{"id":"5aef7c98-5527-4586-97f1-f1d49dd8cf3b","arxiv_id":"2608.04354","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Every finite stochastic matrix is a marginal of a larger unistochastic matrix, so classical Markov transitions can be represented by unitary quantum evolution with a fixed ancilla.","lead":"This paper proves that any finite stochastic matrix, a table of probabilities used for Markov chains, can be written as a sum of squared entries of a larger unitary matrix. The result connects classical random processes to the probability calculus of quantum mechanics, and gives an explicit recipe for embedding one into the other.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central construction is valid; the only load-bearing proof defect is a missing complex conjugate in Eq. (51), which breaks the orthonormality calculation for complex Θ but is trivially fixable.","rationale":"The paper's central claim—that every finite column-stochastic matrix is the marginal of a larger unistochastic matrix—is correct. I verified the construction step by step: the embedding into a square stochastic matrix, the Kraus decomposition, the orthonormality of the N columns (up to the conjugate typo), the Gram-Schmidt completion, and the marginalization formula all work. The Born Representation Theorem is essentially a one-line construction and is also correct. The reader's CONDITIONAL verdict is appropriate: the theorem is sound but the novelty framing overstates prior work (the paper itself cites Gopalkrishnan 2026 and Schmidt 2021), and the application section contains a wrong Collatz example. The only technical proof defect I find that touches the central argument is the missing complex conjugate in Eq. (51). It is a one-character typo and does not threaten the theorem's truth—the columns are orthonormal regardless, because they have disjoint support in different k-blocks and their norms are 1 by column normalization—but as printed the proof is not valid for the complex Θ that Eq. (39) explicitly allows. Since this is a minor, easily repaired error, I do not move the verdict; UNCHANGED (CONDITIONAL) is right. The concrete test above settles the issue.","tokens_in":8404,"tokens_out":25614,"duration_ms":233403,"concrete_test":"Verify the corrected Eq. (51): replace sum_{i,k} K_{k,im} K_{k,ij} by sum_{i,k} overline{K_{k,im}} K_{k,ij}, and check that it equals sum_k (K_k†K_k)_{mj} = δ_{mj}. To see the printed version fail, take N=2, Θ = [[i/√2, 1/√2],[1/√2, -i/√2]] (which satisfies column normalization), form K_k = Θ P_k, and evaluate both expressions for (m,j) = (1,1): the printed expression gives sum_i Θ_{i1}^2 = i^2/2 + 1/2 = 0, not 1, while the correct expression gives 1.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In Theorem 3.3, the N columns of U defined by Eq. (50) must be orthonormal so that Gram-Schmidt can extend them to a unitary. Eq. (51) attempts to prove this by computing (U†U)_{(m1),(j1)} = sum_{i,k} K_{k,im} K_{k,ij}. This omits the complex conjugate: since (U†)_{(m1),(ik)} = overline{(U)_{(ik),(m1)}} = overline{K_{k,im}}, the product should be overline{K_{k,im}} K_{k,ij}. Without the conjugate, the claimed equality to sum_k (K_k†K_k)_{mj} is not valid when Θ (Eq. 39) is complex, as the paper explicitly permits. The error is confined to the displayed calculation: the columns are in fact orthonormal, because column (m,1) has support only in the k=m block while column (j,1) has support only in the k=j block, and the norm of column (j,1) is sum_i |Θ_{ij}|^2 = 1 by the column-stochastic normalization (41). Thus the theorem's conclusion survives, but the proof as printed contains a genuine gap at its key step. The fix is to add the conjugate in Eq. (51), or equivalently to choose the real square root Θ_{ij} = sqrt(\\tilde Γ_{ij}), which is always available.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves two structural theorems connecting stochastic matrices to quantum probability formulas. Theorem 3.1 (the Born Representation Theorem) states that every M x N stochastic matrix Gamma can be written entrywise as Gamma_ij = tr(E_i P_j), where {E_i} is a POVM and {P_j} is a PVM. Corollary 3.2 shows that if the POVM is also a PVM, then Gamma is unistochastic. Theorem 3.3 (the Unistochastic Theorem) constructs, for any M x L stochastic matrix Gamma and N = max(M,L), an N^2 x N^2 unitary U and two PVMs such that the matrix GammaGamma with entries GammaGamma_{(ik),(jl)} = Tr(P'_{(ik)} P_{(jl)}) = |U_{(ik),(jl)}|^2 is unistochastic and Gamma_ij = sum_k GammaGamma_{(ik),(j1)}, i.e. every finite stochastic matrix is a marginal of a unistochastic matrix of bounded dimension. The proof is constructive and self-contained, avoiding black-box use of Naimark or Stinespring dilation theorems, and the paper closes with applications to deterministic and probabilistic Markov chains.","tokens_in":8659,"tokens_out":6843,"duration_ms":70110,"significance":"If