{"id":"63ee0c37-afc6-49fd-84f9-2d828e5f6426","arxiv_id":"2411.18036","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A quasi U-tile structure is introduced as a sufficient condition for strongly uncompletable product bases, yielding UPBs that are SUCPBs in one, two, or all three bipartitions of tripartite systems.","lead":"This paper constructs new examples of quantum product bases that cannot be completed or extended in some, but not all, ways of splitting a tripartite system. It also builds a smaller example that remains uncompletable under every bipartition, improving a known construction.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 1 rests on an unproved support-classification: rank-one, zero-sum tile matrices are asserted, not shown, to be confined to one new tile; all main SUCPB theorems inherit this gap.","rationale":"The reader's weakest assumption identifies exactly the decisive unproved step in Lemma 1, so I agree with the conditional verdict. The explicit examples, especially the d=6 table, provide real evidence that the constructions are nonempty, but they do not establish the general classification of rank-one, zero-sum tile matrices that Lemma 1 needs. The concern is a correctness risk rather than a style issue: it is addressable by supplying the missing support-classification lemma or by a computational verification on the concrete families. Until that is done, the SUCPB claims in Theorems 1, 3, and 4, including the headline d^3−3d^2+1 construction, rest on an unproved structural assertion. The verdict should remain CONDITIONAL.","tokens_in":17878,"tokens_out":20446,"duration_ms":192381,"concrete_test":"For the d=6 instance of U^4_A|BC from Theorem 4 (Eq. (16) and Fig. 11), enumerate all subsets S of the old tiles whose union is a rectangle. For each S, write M=uv^T with u_r v_c constant on each tile, impose Sum(M)=0, and solve symbolically. If any nonzero solution uses tiles from at least two of the l_j groups defined in Theorem 4's proof, Lemma 1's 'only m cases' assertion is false and Theorem 4's all-bipartition SUCPB claim collapses. If no such solution exists, repeat the enumeration on Fig. 2 and on the Theorem 1 partitions to confirm the classification is repairable.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Lemma 1 asserts, just before Eq. (5), that if Sum(M)=0 and rank(M)=1 then 'there are only m cases', meaning any product state in H⊥_US has support inside a single l_j. This is the only step that bounds the span of all product states in the complement by n−m<n−1, and it is exactly what makes U_S an SUCPB. The quasi U-tile conditions (i)–(iii) do not transparently imply this classification. A rank-one matrix has support equal to a product of a row set and a column set. Since old tiles are grouped into l_j by shared row or column indices, such a support can cut through several l_j, using only some tiles of each, without containing any l_j as a full rectangle. Condition (i) only forbids extending an entire l_j to a larger tile, not extending a proper subrectangle of it with tiles from another group, and condition (iii) only constrains the coarse structure T_U. The proof gives no argument excluding partial multi-l_j supports. If such a support exists, the dimension count Dim(O_1+...+O_m)=n−m is invalid, and the SUCPB conclusions of Theorems 1, 3, and 4 inherit the gap.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a geometric sufficient condition, called a quasi U-tile structure, for constructing strongly uncompletable product bases (SUCPBs). Lemma 1 claims that any bipartite tile structure satisfying this condition yields an SUCPB of size d1d2 - n + 1. The authors then apply this lemma to tripartite systems: they construct UPBs in C^d ⊗ C^d ⊗ C^2 that are SUCPBs in two bipartitions, UPBs in C^d ⊗ C^2 ⊗ C^2 that are SUCPBs in at most one bipartition, and a UPB in C^d ⊗ C^d ⊗ C^d of size d^3 - 3d^2 + 1 that is an SUCPB in every bipartition. The paper's main advertised achievement is this smaller-cardinality all-bipartition SUCPB, together with a systematic survey of the possible numbers of bipartitions in which a UPB can be an SUCPB.","tokens_in":18164,"tokens_out":15015,"duration_ms":127211,"significance":"If Lemma 1 is correct, the quasi U-tile criterion is a useful