proof in the appendix is not valid.\n\nThere's also a smaller overreach in the abstract's \"robustly\": Proposition 3 is conditional on an unspecified robust self-testing function f(N, ε) and applies only to rBB84_N; the analogous rCHSH_N statement is explicitly left to future work. That's a fair limitation to state, but it should be labeled as such rather than folded into the summary claim.\n\nFor a referee: the paper deserves review, because the rBB84_N result is substantial and the rCHSH_N flaw is concrete and likely fixable. But the current version should not be accepted until Proposition 2 is repaired or removed from the claims.","headline":"The rBB84_N threshold is a clean, citable result, but the rCHSH_N proof has a false bound in Eq. (B17) that invalidates Proposition 2 as written.","tokens_in":23016,"tokens_out":5500,"would_cite":true,"duration_ms":49180,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Parallel repetition of routed Bell tests makes the critical detection efficiency drop exponentially to 1/2^N.","keywords":["routed Bell experiments","parallel repetition","detection efficiency","non-jointly measurable measurements","device-independent quantum key distribution","self-testing","CHSH inequality","BB84 inequality"],"falsifier":"For $N=4$, evaluate the largest eigenvalue of every operator $\\sum_{y: b_y \\neq \\varnothing} ( \\bigotimes_j A^{y_j}_{b_{y,j}} - q I )$ at $q = 1/\\sqrt{2}$; if any eigenvalue exceeds $1 - 1/\\sqrt{2}$, the jointly-measurable bound in Proposition 1 is false.","tokens_in":21892,"feed_emoji":"🔑","tokens_out":12574,"duration_ms":108534,"temperature":0.7,"pith_summary":"This paper tries to establish that the detector-efficiency barrier to long-distance quantum correlations can be broken exponentially by repeating routed Bell experiments in parallel. Concretely, for any number $N$ of parallel copies, the two strategies $rBB84_N$ and $rCHSH_N$ certify that the distant device's measurements are non-jointly measurable—the property that blocks an eavesdropper from faking the statistics—whenever the distant detection efficiency exceeds $\\eta^* = 1/2^N$. Because the distant device has $2^N$ measurement settings, this threshold cannot be lowered for these strategies, so the exponential decrease is optimal. If true, the result is a concrete route toward loophole-free certification and device-independent quantum key distribution over much longer distances, at the price of preparing $2^N$-dimensional entangled states and keeping the near-source devices near-perfect.","feed_headline":"Parallel Bell tests cut detection threshold to 1/2^N","feed_subtitle":"The distant device's required efficiency falls exponentially while staying optimal—a route to long-range quantum keys.","key_machinery":"The load-bearing object is the $q$-penalized $N$-product Bell inequality, $B_N(q) = B_N - (q/2^N)\\sum_{y,b} p_B(b|y,1)$, a generalized long-path inequality that discounts every click at the distant device. Self-testing is the second ingredient: maximal violation of the $N$-product CHSH inequality between $A$ and $B_0$ implies, up to an isometry and a junk state, the reference strategy, yielding two consequences used throughout: Alice's marginals are uniform and Bob's remotely prepared states are Alice's reference projectors. Under a jointly measurable model at $B_1$, the $q$-penalized value becomes a sum of terms involving operators $C_{\\vec b}$ built from tensor products of projections; bounding the largest eigenvalue of these operators through a norm inequality for sums of positive semidefinite operators gives the jointly measurable ceiling $(1-q)/2^N$. Comparing the ceiling with the ideal value $(1-q)\\eta$ yields $\\eta^* = 1/2^N$; the $rCHSH_N$ case follows the same template with two-qubit density operators replacing projections.","core_discovery":"The central claim is that the $N$-fold parallel repeated routed strategies $rBB84_N$ and $rCHSH_N$ achieve the optimal detection-efficiency threshold $\\eta^* = 1/2^N$ for certifying that the far device $B_1$ performs non-jointly measurable (NJM) measurements, assuming the source and the near devices $A$ and $B_0$ are perfect. The proof introduces $q$-penalized versions of the $N$-product BB84 and CHSH inequalities, in which every click of $B_1$ is penalized. Self-testing from the maximal violation of the $N$-product CHSH inequality between $A$ and $B_0$ forces Alice's outcomes to be uniform and fixes Bob's remotely prepared states; inserting this into a jointly measurable model of $B_1$ bounds any such model by $(1-q)/2^N$, while the ideal strategies achieve $(1-q)\\eta$. Hence NJM is certified whenever $\\eta > 1/2^N$, and since $B_1$ has $2^N$ settings the threshold is tight.","pith_inferences":["Beyond the paper: because NJM certification is the cryptographic precondition, the same $1/2^N$ law should set the threshold for positive key rates in DIQKD, but proving that requires a full finite-key analysis the paper does not perform.","Beyond the paper: the tightness argument suggests a general trade-off: any routed strategy giving $B_1$ $m$ settings cannot certify NJM below $1/m$, so the exponential gain is bought by exponential dimension and settings, meaning a practical