{"id":"1eddb95f-f3f5-4021-ae01-fb4d44d14dab","arxiv_id":"2509.10949","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Any empirically-adequate, linearity-preserving complex-valued quasiprobability representation of a finite-dimensional, tomographically-local GPT decomposes as Q(T) = χ_B ∘ C(T) ∘ φ_A.","lead":"The paper proves that every complex-valued quasiprobability representation of a finite-dimensional, tomographically-local generalized probabilistic theory factors into two linear maps sandwiching the complexified process, with the two maps fixed by how the representation acts on states and on the identity. This gives a common structural form for Kirkwood-Dirac and other complex quasiprobability representations, extending an earlier real-valued result.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified; the structure theorem is sound under its stated assumptions.","rationale":"The reader's verdict of ACCEPT with moderate confidence is appropriate. I read the full text and checked the proof chain behind Corollary III.4: the decomposition of processes from Lemma II.1, the use of empirical adequacy on closed diagrams, the construction of χ_A and φ_A, and the injectivity argument via the left-inverse relation. No internal inconsistency or unsupported leap was found. The strongest claim in the reader's summary matches the theorem: a complex-valued quasiprobability representation is fixed by its action on states and on the identity, and in quantum theory this reduces to frames and dual frames. The weakest assumption identified by the reader, tomographic locality, is indeed explicit in the theorem statement and is the main limiting condition. However, a closer look suggests the proof may only require the single-system spanning properties of states and effects, which hold for every GPT in the paper's sense; if so, tomographic locality is stronger than necessary but still not a threat to the theorem as stated. The only other notable caveat is that Definition II.17 states linearity-preservation for arbitrary real linear combinations without spelling out the domain when a combination is not itself a process; this is a minor formal imprecision, easily resolved by working with the linear span, and it does not change the verdict. Therefore no adjustment is recommended.","tokens_in":36549,"tokens_out":46714,"duration_ms":441347,"concrete_test":"Re-derive the used direction of Lemma II.1 without invoking tomographic locality, using only the GPT property that states span A and effects span A^*; if the resolution of the identity (Eq. 68) and the left-inverse computation (Eq. 75) still go through, then the stated assumption is not load-bearing and the proof is confirmed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"No significant objection identified. The central factorization in Corollary III.4 is established by the proof: Lemma II.1 gives a real-coefficient resolution T = Σ r_ij s_i∘e_j; empirical adequacy fixes M on closed diagrams; semi-functoriality and linearity-preservation justify expanding M(T); the maps χ_A (Eq. 72) and φ_A (Eq. 74) are well-defined because states and effects span their spaces (Lemmas A.1 and A.2), and Eq. (75) proves φ_A∘χ_A = id, so χ_A is injective. For semi-functors, D_A = χ_A∘φ_A, and since D_A∘χ_A = χ_A the images coincide, giving φ_A = χ̄_A^{-1}∘D_A. No hidden circularity or missing step was found. The only genuinely restrictive assumption is tomographic locality, which the paper explicitly assumes; if it failed, Lemma II.1 as stated would not be available. In fact, the proof appears only to need single-system spanning of states and effects, so tomographic locality may be stronger than necessary, but this does not weaken the theorem as stated.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper proves a structure theorem for complex-valued quasiprobability representations of finite-dimensional, tomographically-local generalized probabilistic theories (GPTs). Working in the process-theoretic framework of Schmid et al., the authors define a complex-valued quasiprobability representation as an empirically adequate, linearity-preserving semi-functor Q from a GPT into the process theory FinQuasiSubStoch_C. The main result (Corollary III.4, with Theorems III.1 and III.3 for the functorial and semi-functorial cases respectively) states that any such representation factorizes as Q(T) = chi_B ∘ C(T) ∘ phi_A, where for each system A the map chi_A is an injective complex-linear map fixed by the action of Q on states, and phi_A = chi_A^{-1} ∘ Q(id_A) is fixed by the action of Q on the identity. In the functorial case this reduces to Q(T) = chi_B ∘ C(T) ∘ chi_A^{-1} with chi_A invertible. For quantum theory the authors show that the pair (chi_A, phi_A) corresponds to a choice of frame and dual frame, and that the representation of the identity is an idempotent matrix, reducing to the identity exactly in the functorial case. The proof is constructive and diagrammatic, building on a real-coefficient resolution of the identity supplied by tomographic locality (Lemma II.1) and on a detailed development of the complexification functor, including proofs that it is faithful and