{"id":"36895909-3c83-44fa-99d9-03a44c258d74","arxiv_id":"1908.07694","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper derives tight, SDP-computable majorization bounds for the probability vector of a post-test measurement conditioned on a pre-test outcome, and uses them to outer-approximate arbitrary uncertainty regions.","lead":"This paper introduces a \"complementary information principle\" that bounds the possible outcomes of a second measurement after the first measurement's result is known. It gives computable, optimal bounds under majorization and uses them to approximate uncertainty regions for any uncertainty measure.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"One-step flatness construction of the optimal upper bound t may fail to produce a valid probability vector; Theorem 2's explicit t needs an iteration or an additional proof.","rationale":"The reader identifies the i.i.d.-copies assumption as the weakest point, but that assumption is explicit in the paper's black-box framework and is a reasonable idealization; sequential disturbance is outside the stated model. The more load-bearing issue is internal to the proof of the central theorem: the upper bound t is constructed by a single flatness step, yet the flatness process for constructing least concave majorants is generally iterative, and the paper provides no proof that one pass suffices for the special s_k generated by the quantum SDPs. A concrete numerical counterexample to the general algorithm shows the stated rule can yield a vector not in P_n; if such an s_k is realizable in the quantum setting, Theorem 2's t would not be the true supremum, weakening the claimed optimality. The mathematical existence of a correct t is protected by lattice completeness, so the issue is repairable by invoking the iterative algorithm, which is why the verdict should remain CONDITIONAL rather than outright reject. The paper does contain independent support: explicit qubit analytical formulas, a qutrit example, and detailed comparisons with UUR and DSMUR, all of which are unaffected by this construction issue. Still, the universal uncertainty region and the joint-uncertainty bounds inherit the correctness of t, so the proof gap warrants a conditional acceptance pending clarification or correction.","tokens_in":26931,"tokens_out":14621,"duration_ms":226177,"concrete_test":"For n=5, generate many random rank-1 projective measurements M,N and a random pre-test distribution p (e.g., 100 instances). Solve the SDPs in Eq. (6) to obtain s_k for k=1..5, set S_k=s_k-s_{k-1}, and apply the paper's one-step flatness rule from Eq. (7). Verify (i) the resulting T_k are nonincreasing and nonnegative, and (ii) for all k, the cumulative sum t_k is at least s_k. If any instance fails, iterate the flatness process until it converges and compare: if the iterative t differs from the one-step t, Theorem 2's explicit formula is not correct as stated. A targeted instance with S increments designed so that S_{j+1} exceeds the averaged block value would be a decisive probe.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central constructive claim is that Theorem 2's optimal upper bound t is obtained from the SDP values s_k by a single flatness step (main text after Theorem 1, Eq. (7); SM proof of Theorem 2). The standard flatness process is iterative: after averaging the block S_i,...,S_j, a new violation S_{j+1} > a can appear, so one pass need not yield a nonincreasing probability vector. For example, increments S=(0.4,0.1,0.2,0.3,0), with cumulative s_k=(0.4,0.5,0.7,1,1), have first violation at j=3; the stated rule selects i=2, a=0.15, producing T=(0.4,0.15,0.15,0.3,0), which is not in P_n and does not satisfy the required monotonicity. The paper gives no argument that s_k arising from the quantum SDPs in Eq. (6) avoid this phenomenon for n>=5. Since t is the key object in Theorem 2, and later in the universal uncertainty region and joint-uncertainty bounds, a t outside P_n, or one that is not the true least upper bound, invalidates the claimed optimality. The existence of a unique supremum is guaranteed by completeness of the majorization lattice, so the conceptual framework survives; the flaw is in the constructive specification of t.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a 'complementary information principle' for sequential black-box testing of two incompatible measurements. For a fixed pre-test measurement M with outcome distribution p, the set S(M,p) of states compatible with p defines a set Q of post-test probability vectors for measurement N. The authors claim that Q has a unique greatest lower bound r and least upper bound t under majorization, that both can be computed by semidefinite programs (Theorem 1), and that they are optimal because the majorization lattice is complete (Theorem 2). They then use r and t to outer-approximate uncertainty regions for arbitrary Schur-concave uncertainty measures, prove exactness for qubits, and derive state-independent bounds on general joint uncertainty measures, with comparisons to Maassen-Uffink and direct-sum/direct-product majorization bounds.","tokens_in":27163,"tokens_out":9821,"duration_ms":275768,"significance":"If the main theorem is correct, the paper provides a measure-independent characterization