{"id":"c8a7bde4-c667-4ca4-bad9-ee4f4750b9ff","arxiv_id":"2509.08623","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Any nonlocal quantum behavior certifies some device-independent randomness when all input pairs are used for generation; the input-averaged guessing probability is a faithful, monotonic nonlocality measure.","lead":"This paper proves that any quantum correlation that violates a Bell inequality can certify some device-independent randomness if the protocol uses every measurement setting for generation instead of spot-checking one fixed pair. It introduces the input-averaged guessing probability as a faithful nonlocality measure and computes it analytically for the CHSH test.","discovery_kind":"unification","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 3's proof asserts that joint non-disturbance (Prop. 1) implies individual non-disturbance (Eq. B12); this implication is not generally valid and is essential for the sequential construction of P_JPD.","rationale":"The reader's weakest_assumption is exactly the non-disturbance step: Proposition 1 proves joint non-disturbance, but the proof needs individual non-disturbance (Eq. B12) to make the sequential construction of P_JPD reproduce the quantum marginals. I agree with this identification. The gap is load-bearing because Theorem 3 is the central claim and without (B12) the locality conclusion does not follow. However, the concern is a proof gap, not a demonstrated falsehood; the theorem may still be true and the step may be repairable using the full strength of the perfect-guessing condition. Therefore the CONDITIONAL verdict is appropriate and no change is needed.","tokens_in":30623,"tokens_out":16810,"duration_ms":167370,"concrete_test":"Run an SDP search (NPA level 2 or 3) in the (2,2,2,2) scenario for a tripartite quantum strategy {rho_ABE, Pi_x^a, Lambda_y^b, Gamma_{x,y}^{a,b}} satisfying the perfect-guessing equalities (B2): sum_{a,b} Tr(Pi_x^a⊗Lambda_y^b⊗Gamma_{x,y}^{a,b} rho_ABE)=1 for all x,y, and optimize, for fixed x, the trace distance ||Phi_x⊗1(rho_AB)-rho_AB||_1. If a strategy with nonzero distance exists, Eq. (B12) is false and the proof's sequential construction does not go through; if the maximum over strategies is zero, then the missing lemma is true and should be stated and proved. As a perturbation check, also verify the marginals P_JPD(a_2,b_1|x_2,y_1) computed from (B10) against p(a_2,b_1|x_2,y_1) in any candidate strategy.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorem 3 constructs a classical joint distribution P_JPD in Eq. (B10) by sequentially applying Alice's and Bob's measurements to a single copy of rho_AB. To show that its marginals reproduce the quantum behavior, it uses Eq. (B12): rho_AB(k,l)=rho_AB for all k,l, which requires that each individual measurement channel Phi_x: rho -> sum_a Pi_x^a rho Pi_x^a and Psi_y leave rho_AB invariant. Proposition 1 (Eq. B4) proves only the joint invariance (Phi_x⊗Psi_y)(rho_AB)=rho_AB for each pair (x,y). Joint invariance does not imply local invariance: e.g., for rho_AB=(|00><00|+|11><11|)/2, the (non-measurement) channels Phi_x=X(·)X and Psi_y=X(·)X satisfy (Phi_x⊗Psi_y)(rho_AB)=rho_AB, while Phi_x⊗1(rho_AB)≠rho_AB. The manuscript asserts 'Evidently... the local measurements do not disturb the state' (after Eq. B11) without using the perfect-guessing condition (B2) beyond Proposition 1. If (B12) fails, the equalities in (B13) fail for pairs with k≠l, and P_JPD need not reproduce p; the immunization argument collapses. This is the key step connecting perfect guessing to locality and hence the central no-bound-randomness claim.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies device-independent randomness certification against quantum adversaries and distinguishes two protocol structures: spot-checking, where a fixed input pair is used for generation, and amplification-style protocols, where all input pairs are used. The authors introduce the average guessing probability P_g^(q)(p, mu_XY) and claim (Theorem 2/3) that any nonlocal bipartite quantum behavior satisfies P_g^(q)(p, mu_XY)<1 for any full-support input distribution, i.e., quantum nonlocality is sufficient for randomness amplification and there is no 'bound randomness' against quantum adversaries. They then argue that the average guessing probability is a faithful and monotonic measure of nonlocality (Appendix D), use