strong monoidal.","tokens_in":1283,"tokens_out":1373,"duration_ms":410718,"significance":"If the result holds, it is a significant structural characterization: the content of any complex-valued quasiprobability representation of a finite-dimensional tomographically-local GPT is fully contained in the representation's action on states and on the identity. This unifies previously disparate examples, including Kirkwood-Dirac representations, and extends the earlier structure theorem of Schmid et al. in three directions: complex-valued targets, semi-functorial rather than only functorial maps, and codomains that may be infinite-dimensional. The paper gives explicit constructive proofs of the auxiliary categorical facts (the complexification functor is a faithful strong monoidal functor, and complexification preserves generating sets of states and effects), which strengthens confidence in the main claims. I checked the central steps: the definition of chi_A and phi_A from resolutions of the identity, the left-inverse argument establishing injectivity, the right-inverse argument in the functor case, and the idempotent-splitting argument in the semi-functor case all cohere, and I found no circularity.","major_comments":[],"minor_comments":[{"comment":"There are typographical errors in proper names: 'Eugine Wigner' should be 'Eugene Wigner' and 'Kolmogovo's axioms' should be 'Kolmogorov's axioms'.","section":"Abstract and Section II.B"},{"comment":"The notation for the surjective corestriction is confusing: in the text both chi_A and its corestriction are written with the same symbol, although the PDF appears to use an overline for the corestriction. Please introduce overline-chi_A explicitly and write phi_A = (overline-chi_A)^{-1} ∘ N(id_A) consistently throughout the statements and proofs.","section":"Theorems III.3 and Corollary III.4"},{"comment":"Lemma II.1 is the load-bearing bridge from tomographic locality to the resolution of the identity used throughout the proof, but it is quoted without proof. Since the lemma carries so much weight, adding a proof sketch or a precise statement of the version proved in Ref. [26] would substantially improve the paper's self-containedness.","section":"Lemma II.1"},{"comment":"The sentence containing '|Λ| = d^2 > d' is slightly inaccurate for d = 1; it should be written as '|Λ| = d^2 (and |Λ| > d for d > 1)' or simply '|Λ| = d^2'.","section":"Section III.D"},{"comment":"The phrase 'acts as the composition of the standard embedding and the identity' is awkward; the intended meaning is simply M(p) = e_C(p) for every closed diagram p in G(I,I), and this could be stated more directly.","section":"Definition II.18"}],"recommendation":"minor_revision","confidential_remarks":"The manuscript relies heavily on the authors' own earlier framework, especially Ref. [26], but the new factorization theorem is obtained by direct construction from the representation itself, and the dependence is transparent rather than hidden. I see no reason to question novelty or fit with the journal's scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Punchline: this is a real extension, not a repackaging. Schmid et al. proved a structure theorem for real-valued, functorial, identity-preserving, finite-dimensional representations; this paper relaxes all three constraints at once—complex field, semi-functors, infinite-dimensional codomain—and shows that any such representation of a finite-dimensional tomographically-local GPT factors as Q(T) = χ_B ∘ C(T) ∘ φ_A, with χ_A fixed by states and φ_A fixed by χ_A and Q(id_A). The proof is constructive and mostly self-contained given their prior framework. I checked the main steps: Lemma II.1 gives the resolution of the identity; empirical adequacy fixes scalars; injectivity of χ_A follows from the left inverse; the semi-functor case uses the idempotent splitting D_A = χ_A ∘ φ_A. The stress-test note is right: no hidden circularity found.\n\nWhat's new: complex-valued KD-type representations are the obvious motivation, and the categorical factorization cleanly subsumes the frame-and-dual-frame picture for quantum theory. The KD functor from Ref. [49] appears as a special case. That is a genuinely useful organizing result for the applications listed (metrology, scrambling, quantum computation). The proof of the complexification functor's strong monoidal property is spelled out in the appendix, which is more than this literature usually does.\n\nSoft spots, in proportion: (1) Tomographic locality is load-bearing. Lemma II.1 is quoted from Ref. [26]; if a GPT has holistic degrees of freedom the decomposition fails. The paper says this clearly, so it's not a hidden flaw, but it limits the theorem's scope. The stress-test note suggests single-system spanning might suffice; I didn't verify that, but it does not affect the theorem as stated. (2) The paper builds heavily on the authors' own framework—Refs. [26] and [49]. That's legitimate since those are the prior results being generalized, but a reader who doesn't accept that framework will have to do extra work. Lemma II.1 in particular is imported without proof. (3) The title says \"physical theories,\" but the theorems apply to sequential composition only; parallel composition is left to future work. They acknowledge this, but it means the representation of composite systems is not fully captured. (4) Minor: the \"infinite-dimensional\" extension is only for semi-functors; Corollary III.2 shows functors must land in finite-dimensional spaces, so the title's breadth slightly overstates the functor case.