of conditional post-test uncertainty and a general method for outer-approximating uncertainty regions, with exact qubit results. The explicit SDP formulations, the analytic qubit bounds, and the detailed comparisons with Maassen-Uffink and direct-sum/direct-product majorization relations are useful and go beyond previous work that fixes a particular uncertainty measure. The conceptual move of conditioning on partial pre-test information instead of assuming full state knowledge is also worthwhile. However, the constructive specification of the optimal upper bound t is not fully justified as written, and this is load-bearing for Theorem 2 and all subsequent applications.","major_comments":[{"comment":"The one-step flatness construction in Eq. (7) does not, in general, produce a non-increasing probability vector, so the proof of Theorem 2 is incomplete for the t defined there. For example, applied to S=(0.4,0.1,0.2,0.3,0), the stated rule gives j=3, i=2, a=0.15, and hence T=(0.4,0.15,0.15,0.3,0), which is not in P_n. The proof of Theorem 2 uses t in P_n as a non-increasing vector and as the least concave majorant of the partial sums; it also asserts t=⋁Q only after arguing that t∈P_n. No argument is given that the particular s_k values arising from the quantum SDPs in Eq. (6) avoid this phenomenon for n≥4. The fix is local: the flatness step should be iterated until the vector is non-increasing, or the authors should prove that one pass is terminal for the special class of s_k from Eq. (6). Until then, the optimality of t, and therefore the universal uncertainty region and joint-uncertainty bounds that rely on t, are not established.","section":"Main text, Eq. (7); Supplemental Material, proof of Theorem 2"}],"minor_comments":[{"comment":"The claim that the bounds 'can be efficiently computed' and that s_k 'can still be solved efficiently via parallel computations even for large dimension n' is an overstatement: computing all s_k requires solving sum_{k=1}^n binom(n,k)=2^n-1 independent SDPs, and parallelization does not reduce the total computational work. The complexity statement should be restated honestly; this does not affect the mathematical validity of the bounds.","section":"Main text, paragraph after Theorem 2"},{"comment":"The printed s vector in the n=4 example sums to 1.0001 rather than 1; the rounding should be cleaned up or the numerical precision stated.","section":"Supplemental Material, Remark 2"},{"comment":"The simplification ~R_p(f,g)={(f(p),y)| g(t)≤y≤g(r)} is presented as an equality, but it implicitly assumes that every intermediate uncertainty value between g(t) and g(r) is achieved by some q with r≺q≺t. As written this is an outer approximation; a short justification or an explicit inclusion sign would be clearer.","section":"Main text, Eq. (12)"},{"comment":"The phrase 'two independent and identically distributed resources ρ' is potentially confusing; it should be 'two independent copies of the same state ρ', since the derivation assumes the pre-test and post-test are performed on fresh copies rather than sequentially on one physical system.","section":"Main text, opening of the framework"},{"comment":"In the contradiction argument, the proof writes inequalities for '∀l∈{1,...,i−1}' and '∀l∈{j,...,n}' even though l was fixed; this quantifier sloppiness should be corrected for readability.","section":"Supplemental Material, proof of Theorem 2"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is the first real step beyond state-independent UUR and DSMUR. Conditioning on the pre-test distribution p and using SDPs to get tight majorization bounds r and t on the post-test q is a genuinely new idea, and the framework is useful beyond any single entropy measure. The lower bound r is solid, the universal uncertainty region construction is natural, and the qubit examples show the method beating Maassen-Uffink. Worth engaging seriously.\n\nThe genuinely new content is Theorem 1's SDP characterization and Theorem 2's claim that r and t are the lattice infimum and supremum of the feasible post-test set. The proof that r is concave and a valid lower bound is correct. The applications to outer-approximating uncertainty regions and bounding joint uncertainties follow cleanly once such bounds exist. I also appreciate that nothing is fitted: the bounds come from exact SDP optimizations, and the lattice-completeness argument is imported correctly from the majorization literature.\n\nThe main soft spot is the construction of t. The paper defines t by a single averaging step (Eq. 7). The flatness process in the majorization lattice literature is iterative: after averaging one block, a new violation can appear. The stress-test example S = (0.4, 0.1, 0.2, 0.3, 0) illustrates this directly — the proposed single-pass t is not non-increasing, so it is not even in P_n. The proof of Theorem 2 in the supplemental material relies on properties of this one-step t, so as written the optimality argument does not go through. The existence of a true supremum is not in doubt, because the majorization lattice is complete, but the claimed explicit construction needs either the full iterative algorithm or a proof that the specific s_k values from the SDPs cannot produce a second violation. This is likely fixable by citing the standard flatness process properly, but it is a real gap in the current version.