and genuinely new tool for constructing SUCPBs, and the reported examples answer the open question about existence of UPBs that are SUCPBs in only one or two bipartitions. The explicit constructions are nontrivial and the cardinality improvement from the previously known all-bipartition SUCPB to d^3 - 3d^2 + 1 is a concrete contribution. The paper does not fit parameters and does not assume its main conclusion as an input; it builds on the established U-tile characterization of UPBs from Ref. 28 and the SUCPB criterion from Ref. 29, which is a reasonable foundation if those results are correctly invoked. However, the central lemma's decisive step is presently asserted rather than proved, and several later verifications are delegated to visual inspection of figures, so the significance of the paper depends on closing these gaps.","major_comments":[{"comment":"The load-bearing step of Lemma 1 is the sentence 'From the conditions (i)-(iii), if Sum(M)=0 and rank(M)=1, there are only m cases for the matrix M.' This classification is not proved, and it is not a direct consequence of the stated conditions. A rank-one matrix has support equal to a product of a row set and a column set; such a support can intersect several groups l_j while containing none of them as a full rectangle. Condition (i) only forbids extending an entire l_j to a larger tile, and condition (iii) only constrains the coarse tile structure T_U. Neither condition, as written, rules out a support that uses only some tiles from each of several l_j and is closed under the row-column Cartesian product. Without Eq. (5), the dimension count Dim(O_1+...+O_m)=n-m is unsupported, and the conclusion Dim(H⊥_US) ≤ n-5, hence the SUCPB property, does not follow. The same gap is inherited by Theorems 1, 3, 4 and Proposition 3, which all invoke Lemma 1. Please provide a complete proof of the support classification, or add and justify an explicit extra condition on the quasi U-tile structure that makes it true.","section":"Lemma 1, proof, Eq. (5)"},{"comment":"The proofs that the new tile structures ∪ l_j are U-tile structures in each bipartition are asserted rather than demonstrated. For Proposition 3 this is a visual check in Fig. 11, and for Theorem 4 it is stated as 'they form a U-tile structure' without a general combinatorial argument. To apply Lemma 1, condition (iii) must be verified for all d, which requires showing that no special rectangle in the new tile structure can be decomposed into two smaller special rectangles or tiles. The current text does not supply such a verification, and the analogy with the d=6 case is not a proof for general d. This point is load-bearing for the claim that the constructed sets are SUCPBs in every bipartition.","section":"Section V, proofs of Proposition 3 and Theorem 4"},{"comment":"The proof of Theorem 2 states that the four states |ψ1⟩,...,|ψ4⟩ are orthogonal product states in H⊥_{U2_A|BC}, but it does not verify their orthogonality or their containment in the complementary subspace for general d. Since the conclusion that U2 is not an SUCPB in any bipartition rests on these four states spanning a four-dimensional complement, this verification should be given explicitly. The same applies to the analogous four-state argument in Example 2, where the states are listed without a check that they are mutually orthogonal and orthogonal to all members of the UPB.","section":"Theorem 2, proof and general construction U2"}],"minor_comments":[{"comment":"There are several typos: 'sized3 −3d2 +1' should read 'size d^3 − 3d^2 + 1', and 'bipartions' should be 'bipartitions'.","section":"Abstract and Introduction"},{"comment":"The sentence 'In Ref. ??, Shi et al. proposed a UPB of size 200...' contains a missing reference placeholder. This must be filled in before publication.","section":"Section V, after Proposition 3"},{"comment":"The displayed formula for A5 has a typographical error: it reads '|β_j⟩_A |η_s⟩_V |1⟩_C}' with a subscript V and an extra brace. It should presumably be '|β_j⟩_A |η_s⟩_B |1⟩_C'.","section":"Eq. (14), general construction for Theorem 3"},{"comment":"The definition of SUCPB is awkward: 'for all Hext' is unclear. The intended meaning appears to be that the incompletability holds for every bipartition of the