\n\nVerdict: this deserves a serious referee. The main theorem is clearly stated, the proofs are checkable, and the limitations are disclosed. I'd bring it to our reading group and would cite it when the KD categorical picture comes up.","headline":"Genuinely new structure theorem for complex-valued quasiprobability representations, with clean explicit proofs and one important caveat (tomographic locality); deserves serious refereeing.","tokens_in":37280,"tokens_out":1833,"would_cite":true,"duration_ms":14666,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Every complex-valued quasiprobability representation of a tomographically local GPT is fixed by its action on states and the identity channel.","keywords":["quasiprobability representations","complex-valued quasiprobabilities","Kirkwood–Dirac distributions","generalized probabilistic theories","structure theorem","semi-functors","tomographic locality","frame representations"],"falsifier":"Build a finite-dimensional GPT whose two-party state space strictly contains the span of product states (so tomographic locality fails), define a linearity-preserving, empirically adequate complex-valued quasiprobability representation of it, and check whether every process matrix factors as $Q(T)=\\chi_B\\circ C(T)\\circ\\phi_A$ with $\\phi_A=\\overline{\\chi_A}^{-1}\\circ Q(\\mathrm{id}_A)$; a single non-factorizing example would refute Corollary III.4 as stated.","tokens_in":36359,"feed_emoji":"⚛️","tokens_out":9209,"duration_ms":81211,"temperature":0.7,"pith_summary":"Quasiprobability representations, especially complex-valued Kirkwood–Dirac distributions, are widely used tools, but their structural constraints were previously known only in the real-valued, identity-preserving case. This paper proves that any complex-valued quasiprobability representation of a finite-dimensional, tomographically local generalized probabilistic theory factors system-by-system as $Q(T) = \\chi_B \\circ C(T) \\circ \\phi_A$, where $\\chi_A$ is an injective complex-linear map fixed by the representation's action on states and $\\phi_A$ is fixed by its action on the identity. In quantum theory this recovers the familiar frame-and-dual-frame description, and in the functorial case it forces the codomain dimension to match the complexified system dimension. If correct, the entire freedom in such representations is concentrated in per-system choices of states and identity images.","feed_headline":"Complex quasiprobability maps factor into two simple pieces","feed_subtitle":"For any tomographically local GPT, the whole representation is fixed by states and the identity channel.","key_machinery":"The workhorse is the complexification functor $C:\\mathrm{Vect}_{\\mathbb{R}}\\to\\mathrm{Vect}_{\\mathbb{C}}$, which sends a real vector space $W$ to $W\\oplus W$ with $i(w_1,w_2)=(-w_2,w_1)$ and an $\\mathbb{R}$-linear map $f$ to its unique $\\mathbb{C}$-linear extension $f^{\\mathbb{C}}$. The paper also uses a semi-functor as the notion of representation: a map between process theories that preserves sequential composition but not necessarily the identity, so the image of each identity is an idempotent. Tomographic locality, stated as Lemma II.1, supplies a resolution of the identity as a finite linear combination of measure-and-prepare processes; this resolution lets the authors define $\\chi_A$ and $\\phi_A$ from states and effects alone, and the complexification functor lets those real-data maps be read as $\\mathbb{C}$-linear maps. The key identity is the factorization $Q(T)=\\chi_B\\circ C(T)\\circ\\phi_A$, with $\\phi_A=\\overline{\\chi_A}^{-1}\\circ Q(\\mathrm{id}_A)$, where $\\overline{\\chi_A}$ is the invertible corestriction of $\\chi_A$ onto its image.","core_discovery":"The central claim, Corollary III.4, states that every empirically adequate, linearity-preserving complex-valued quasiprobability representation $Q:\\mathcal{G}\\to\\mathrm{FinQuasiSubStoch}_{\\mathbb{C}}$ of a tomographically local, finite-dimensional GPT can be written as $Q(T)=\\chi_B\\circ C(T)\\circ\\phi_A$ for every process $T:A\\to B$, where $C$ is the complexification functor from real to complex vector spaces. For each system $A$, $\\chi_A$ is an injective $\\mathbb{C}$-linear map uniquely determined by the action of $Q$ on states, and $\\phi_A = \\overline{\\chi_A}^{-1}\\circ Q(\\mathrm{id}_A)$, with $\\overline{\\chi_A}$ the invertible corestriction onto the image of $\\chi_A$. Consequently the representation is fully determined by its values on states and on the identity. In the quantum case this says each system's representation is exactly a choice of frame and dual frame: states pick out $\\chi_A$, while effects, equivalently the identity's representation, pick out $\\phi_A$. The theorem covers