the results stand, the paper gives a clean and explicit sense in which arbitrary finite Markovian stochastic dynamics can be embedded into unitary quantum dynamics followed by a partial trace-like marginalization. The construction is elementary, the dimension bound N^2 is explicit, and the proofs are checkable step by step. The paper is not claiming new empirical predictions or a new physical theory; its value is structural, and it may be of interest to philosophers of physics and researchers working on stochastic-quantum correspondences. The self-contained nature of the proofs is a definite strength, as is the explicit marginalization formula in Eq. (35).","major_comments":[{"comment":"The displayed computation of (U†U)_{(m1),(j1)} omits a complex conjugation. Since (U†)_{(m1),(ik)} = overline{(U)_{(ik),(m1)}} = overline{K_{k,im}}, the middle expression should be sum_{i,k} overline{K_{k,im}} K_{k,ij}, not sum_{i,k} K_{k,im} K_{k,ij}. As printed, the equality to (K_k† K_k)_{mj} = delta_{mj} is valid only for real Theta, whereas Eq. (39) explicitly permits complex-valued entries. This is a genuine gap in the proof of orthonormality of the N columns defined by Eq. (50), which is load-bearing for the subsequent Gram-Schmidt extension to a unitary. The gap is easily repaired by inserting the conjugate, or equivalently by choosing the real square root Theta_{ij} = sqrt(tilde{Gamma}_{ij}), and the theorem's conclusion remains correct; the authors should implement one of these fixes in the final version.","section":"Section 3, Eq. (51)"}],"minor_comments":[{"comment":"The statement that unit-trace POVM elements imply Gamma is doubly stochastic should explicitly note that this condition forces M = N, since sum_i tr(E_i) = tr(1) = N.","section":"Section 3, after Eq. (22)"},{"comment":"The footnote asserts that Schmidt (2021) contains an error in its construction of a partial isometry analogous to Eq. (50). Since this is a specific claim about another paper, it would be helpful to state the nature of the error or soften the claim unless a detailed comparison is provided.","section":"Footnote 1"},{"comment":"The marginalization formula fixes the ancilla index in the second factor to 1. This is legitimate by construction, but the authors should make explicit that the representation is tied to this particular ancilla state; otherwise readers may assume a partial trace over an arbitrary ancilla state, which would not follow from the displayed formula.","section":"Section 3, Eq. (35)"},{"comment":"The example is clear, but the sentence 'one can, in principle, write down a 25 x 25 unitary matrix U' would benefit from a pointer to the explicit construction in Theorem 3.3, since the theorem already gives the algorithm for doing so.","section":"Section 4, Collatz example"}],"recommendation":"minor_revision","confidential_remarks":"The paper's central claim appears correct and the proof is repairable with a small, localized correction. The main reservation is the reliance on an unpublished MPhys project report for the dimension bound N^2 in Theorem 3.3; although the proof is self-contained, the attribution to Gopalkrishnan (2026) may need editorial scrutiny. The footnote about Schmidt (2021) should also be substantiated. Overall, the paper is suitable for publication after a minor revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper's central construction is valid, but the title oversells it. Theorem 3.3, the Unistochastic Theorem, is explicitly credited in a footnote to Gopalkrishnan (2026) for the N² bound and to Schmidt (2021) for a similar theorem; Theorem 3.1 is essentially a definitional identity, defining the POVM directly from the matrix entries. What the paper actually contributes is a self-contained, elementary proof of these known results, and that has real value for people who want the construction without hunting through project reports or flagged prior proofs.\n\nThe proof strategy is genuinely nice: pad the rectangular stochastic matrix to square, take any entrywise square root, build Kraus operators from it, and then assemble N orthonormal columns of a larger unitary from those Kraus matrices. The marginalization formula (35) is explicit and easy to verify. The writing is clear, and the author is honest about the prior literature, which I appreciate.\n\nThe soft spots are real but not load-bearing. First, Eq. (51) omits a complex conjugate in the inner product calculation. As printed, the orthonormality proof only goes through for real Θ. The conclusion is still true—the columns have disjoint block support and each has unit norm by column-stochasticity—so the fix is trivial, but it is a genuine gap in the displayed proof. Second, the Collatz example in Section 4 is internally inconsistent: the rule (3k+1) mod 5 sends 3 to 0, while the text and matrix (61) send 3 to 2. That is a sloppy application example, not a flaw in the theorem. Third, the repeated use of “new theorems” is too strong given the paper's own citations. The honest framing would be “self-contained proofs of known results, with a tightened bound.”