multipartite system; the text should say this explicitly.","section":"Section II, definition of SUCPB"},{"comment":"In the proof of Theorem 4, the index set for k and i is written as Z6 even though the statement is for general d ≥ 6. This should be Z_d.","section":"Theorem 4, proof"},{"comment":"The text says the 4 × 4 tile structure T^{V2}_{A|BC} is shown in Fig. 7, but the figure labeled Fig. 8 is the one displaying a 4 × 4 tile structure; the references to the figures are inconsistent.","section":"Example 2 and Fig. 7/Fig. 8"}],"recommendation":"major_revision","confidential_remarks":"The central gap in Lemma 1 is serious enough that the paper should not be accepted in its current form, but the overall approach is plausible and the constructions are explicit, so a major revision with a complete proof of the support classification could make the paper publishable. The heavy reliance on Refs. 28 and 29 is acceptable if correctly applied, but the authors should double-check that the multipartite U-tile correspondence is indeed proved in those references and state it precisely. The missing 'Ref. ??' placeholder suggests the manuscript was not fully polished, and the proofs would benefit from replacing figure-based verification of U-tile properties with explicit combinatorial arguments."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a real construction paper, not a repackaging. It gives the first UPBs that are SUCPBs in exactly one or exactly two bipartitions, and it brings the all-bipartition SUCPB UPB in C6^3 down from 200 to 109 states (general size d^3 - 3d^2 + 1). The quasi U-tile sufficient condition is a reasonable geometric tool, and the explicit tables, especially the C6 table, are concrete and checkable.\n\nThe main soft spot is Lemma 1. The lemma asserts that for a rank-one matrix M with Sum(M)=0 built on a quasi U-tile structure, the support must lie inside one of the new tiles l_j. That is the only statement that bounds the span of product states in the complement by n-m < n-1, so all the SUCPB conclusions in Theorems 1, 3, and 4 rest on it. The proof just says 'from conditions (i)-(iii) there are only m cases'. I don't see that as immediate: a rank-one support is a row times a column, and it can cut through several l_j using parts of each, and nothing in (i) or (iii) transparently rules that out. This needs a real argument or a corrected statement. It may well be true; the small examples are consistent with it. But as written it's a genuine gap, not a typo.\n\nAlso: the proofs lean heavily on figures to assert that the new tile structures are U-tile structures. That is standard in this area, but should be backed by a readable induction or a checkable pattern for the general d. There are the usual typos (e.g., 'UPB sized3' in the abstract, a broken \\cite{??} in the proof of Proposition 3, and the odd 'dimension d of the space is even' phrasing in the introduction). Those are minor.\n\nThe self-citation load is real but not abusive: the paper leans on the authors' earlier U-tile theorem and SUCPB criterion without proof, and both are published; that's acceptable in a construction paper.\n\nWho should read it: people working on UPBs, bound entanglement, and strong nonlocality. It gives explicit examples that are useful even before the general lemma is fixed. If Lemma 1 gets a rigorous proof, this is a solid contribution. Worth sending to a serious referee now, with the request to scrutinize Lemma 1. I'd accept it conditionally.","headline":"A useful construction paper that likely answers the one/two-bipartition SUCPB question and shrinks the known all-bipartition example, but the proof of Lemma 1 skips the one step that makes the whole thing work.","tokens_in":18714,"tokens_out":2071,"would_cite":true,"duration_ms":18334,"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":"The paper proves that a 'quasi U-tile' tile structure is a sufficient condition for an orthogonal product set to be a strongly uncompletable product basis, and uses it to construct a UPB of size $d^3-3d^2+1$ in…","keywords":["unextendible product basis","strongly uncompletable product basis","quasi U-tile structure","U-tile structure","tripartite quantum systems","product basis","bound entanglement","tile structure"],"falsifier":"Find