it to show that the I_4422 inequality certifies randomness at its nonlocality detection-efficiency threshold (Appendix C), provide an analytical formula for the average guessing probability in CHSH (Appendix E), and demonstrate improved amplification rates via a numerical entropy calculation (Appendix F). The central proof is an immunization argument that builds a classical joint distribution from a hypothetical perfect-guessing tripartite strategy.","tokens_in":30783,"tokens_out":16623,"duration_ms":185692,"significance":"If the central theorem holds, it is a conceptually important result: it complements the recent finding that maximum nonlocality can be useless for fixed-setting spot-checking protocols by showing that the same resource is sufficient in amplification protocols with randomized inputs. It also gives a concrete operational meaning to the average guessing probability as a nonlocality measure and yields a practical improvement for the I_4422 inequality. The paper contains explicit calculations, a candidate analytical formula, an SOS decomposition for a new Bell expression, and numerical SDP checks. However, the main theorem's proof has a load-bearing gap, and the advertised analytic CHSH result relies on an imported conjectured formula. These issues currently prevent acceptance.","major_comments":[{"comment":"The proof of Theorem 3 hinges on the claim that the local measurements do not disturb rho_AB. Proposition 1 establishes only the joint non-disturbance (Phi_x⊗Psi_y)(rho_AB)=rho_AB for each pair. The sequential construction of P_JPD in Eq. (B10) requires the individual conditions Phi_{x_k}(rho_AB)=rho_AB and Psi_{y_l}(rho_AB)=rho_AB; without them the equality in Eq. (B13) holds only for the first pair and not for later pairs. Joint invariance does not imply local invariance for general channels; for example, for rho_AB=(|00><00|+|11><11|)/2, the unitary channels Phi_x=Psi_y=X(·)X satisfy joint invariance while local invariance fails. The sentence 'Evidently...' after Eq. (B11) is therefore not a proof. Additionally, Eq. (B6) uses rho_AB=tr_E[1⊗1⊗Gamma_{(x,y)}(rho_ABC)], which is not true for arbitrary POVM measurement channels unless Gamma is projective or a dilation argument is supplied.","section":"Appendix B, Eqs. (B4)-(B13)"},{"comment":"The advertised analytical expression for the average guessing probability in CHSH, Eq. (8) in the main text and Eq. (E2) in the appendix, depends on the maximum quantum value of the double-tilted CHSH expression, Eq. (E5), imported from Ref. [34]. That reference is a preprint whose title indicates a conjectured analytical solution. The manuscript treats Eq. (E5) as an established bound, and the only verification offered is numerical NPA coincidence to 10^{-9}. Since the claim 'analytically compute the average guessing probability' is one of the paper's main results, the formula must either be proved or explicitly flagged as conditional on the conjecture.","section":"Appendix E, Eq. (E5) and Theorem 4"},{"comment":"The monotonicity proof is not rigorous. In the input-substitution case, Eqs. (D10)-(D12) relabel x_i and x_j without a corresponding transformation of the input weights, and Eq. (D12) asserts an equality to a maximum that does not follow from the preceding optimization. The coarse-graining argument is also only sketched. Since Lemma 4's claim that the average guessing probability defines a nonlocality measure depends on Lemma 3, these arguments need to be made rigorous or replaced by a reference to a known resource-theoretic result.","section":"Appendix D, Lemma 3"}],"minor_comments":[{"comment":"The phrase 'the detection efficiency threshold for randomness generation is never lower than that for nonlocality detection' appears to be the opposite of what Theorem 3 implies. If every nonlocal behavior certifies randomness on average, the randomness threshold is no higher (in fact equal) to the nonlocality threshold. Please clarify.","section":"Abstract and Section 'Detection efficiency thresholds'"},{"comment":"There is a typo in Eq. (C1): 'p(A_4-1)' should presumably be 'p(A_4=1)', and similarly for B_4. The formula should be checked carefully.","section":"Appendix C, Eq. (C1)"},{"comment":"The captions use x=y∈{0,1,2,3} while the text uses {1,2,3,4}. Please standardize the indexing.","section":"Figures 1 and 2"},{"comment":"The statement 'sufficient resource for DI randomness