sweet spot likely lies at moderate $N$.","Beyond the paper: the norm-bound technique, applied here to product BB84 and CHSH inequalities, may extend to other self-testable Bell inequalities; a testable prediction is that the same $1/2^N$ scaling appears for any inequality whose maximal violation self-tests product measurements.","Beyond the paper: the numerical $N=2$ robustness comparison suggests that higher-dimensional parallel strategies tolerate source noise better than qubit strategies with the same number of distant settings, a trend that could be checked systematically for $N=3$ with improved optimization methods."],"forward_implications":["Each additional parallel copy halves the detection efficiency needed to certify non-jointly measurable measurements at the distant device, from $1/2$ at $N=1$ to $1/2^N$ in general.","The threshold is tight for these strategies: because $B_1$ has $2^N$ settings, no certification of NJM measurements at that device can succeed at efficiency below $1/2^N$.","The certification step on which secure device-independent quantum key distribution relies can in principle run at this exponentially lower efficiency, although a full key-rate analysis is not part of the paper.","The cost of the exponential gain is exponential resources: the shared state has local dimension $2^N$, and the distant device needs $2^N$ measurement settings.","For $N=2$, numerical relaxations show that both parallel strategies tolerate source noise better than their single-copy counterparts, and better than a qubit-based strategy with four distant settings."],"supporting_citations":[{"why":"Supplies the parallel-repetition method that drives critical detection efficiency down as $(C/Q)^N$, which this paper adapts to routed Bell experiments.","marker":"[16]"},{"why":"Introduces routed Bell experiments and the idea of using near-source correlations between $A$ and $B_0$ to lower the distant device's threshold.","marker":"[17]"},{"why":"Establishes the single-copy $\\eta^*=1/2$ threshold for binary routed strategies and the jointly measurable model that must be beaten for security.","marker":"[22]"},{"why":"Shows that the single-copy rBB84 and rCHSH strategies support DIQKD at $\\eta^*\\approx 1/2$, the baseline that the exponential threshold extends.","marker":"[26]"},{"why":"Provides the self-testing statement for maximal violation of the $N$-product CHSH inequality that yields uniform marginals and the remote-state form used in the proofs.","marker":"[29]"},{"why":"Supplies the operator-norm bound for sums of positive semidefinite operators used to control the eigenvalue of the operators $C_{\\vec b}$.","marker":"[36]"}],"fun_headline_variants":["Parallel Bell tests: detection threshold drops to 1/2^N","Exponential reduction: N Bell tests need only 1/2^N efficiency","Long-range QKD: parallel tests hit optimal 1/2^N threshold","Routed Bell tests: efficiency requirement falls as 1/2^N","Optimal scaling: parallel Bell tests achieve 1/2^N threshold"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proofs require the near-source devices $A$ and $B_0$ to achieve the exact maximum quantum violation of the $N$-product CHSH inequality, because the self-testing relations used to bound jointly measurable models hold only under that perfection.","fun_headline_variants_meta":{"raw":{"variants":["Parallel Bell tests: detection threshold drops to 1/2^N","Exponential reduction: N Bell tests need only 1/2^N efficiency","Long-range QKD: parallel tests hit optimal 1/2^N threshold","Routed Bell tests: efficiency requirement falls as 1/2^N","Optimal scaling: parallel Bell tests achieve 1/2^N threshold"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000287,"raw_usage":{"total_tokens":1658,"prompt_tokens":890,"completion_tokens":768,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":506,"completion_tokens_details":{"reasoning_tokens":666}},"tokens_in":506,"tokens_out":768,"duration_ms":7116,"temperature":1.0,"reasoning_tokens":666,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T10:35:17.428535+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For $N=4$, evaluate the largest eigenvalue of every operator $\\sum_{y: b_y \\neq \\varnothing} ( \\bigotimes_j A^{y_j}_{b_{y,j}} - q I )$ at $q = 1/\\sqrt{2}$; if any eigenvalue exceeds $1 - 1/\\sqrt{2}$, the jointly-measurable bound in Proposition 1 is false.","supporting_citations":[{"cited_title":"Massar and S","cited_arxiv_id":null,"evidence_quote":"Supplies the parallel-repetition method that drives critical detection efficiency down as $(C/Q)^N$, which this paper adapts to routed Bell experiments."},{"cited_title":"Chaturvedi, G","cited_arxiv_id":null,"evidence_quote":"Introduces routed Bell experiments and the idea of using near-source correlations between $A$ and $B_0$ to lower the distant device's threshold."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the single-copy $\\eta^*=1/2$ threshold for binary routed strategies and the jointly measurable model that must be beaten for security."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the operator-norm bound for sums of positive semidefinite operators used to control the eigenvalue of the operators $C_{\\vec b}$."}],"review_version":1}