a quasi U-tile structure and a solution of the zero-sum, rank-one conditions on its matrix $M$ whose nonzero entries touch two different new tiles. The simplest place to look is the paper's $3\\times 5$ example with six tiles, where solving the equations for the six coefficients $a_1,\\dots,a_6$ would either confirm or refute the 'only $m$ cases' assertion; a counterexample would directly falsify Lemma 1 and the SUCPB claims built on it.","tokens_in":17682,"feed_emoji":"🧩","tokens_out":12049,"duration_ms":95450,"temperature":0.7,"pith_summary":"The paper gives a geometric sufficient condition for a set of orthogonal product states to be a strongly uncompletable product basis (SUCPB): the underlying tile structure must be a quasi U-tile structure. With this condition the authors settle the trichotomy question left open since 2003: in tripartite systems a UPB may be SUCPB in two bipartitions, in at most one, or in every bipartition. They build examples of each, and the headline construction is a UPB in $\\mathbb{C}^d\\otimes\\mathbb{C}^d\\otimes\\mathbb{C}^d$ with $d^3-3d^2+1$ states that is SUCPB in every bipartition, a smaller cardinality than the previous known example. The proof machinery is Lemma 1, which shows that in a quasi U-tile structure any product state orthogonal to the stopper state is forced to lie in one of $m$ mutually orthogonal blocks, so the product states left in the complement cannot span it.","feed_headline":"109-state product basis uncompletable in every cut","feed_subtitle":"A quasi U-tile rule constructs bases that resist completion in each bipartition, improving the old 200-state bound.","key_machinery":"The load-bearing object is the quasi U-tile structure: a tile structure whose tiles can be grouped into at least five new tiles, each new tile formed by old tiles sharing a row or column index, each new tile non-extendable except to the whole structure, and the new-tile structure itself a U-tile structure. Lemma 1 attaches to each tile a family of Fourier-type product states and adds the uniform stopper state $|S\\rangle$; any product state in the complement corresponds to a rank-one matrix with zero sum of entries. The quasi U-tile conditions are used to force that matrix's support into a single new tile, and because the new tiles are mutually orthogonal, all product states in the complement lie in a space of dimension at most $n-m\\le n-5$, strictly below the complement dimension $n-1$. This dimension shortfall is exactly the defining property of an SUCPB.","core_discovery":"The central claim is that a $d_1\\times d_2$ tile structure is a quasi U-tile structure—its $n$ tiles can be partitioned into $m\\ge 5$ new tiles, each new tile a union of old tiles sharing a common row or column index, none extendable except to the whole structure, and the new tiles themselves forming a U-tile structure—then the associated orthogonal product set is an SUCPB of size $d_1d_2-n+1$. In the tripartite setting the paper claims that by viewing each of the three bipartitions separately, the quasi U-tile criterion yields UPBs that are SUCPBs in two bipartitions, in at most one, and in every bipartition, completing the classification. The main object is the UPB $U^4$ in $\\mathbb{C}^d\\otimes\\mathbb{C}^d\\otimes\\mathbb{C}^d$ of size $d^3-3d^2+1$, built from vertical, horizontal, and crossed tile states plus the uniform stopper state, which is SUCPB in every bipartition for $d\\ge 6$.","pith_inferences":["The quasi U-tile criterion is formulated for rectangular tiles, but the same rank-one/zero-sum mechanism should extend to cube-like tiles in N-partite systems with $N\\ge 4$, producing SUCPBs in every bipartition by applying the argument cut by cut.","The new construction requires $d\\ge 6$, so the minimal size of a UPB that is SUCPB in every bipartition in dimensions $d=3,4,5$ remains open; the bound $d^3-3d^2+1$ suggests the true scaling is $d^3-O(d^2)$, and combinatorial searches over tile partitions could test smaller cases.","The quasi U-tile recipe decouples the SUCPB property from the explicit vector details: any tile partition satisfying the three geometric conditions yields the same dimension shortfall, so one can search for new SUCPBs by purely combinatorial data on tiles."],"forward_implications":["For