amplification' should be supported by an explicit protocol statement or a precise reference. P_g^(q)(p,mu)<1 is a necessary precondition, but by itself it does not automatically yield amplification of an SV source; the manuscript should make clear how the existing protocols of Refs. [5-11] convert this into full amplification.","section":"Theorem 2"},{"comment":"There is a typo: 'adversary's adversary's side information' should be 'adversary's side information'.","section":"Appendix A.2"},{"comment":"There are minor notation issues in the derivation, e.g., 'p^{A|X,Z,E}(a|A0,z=0,e=a)' is used before being defined, and some parentheses in the displayed f_phi formula are unbalanced in the main text.","section":"Appendix E, Eq. (E13)"}],"recommendation":"major_revision","confidential_remarks":"I believe the main no-bound-randomness claim is likely correct and that the proof gap in Appendix B may be repairable with a careful dilation/fixed-point argument. However, as written, the central theorem is not established, and the analytic CHSH claim depends on a conjectured external result. These are not merely cosmetic issues, so major revision is appropriate."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First thing you should know: this is a serious paper, not a crank submission. The main conceptual claim is that any nonlocal quantum behavior certifies randomness when inputs are drawn from a full-support distribution (amplification structure), in contrast to the fixed-input-pair spot-checking case. If true, that is an important clarification of the nonlocality/randomness resource relation. But the proof of the central theorem (Theorem 3, Appendix B) has a gap that is load-bearing, and the current text does not close it.\n\nWhat is genuinely new: the input-averaged guessing probability as a faithful monotone measure of nonlocality (Lemmas 3–4 plus Appendix D), the analytic CHSH average guessing probability (Theorem 4, modulo the imported conjecture from [34]), and the correction to the I_4422 detection-efficiency story — the paper shows you can certify randomness down to the 61.8% threshold using the average measure, contrary to the fixed-setting conclusion of Zhang et al. These are real contributions and will be useful even if Theorem 3 needs work.\n\nThe soft spots, in order of seriousness:\n\n1. The stress-test concern is correct. Proposition 1 proves joint non-disturbance (Π_x⊗Λ_y)(ρ_AB)=ρ_AB. Eq. (B12) requires each local channel Π_x and Λ_y to leave ρ_AB invariant. That implication does not follow from the stated argument. The paper says 'Evidently' and moves on. This is the bridge from perfect guessing to locality, so Theorem 3 is not actually established by the text. The counterexample in the note uses non-measurement channels, so it does not settle the question for measurement channels, but the burden is on the authors to prove the implication or restructure the construction. A referee should demand this.\n\n2. Main-text Eq. (8) states equality where Appendix E proves a lower bound and only numerical tightness at NPA level 3 (1e-9). That is a mismatch that should be corrected.\n\n3. The analytic CHSH formula leans on a conjectured value from [34]. The numerics make it believable, but the bound is conditional on a conjecture.\n\n4. Theorem 2's SV-source bound has a dimension error for binary inputs: (1/2±ε) log2 |X||Y| exceeds 1 when |X||Y|=4. Likely a typo, but it is in a theorem statement.\n\n5. No code or data shipped for the SDP numerics. For a paper with this many plots, that is a reproducibility gap.\n\nAppendix D's monotonicity argument for input substitution is also not fully rigorous as written.\n\nWho is this for? Anyone working on DI randomness expansion or amplification, and on nonlocality measures. It deserves a serious referee — I would send it out — but the referee should have time to check Appendix B carefully. If the gap is fixable, this becomes a strong paper. As is, the central theorem should be treated as conditional.