\n\nWho gets value from this? Someone teaching or writing about dilations of stochastic matrices into unitary evolutions, or someone who wants a verified reference for the N² construction. It is not a breakthrough, but it is a useful exposition of a genuine result. I would not cite it in my own work when the original sources are available, but I would send it to a referee: the mathematics is checkable, the topic is of interest, and the remaining issues are exactly the kind a referee should catch and have fixed.\n\nSend it out, but with a referee who will demand the conjugate in Eq. (51) and a corrected Collatz example. After those fixes, it is a solid, if modest, contribution.","headline":"A clean, mostly expository proof that every stochastic matrix is a marginal of a unistochastic one; the main theorem is already in the cited literature, and the printed proof has a small fixable gap.","tokens_in":9233,"tokens_out":2544,"would_cite":false,"duration_ms":26486,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["15A51","81P16","47A20"],"pacs":[],"model":"deepseek-v4-flash","headline":"Every finite stochastic matrix is the marginal of a larger unitary matrix.","keywords":["stochastic matrices","unistochastic matrices","Born rule","projection-valued measures","positive-operator-valued measures","Kraus operators","unitary dilation","Markov chains"],"falsifier":"Follow the paper's construction for a small matrix with irrational entries, e.g. the $2\\times 2$ doubly stochastic matrix $\\begin{pmatrix}1/3&2/3\\\\2/3&1/3\\end{pmatrix}$: choose $\\Theta$ entrywise, form $K_i=\\Theta P_i$, assemble the four columns $U_{(ik),(j1)}=K_{k,ij}$, extend to a $4\\times 4$ unitary by Gram-Schmidt, and check numerically whether $\\sum_k |U_{(ik),(j1)}|^2$ equals $\\Gamma_{ij}$ and whether $U^\\dagger U=\\mathbb{1}$. Any failure of unitarity or of the marginal sum would refute Theorem 3.3.","tokens_in":8159,"feed_emoji":"🎲","tokens_out":12363,"duration_ms":115424,"temperature":0.7,"pith_summary":"The paper aims to prove that stochastic matrices—the finite tables of transition probabilities used for Markov chains—are not merely classical objects. Its first theorem shows that every entry of any stochastic matrix can be written as $\\operatorname{tr}(E_i P_j)$, a trace of one matrix from a POVM with one from a PVM, which is the shape of the quantum probability formula. Its second, stronger theorem embeds any $M\\times L$ stochastic matrix $\\Gamma$ into an $N^2\\times N^2$ unistochastic matrix, $N=\\max(M,L)$, from which $\\Gamma$ is recovered by summing over one index. The constructive proof means any finite Markov chain can be represented as the coarse-grained measurement statistics of a unitary process on a Hilbert space of bounded dimension. If correct, this gives every finite classical stochastic dynamics a quantum-mechanical implementation without altering its transition probabilities.","feed_headline":"Every stochastic matrix is a margin of a larger unitary one","feed_subtitle":"The proof builds an explicit unitary dilation, so any Markov chain's transition probabilities match quantum measurements.","key_machinery":"The proof works by an explicit column construction. After embedding $\\Gamma$ in a square stochastic matrix $\\tilde\\Gamma$, the author chooses any entrywise square root $\\Theta$ with $\\tilde\\Gamma_{ij}=|\\Theta_{ij}|^2$; column normalization makes $\\Theta^\\dagger\\Theta$ diagonal with ones. Defining Kraus operators $K_i=\\Theta P_i$ for the diagonal PVM $P_i$, the identity $\\sum_i K_i^\\dagger K_i = \\mathbb{1}$ follows, and the components $K_{k,ij}$ become the entries of $N$ columns of the $N^2\\times N^2$ matrix $U$. Orthonormality of these columns is exactly the Kraus identity, so Gram-Schmidt extends them to a full unitary. The $N^2$-member PVMs $P_{(ij)}=P_i\\otimes P_j$ and $P'_{(ij)}=U^\\dagger P_{(ij)}U$ then produce the unistochastic matrix $|U_{(ik),(jl)}|^2$, and the sum over $k$ at $jl=(j1)$ returns $\\Gamma_{ij}$.","core_discovery":"Theorem 3.3 is the central result: for any $M\\times L$ stochastic matrix $\\Gamma$ with $N=\\max(M,L)$, there exists an $N^2\\times N^2$ unitary matrix $U$ and projection-valued measures $P_{(ij)}$ and $P'_{(ik)}=U^\\dagger P_{(ik)}U$ such that the array $\\Gamma\\Gamma_{(ik),(jl)} = \\operatorname{Tr}(P'_{(ik)}P_{(jl)}) = |U_{(ik),(jl)}|^2$ is unistochastic and $\\Gamma_{ij} = \\sum_{k=1}^N \\Gamma\\Gamma_{(ik),(j1)}$. The original stochastic matrix is therefore the marginal of a larger unitary object, with the ancillary index fixed at $1$. The paper first proves the weaker Born Representation Theorem, $\\Gamma_{ij}=\\operatorname{tr}(E_iP_j)$ with a POVM and a PVM, and