\n\nMinor issues: the efficiency claim overreaches, since each s_k requires solving C(n,k) SDPs, which is exponential in the middle. Parallelism does not change the total complexity. The concluding claim that the method extends to POVMs without modification is unsupported. The sentence about unpublished photonic data fitting well is not citable evidence and should be removed or replaced. The i.i.d. copies assumption is stated explicitly and is fine; the bounds do not apply to sequential measurements on the same physical copy, and the paper does not claim otherwise.\n\nThis paper is for researchers in majorization-based uncertainty relations, entropic uncertainty regions, and resource theories with partial knowledge of the state. It deserves a serious referee. With the flatness construction patched and the complexity statement corrected, it would be a solid, publishable contribution. Recommendation: send to peer review, and ask the authors to fix the construction of t.","headline":"A genuinely new conditional majorization approach to uncertainty regions, but the one-step flatness formula for the upper bound does not obviously produce a valid probability vector and needs to be fixed.","tokens_in":27737,"tokens_out":6000,"would_cite":true,"duration_ms":547571,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P45","81P15"],"pacs":["03.67.-a"],"model":"deepseek-v4-flash","headline":"The information gained from the first of two incompatible measurements fixes the tightest possible majorization bounds on the outcome distribution of the second, and those bounds can be computed by semidefinite programming.","keywords":["complementary information principle","majorization","uncertainty relations","semidefinite programming","quantum measurements","uncertainty regions","joint uncertainty","information causality"],"falsifier":"Concretely, prepare a known state ρ, compute p from the Born rule, solve the two SDP families to obtain r and t, then measure N on many fresh copies and estimate the empirical post-test vector q; if the empirical Lorenz curve ever falls outside the envelope r≺q≺t beyond sampling error, the claimed bounds are wrong for that scenario.","tokens_in":26696,"feed_emoji":"⚛️","tokens_out":12197,"duration_ms":117605,"temperature":0.7,"pith_summary":"The paper aims to show that complementarity between two incompatible measurements can be stated directly on the outcome probability vectors, without selecting an entropy or other scalar uncertainty measure. The claim is that once a first measurement returns a fixed outcome distribution p, every quantum state compatible with p produces a second-measurement outcome q that lies between two unique probability vectors r and t in the majorization order, and these are the tightest possible bounds. The paper proves that r and t can be computed efficiently with semidefinite programs, and uses them to build explicit outer approximations for the full uncertainty region of any reasonable uncertainty measure. If correct, this supplies a measure-independent benchmark for uncertainty relations and fundamental limits for all joint uncertainty measures.","feed_headline":"Pre-test outcome fixes the tightest bounds on post-test uncertainty","feed_subtitle":"A semidefinite program finds the exact probability envelope, with no entropy measure chosen","key_machinery":"Majorization is the ordering that carries the argument: for probability vectors x and y, x≺y when the sum of the k largest entries of x does not exceed that of y for every k, equivalently the Lorenz curve of x lies below that of y. Comparison is done via marginal majorization, which fixes the pre-test vector p and orders post-test outcomes by majorization. The set of n-outcome probability vectors under this order is a complete lattice—every subset has a unique infimum and supremum—which is why r and t exist. The semidefinite programs compute, for each k, the smallest and largest possible value of the sum of the k largest post-test probabilities over all states satisfying the pre-test Born-rule constraints; these produce the lower envelope r and a raw upper envelope s. Because s need not be concave, the flatness process replaces it by the least concave curve above it, giving the optimal upper bound t.","core_discovery":"The central discovery is Theorem 2: for a fixed pre-test measurement M and outcome distribution p, the set Q of post-test probability vectors has a unique greatest lower bound r and unique least upper bound t under majorization, so every q in Q satisfies r≺q≺t. Geometric language makes this vivid: the Lorenz curve of every admissible post-test distribution lies between the Lorenz curves of r and t, and the two curves form the tightest envelope allowed by the pre-test data. The lower bound r comes from minimizing, over all states compatible with p, the maximal sum of the k largest post-test probabilities for every k, while the upper bound t is obtained by taking the corresponding maxima and then applying a flatness procedure that restores the concavity a Lorenz curve must have. The paper presents this result as a complementary information principle: the classical information gained in the pre-test confines the uncertainty of the