\n\nRecommendation: engage with it, but ask for a rewritten Theorem 3 proof.","headline":"Central no-bound-randomness claim is plausible and worth taking seriously, but Theorem 3 has a real proof gap in the passage from joint to local non-disturbance; the theorem is unproven as written and needs a repair before the paper's main conclusion can be trusted.","tokens_in":31459,"tokens_out":13432,"would_cite":true,"duration_ms":137843,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P13","81P94"],"pacs":["03.65.Ud","03.67.Dd"],"model":"deepseek-v4-flash","headline":"Any amount of quantum nonlocality suffices to certify randomness against a quantum adversary, once randomness is averaged over all input settings.","keywords":["quantum nonlocality","device-independent randomness","guessing probability","bound randomness","randomness amplification","Bell inequalities","CHSH","detection efficiency"],"falsifier":"Construct any nonlocal bipartite behavior p and full-support input distribution mu_XY with an explicit tripartite quantum strategy achieving average guessing probability exactly 1; or, more surgically, exhibit a perfect-guessing tripartite strategy in which the joint measurement leaves rho_AB invariant but a single local measurement (Pi_x tensor identity or identity tensor Lambda_y) changes it. Either would invalidate Theorem 3.","tokens_in":30349,"feed_emoji":"🎲","tokens_out":5932,"duration_ms":67328,"temperature":0.7,"pith_summary":"The paper establishes that quantum nonlocality is both necessary and sufficient for device-independent randomness, provided the protocol does not fix a single measurement setting. It proves that for any nonlocal quantum behavior, an adversary cannot perfectly guess the outputs on average over all inputs: the average guessing probability is strictly below 1. This rules out “bound randomness” in quantum nonlocality and makes the average guessing probability a faithful and monotonic measure of nonlocality, unlike the fixed-setting guessing probability. The same perspective lowers the detection-efficiency threshold for randomness certification from the I_4422 inequality and yields a new analytic formula for the average guessing probability in a CHSH test.","feed_headline":"Quantum nonlocality always yields device-independent randomness","feed_subtitle":"All nonlocal correlations beat an adversary's perfect guess once randomness is averaged over every measurement setting.","key_machinery":"The carrying mechanism is a classical joint probability distribution P_JPD assembled by ordering Alice's and Bob's inputs and applying their POVMs sequentially to a single copy of the marginal state rho_AB. Perfect predictability by Eve forces each joint measurement to leave rho_AB invariant, via the gentle measurement lemma, so the sequential marginals reproduce the original quantum probabilities and yield a local model, contradicting nonlocality. For the CHSH computation, the key object is the double-tilted CHSH expression I_CHSH^(alpha,phi) = alpha cos(phi/2) A0 + alpha sin(phi/2) A1 + I_CHSH; its quantum maximum bounds the tilted correlator combination, and translating the input bias p i","core_discovery":"The central claim is Theorem 3: no nonlocal quantum behavior exhibits bound randomness against a quantum adversary. If a tripartite quantum strategy allowed Eve to guess Alice and Bob's outputs perfectly for every input pair, the proof constructs a classical joint distribution—obtained by running all of their measurements sequentially on one copy of the shared marginal state—whose marginals reproduce the observed behavior, making the behavior local. Since a local behavior cannot win any nonlocal game above its classical value, perfect guessing is incompatible with nonlocality. Theorem 2 then states that for any nonlocal bipartite behavior p and any full-support input distribution mu_XY, the","pith_inferences":["If the theorem is correct, other multi-input Bell inequalities with apparently high randomness thresholds may also certify randomness at their nonlocality threshold once protocols average over all input pairs; I_4422 is one worked example.","The same average-guessing-probability quantity could be applied to device-independent quantum key distribution and to expansion protocols beyond amplification, potentially closing gaps left by fixed-setting security analyses.","A concrete experimental test would be to measure the average min-entropy directly in an existing Bell test with a small but nonzero violation: positive min-entropy would confirm the mechanism, while a finding of zero would challenge it.","The sequential-measurement construction suggests a resource-theoretic reading: quantum nonlocality and certifiable randomness may coincide as resources once the protocol structure is optimized, rather than