derives as a corollary that when the POVM is itself a PVM the matrix is unistochastic, i.e. its entries are squared moduli of entries of a unitary matrix. Theorem 3.3 removes the doubly stochastic restriction by dilation, so arbitrary rectangular stochastic matrices become unistochastic in a larger dimension.","pith_inferences":["Because the proof fixes the ancilla in the state $|1\\rangle$, varying that ancilla state should produce a family of different stochastic matrices from the same unitary, so the embedding is highly non-unique and the ancilla preparation is a free resource not discussed in the paper.","The dilation dimension $N^2$ is an upper bound, not a proven optimum; testing small matrices such as $2\\times 2$ and $3\\times 3$ cases to see whether smaller unitary dilations exist could reveal whether the bound is tight.","The theorem suggests that any finite stochastic process can be quantized with bounded overhead, but because irreversible chains require projections at each step, the representation is a simulation rather than a hidden-variable model that would make irreversibility emerge from unitary evolution alone.","One natural next step, not taken here, is to ask whether the construction extends to time-inhomogeneous Markov chains or to continuous-time generators, where the structure of the Kraus operators might need different treatment."],"forward_implications":["Every finite-state Markov chain has an explicit unitary dilation of dimension at most $N^2$, so its transition probabilities can be reproduced by Born-rule measurements on a quantum system with a fixed ancilla state.","The Born Representation Theorem puts any finite conditional distribution into POVM/PVM trace form, so classical finite statistics obey a generalized quantum measurement rule.","Corollary 3.2 implies that if the POVM side can be promoted to a PVM, the stochastic matrix is unistochastic and hence doubly stochastic; general rectangular stochastic matrices require the extra dilation.","For deterministic reversible processes, the constructed unitary admits a Hamiltonian and a continuous-time Schrödinger evolution interpolating the discrete dynamics; for irreversible or probabilistic chains, exact recovery requires inserting projections at integer times."],"supporting_citations":[{"why":"Supplies the construction of the N columns of U in Eq. (50), which tightens the dilation dimension from N^3 to N^2.","marker":"Gopalkrishnan (2026)"},{"why":"Names and supplies the resolution-of-identity condition that makes the assembled columns of U orthonormal and produces the Kraus decomposition used in the marginalization.","marker":"Kraus (1971)"},{"why":"Defines the quantum-mechanical probability rule whose generalized trace form the two theorems establish for stochastic matrices.","marker":"Born (1926)"},{"why":"The earlier result whose N^3 dilation bound the present theorem improves, serving as the baseline comparison for the stronger claim.","marker":"Barandes 2023, 2025"}],"fun_headline_variants":["All stochastic matrices are margins of unitaries","Stochastic matrices embed in larger unitary ones","Born rule covers all stochastic matrices","Every stochastic matrix is a unitary margin","Unitary dilation makes any stochastic matrix unistochastic"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument assumes that the matrix to be represented is genuinely stochastic—its entries are nonnegative and each column sums to one—because those two properties make it possible to take entrywise square roots and to normalize the columns of the unitary.","fun_headline_variants_meta":{"raw":{"variants":["All stochastic matrices are margins of unitaries","Stochastic matrices embed in larger unitary ones","Born rule covers all stochastic matrices","Every stochastic matrix is a unitary margin","Unitary dilation makes any stochastic matrix unistochastic"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000437,"raw_usage":{"total_tokens":2289,"prompt_tokens":1077,"completion_tokens":1212,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":693,"completion_tokens_details":{"reasoning_tokens":1147}},"tokens_in":693,"tokens_out":1212,"duration_ms":9663,"temperature":1.0,"reasoning_tokens":1147,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T19:09:21.347150+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Follow the paper's construction for a small matrix with irrational entries, e.g. the $2\\times 2$ doubly stochastic matrix $\\begin{pmatrix}1/3&2/3\\\\2/3&1/3\\end{pmatrix}$: choose $\\Theta$ entrywise, form $K_i=\\Theta P_i$, assemble the four columns $U_{(ik),(j1)}=K_{k,ij}$, extend to a $4\\times 4$ unitary by Gram-Schmidt, and check numerically whether $\\sum_k |U_{(ik),(j1)}|^2$ equals $\\Gamma_{ij}$ and whether $U^\\dagger U=\\mathbb{1}$. Any failure of unitarity or of the marginal sum would refute Theorem 3.3.","supporting_citations":[],"review_version":1}