post-test in the tightest possible way, generalizing Heisenberg's complementarity.","pith_inferences":["A natural next step, not taken in the paper, is to use the gap between r and t as a quantitative measure of remaining complementarity and to study how it shrinks as the pre-test measurement becomes more informative; this could sharpen information-causality statements beyond the two extreme cases the paper discusses.","Because the bounds are computed by semidefinite programs, they can serve as a common benchmark for the many published uncertainty relations: any candidate relation is at least as loose as the projection of [r,t], so the projection gives a fair way to compare different entropy-based bounds on one scale.","The flatness process shows the least upper bound is generally not attained by a physical state; characterizing the gap between t and the achievable set Q, and identifying the states that come closest to it, is an open convex-geometric question."],"forward_implications":["For any pair of rank-one projective measurements and any pre-test outcome p, the post-test distribution of every compatible state lies in the majorization interval [r,t], and this interval is computable by parallel semidefinite programs.","The union of these intervals over all p yields an explicit outer approximation of the full uncertainty region for every non-negative Schur-concave uncertainty measure, and for qubits the approximation coincides with the true region.","Every joint uncertainty measure that is non-negative and monotone under doubly stochastic relabeling has state-independent lower and upper bounds given by the paper's formulas, tight for qubits.","The bounds extend to multiple post-test measurements and to POVMs, and give sufficient conditions for resource-theoretic state transformations, such as entanglement and coherence, when only partial information about the state is available."],"supporting_citations":[{"why":"Establishes that majorization supplies a measure-independent notion of uncertainty for a probability vector, motivating the central use of the order in the complementary information relation.","marker":"[13]"},{"why":"Provides the Maassen–Uffink entropic uncertainty relation that the universal uncertainty region is compared against and improved in the qubit case.","marker":"[40]"},{"why":"Defines the postulates for joint uncertainty measures, making the fundamental-limit bounds well-posed.","marker":"[63]"},{"why":"Supplies the completeness of the majorization lattice, which guarantees the unique infimum and supremum r and t exist for every subset of probability vectors.","marker":"[103–105]"},{"why":"Gives the flatness process that converts the raw upper envelope s into the least concave Lorenz curve above it, producing the optimal upper bound t.","marker":"[106]"},{"why":"Hardy–Littlewood–Pólya equivalence between majorization and doubly stochastic matrices is used to pass from the majorization bounds to bounds on every monotone joint uncertainty measure.","marker":"[109]"}],"fun_headline_variants":["Majorization envelope: pre-test data sets tightest post-test bounds","Complementary information principle yields exact uncertainty bounds","Pre-test outcomes pin down the full post-test probability envelope","Pre-test data fixes the exact envelope for post-test uncertainty"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument assumes the black box produces independent, identically distributed copies of one quantum state and that the pre-test and post-test are performed on fresh copies, so the first measurement never disturbs the system that is later measured.","fun_headline_variants_meta":{"raw":{"variants":["Majorization envelope: pre-test data sets tightest post-test bounds","Complementary information principle yields exact uncertainty bounds","Pre-test outcomes pin down the full post-test probability envelope","Pre-test data fixes the exact envelope for post-test uncertainty"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000431,"raw_usage":{"total_tokens":2179,"prompt_tokens":905,"completion_tokens":1274,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":521,"completion_tokens_details":{"reasoning_tokens":1208}},"tokens_in":521,"tokens_out":1274,"duration_ms":10046,"temperature":1.0,"reasoning_tokens":1208,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:00:28.499448+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Concretely, prepare a known state ρ, compute p from the Born rule, solve the two SDP families to obtain r and t, then measure N on many fresh copies and estimate the empirical post-test vector q; if the empirical Lorenz curve ever falls outside the envelope r≺q≺t beyond sampling error, the claimed bounds are wrong for that scenario.","supporting_citations":[{"cited_title":"Operational foundations of complementarity and uncertainty relations","cited_arxiv_id":"1809.03475","evidence_quote":"Defines the postulates for joint uncertainty measures, making the fundamental-limit bounds well-posed."},{"cited_title":"Bengtsson and K","cited_arxiv_id":null,"evidence_quote":"Hardy–Littlewood–Pólya equivalence between majorization and doubly stochastic matrices is used to pass from the majorization bounds to bounds on every monotone joint uncertainty measure."}],"review_version":1}