being distinct operational resources."],"forward_implications":["Any nonlocal quantum behavior, however small the violation, certifies positive min-entropy in a device-independent randomness amplification protocol against a quantum adversary.","The average guessing probability, unlike the fixed-input guessing probability, is faithful and monotonic under single-copy WCCPI operations and therefore serves as a proper nonlocality measure.","For the I_4422 inequality, randomness can be certified for every detection efficiency above 61.8%, matching the nonlocality detection threshold instead of the higher fixed-setting threshold found previously.","The analytic CHSH bound extends the classical fixed-setting formula to biased input distributions and recovers the known formula as the p=1 limit, with numerical tightness confirmed at high NPA level.","The local-model argument does not rely on pseudo-telepathy games, so every nonlocal game, and by extension multipartite nonlocal behaviors, supports randomness amplification in this protocol structure."],"supporting_citations":[{"why":"Shows maximally nonlocal quantum behaviors with fixed-setting guessing probability 1, the negative result the paper refutes by switching to the average guessing probability.","marker":"[12]"},{"why":"Supplies the immunization and three-player game hardness techniques on which the Theorem 3 local-model construction is built.","marker":"[17]"},{"why":"Provides the analytical double-tilted CHSH quantum value used to derive the closed-form average guessing probability in Appendix E.","marker":"[34]"},{"why":"Reports the numerical 90.6% fixed-setting detection threshold for I_4422 that the average-probability calculation replaces with 61.8%.","marker":"[20]"},{"why":"Establishes the 61.8% nonlocality detection threshold for I_4422, the benchmark the paper's randomness threshold matches.","marker":"[21]"},{"why":"Defines the fixed-setting CHSH guessing probability formula recovered as the p=1 limit of the new average formula.","marker":"[1]"},{"why":"Provides the semidefinite-programming method for device-independent lower bounds on conditional von Neumann entropy used in the demonstrated rate improvements.","marker":"[26]"},{"why":"Is the state-of-the-art randomness amplification protocol whose generation rates are improved using the average entropy.","marker":"[6]"}],"fun_headline_variants":["No bound randomness: any nonlocal correlation certifies randomness","Average guessing score proves all nonlocality yields randomness","Any nonlocal behavior defeats a quantum adversary's perfect guess","Nonlocality always beats quantum guessing when averaged over inputs","No bound randomness: averaging inputs makes all nonlocality useful"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"The proof rests on the step where perfect guessing is taken to imply that Alice's and Bob's individual measurements do not disturb their shared marginal state; the paper explicitly proves the joint non-disturbance and asserts the individual form needed for the sequential local model.","fun_headline_variants_meta":{"raw":{"variants":["No bound randomness: any nonlocal correlation certifies randomness","Average guessing score proves all nonlocality yields randomness","Any nonlocal behavior defeats a quantum adversary's perfect guess","Nonlocality always beats quantum guessing when averaged over inputs","No bound randomness: averaging inputs makes all nonlocality useful"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00146,"raw_usage":{"total_tokens":5692,"prompt_tokens":706,"completion_tokens":4986,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":450,"completion_tokens_details":{"reasoning_tokens":4905}},"tokens_in":450,"tokens_out":4986,"duration_ms":38325,"temperature":1.0,"reasoning_tokens":4905,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T20:20:20.876348+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Construct any nonlocal bipartite behavior p and full-support input distribution mu_XY with an explicit tripartite quantum strategy achieving average guessing probability exactly 1; or, more surgically, exhibit a perfect-guessing tripartite strategy in which the joint measurement leaves rho_AB invariant but a single local measurement (Pi_x tensor identity or identity tensor Lambda_y) changes it. Either would invalidate Theorem 3.","supporting_citations":[],"review_version":1}