odd $d\\ge 3$, the set $U^1$ in $\\mathbb{C}^d\\otimes\\mathbb{C}^d\\otimes\\mathbb{C}^2$ is a UPB of size $2d^2-4d+4$ that is SUCPB in the two bipartitions pairing the qubit with one qudit; for even $d\\ge 4$, $U^{1\\prime}$ has size $2d^2-4d+8$ and the same property.","In $\\mathbb{C}^d\\otimes\\mathbb{C}^2\\otimes\\mathbb{C}^2$, the sets $U^2$ (size $4d-4$) and $U^3$ (size $4d-7$) show that UPBs that are SUCPB in no bipartition and in exactly one bipartition both exist, so all three logical cases occur.","The main construction $U^4$ is a UPB of size $d^3-3d^2+1$ in $\\mathbb{C}^d\\otimes\\mathbb{C}^d\\otimes\\mathbb{C}^d$ for $d\\ge 6$ and is SUCPB in every one of the three bipartitions; for $d=6$ this yields 109 states, fewer than the previous 200-state example.","Any orthogonal product set whose tile structure is quasi U-tile is automatically SUCPB by Lemma 1, giving a reusable sufficient condition that avoids checking complement completions case by case."],"supporting_citations":[{"why":"Introduces UPB, UCPB, and SUCPB definitions, the GenTiles constructions, and the open problem about every bipartition.","marker":"[27]"},{"why":"Establishes that a tile structure gives a UPB exactly when it is a U-tile structure, the structural fact the quasi U-tile definition builds on.","marker":"[28]"},{"why":"Provides the criterion that an OPS is SUCPB when product states in its complement cannot span the complement, and supplies the previous UPB that is SUCPB in every bipartition.","marker":"[29]"},{"why":"Supplies the TILES UPB in $\\mathbb{C}^3\\otimes\\mathbb{C}^3\\otimes\\mathbb{C}^2$ that the two-bipartition construction generalizes.","marker":"[30]"}],"fun_headline_variants":["Quasi U-tile trick yields uncompletable bases in every cut","Smaller UPB resists completion in all tripartite cuts","New tile rule builds bases that block completion in every split","Quasi U-tile: a route to strongly uncompletable product bases"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that a product state orthogonal to the constructed basis can only have nonzero coefficients inside one of the $m$ new tiles; if this classification is false, the dimension count that makes the basis strongly uncompletable collapses.","fun_headline_variants_meta":{"raw":{"variants":["Quasi U-tile trick yields uncompletable bases in every cut","Smaller UPB resists completion in all tripartite cuts","New tile rule builds bases that block completion in every split","Quasi U-tile: a route to strongly uncompletable product bases"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000601,"raw_usage":{"total_tokens":2874,"prompt_tokens":1077,"completion_tokens":1797,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":693,"completion_tokens_details":{"reasoning_tokens":1719}},"tokens_in":693,"tokens_out":1797,"duration_ms":11687,"temperature":1.0,"reasoning_tokens":1719,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T11:35:46.226734+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Find a quasi U-tile structure and a solution of the zero-sum, rank-one conditions on its matrix $M$ whose nonzero entries touch two different new tiles. The simplest place to look is the paper's $3\\times 5$ example with six tiles, where solving the equations for the six coefficients $a_1,\\dots,a_6$ would either confirm or refute the 'only $m$ cases' assertion; a counterexample would directly falsify Lemma 1 and the SUCPB claims built on it.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces UPB, UCPB, and SUCPB definitions, the GenTiles constructions, and the open problem about every bipartition."},{"cited_title":"Shi , author X","cited_arxiv_id":null,"evidence_quote":"Establishes that a tile structure gives a UPB exactly when it is a U-tile structure, the structural fact the quasi U-tile definition builds on."},{"cited_title":"Shi , author M","cited_arxiv_id":null,"evidence_quote":"Provides the criterion that an OPS is SUCPB when product states in its complement cannot span the complement, and supplies the previous UPB that is SUCPB in every bipartition."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the TILES UPB in $\\mathbb{C}^3\\otimes\\mathbb{C}^3\\otimes\\mathbb{C}^2$ that the two-bipartition construction generalizes."}],"review_version":1}