semi-functors that need not preserve the identity, extends to complex-valued codomains, and allows infinite-dimensional target spaces; in the narrower functorial case, $\\chi_A$ becomes invertible and the representation is determined by states alone.","pith_inferences":["A testable consequence is that two representations of the same tomographically local GPT that agree on all states and on the identity must agree on every process, so discrepancies in process representations can be detected from those data alone.","If the factorization survives weaker categorical assumptions than symmetric monoidal categories, the structure would appear to be driven by tomographic resolution rather than by the full tensor-product calculus; if it fails, the monoidal structure is load-bearing in a way the current proof does not expose.","The idempotent $Q(\\mathrm{id}_A)$ appearing in the semi-functorial case is a natural candidate for quantifying how far a representation departs from being identity-preserving, which could be connected to overcompleteness of the associated frames."],"forward_implications":["For quantum theory, every complex-valued quasiprobability representation is a frame representation: the action on states fixes a frame for the operator space, and the action on the identity fixes the dual frame.","Any functorial, identity-preserving representation must map each finite-dimensional system to a complex vector space of the same dimension, ruling out infinite-dimensional target spaces in the functorial case.","The full freedom in choosing a representation of a tomographically local GPT is per-system: choose the images of the states and the image of the identity, and the representation of every process is determined.","Classification questions for complex-valued quasiprobability representations, such as which ones are Kirkwood–Dirac-like, reduce to classifying admissible frames, dual frames, and the idempotents arising from the identity.","The semi-functorial version applies to representations that do not preserve the identity, so the theorem covers broader classes of complex-valued representations than earlier real-valued functorial results."],"supporting_citations":[{"why":"Provides the tomographic-locality lemma (Lemma II.1) used to resolve the identity, and the earlier real-valued functorial structure theorem that this work extends.","marker":"[26]"},{"why":"Constructs the Kirkwood–Dirac functorial representations that motivate the theorem and instantiate the quantum case.","marker":"[49]"},{"why":"Establishes that quasiprobability distributions are frame representations, the link used to interpret $\\chi_A$ and $\\phi_A$ as frame and dual frame.","marker":"[18]"},{"why":"Together with [18], supplies the frame-representation theorem for quasiprobability distributions that underlies the identification of representations with frames.","marker":"[17]"},{"why":"Supplies the universal-property complexification construction used to define the complexification functor $C$ and its unique extension property.","marker":"[88]"}],"fun_headline_variants":["Complex quasiprobability: states and identity fix reps","Complex quasiprobability reps: determined by states and identity","Structure theorem: reps factor via states and identity action","States and identity alone fix quasiprobability representations","Two maps fix any complex quasiprobability representation"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is tomographic locality: every composite system's behaviour is fully determined by local measurements on its parts, so the identity process can be expanded as a finite linear combination of measure-and-prepare processes; the paper itself notes that if this fails, composite systems have holistic degrees of freedom invisible to product effects, and the factorization proof does not go through.","fun_headline_variants_meta":{"raw":{"variants":["Complex quasiprobability: states and identity fix reps","Complex quasiprobability reps: determined by states and identity","Structure theorem: reps factor via states and identity action","States and identity alone fix quasiprobability representations","Two maps fix any complex quasiprobability representation"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001285,"raw_usage":{"total_tokens":5306,"prompt_tokens":1059,"completion_tokens":4247,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":675,"completion_tokens_details":{"reasoning_tokens":4168}},"tokens_in":675,"tokens_out":4247,"duration_ms":26651,"temperature":1.0,"reasoning_tokens":4168,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T15:52:13.556766+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Build a finite-dimensional GPT whose two-party state space strictly contains the span of product states (so tomographic locality fails), define a linearity-preserving, empirically adequate complex-valued quasiprobability representation of it, and check whether every process matrix factors as $Q(T)=\\chi_B\\circ C(T)\\circ\\phi_A$ with $\\phi_A=\\overline{\\chi_A}^{-1}\\circ Q(\\mathrm{id}_A)$; a single non-factorizing example would refute Corollary III.4 as stated.","supporting_citations":[],"review_version":2}