{"id":"7f8b715a-9d1e-4252-a950-88a1ec0b6a00","arxiv_id":"2607.24682","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"CP-preserving channels are fully characterized for d≤4; CPDNN maps to qubits (and unital maps from qubits) are CPCP; and trace-distance non-negativity fails strong monotonicity.","lead":"The paper gives necessary and sufficient conditions for CP-preserving quantum channels in dimensions up to 4, an alternative argument that CPDNN maps into qubits are CPCP, and an explicit counterexample showing the trace-distance non-negativity measure is not strongly monotone. It closes several open questions left by Johnston and Sikora’s resource theory of non-negative amplitudes.","discovery_kind":"extension","skeptic_critique":{"model":"moonshotai/kimi-k3","headline":"The unproved comparison-matrix assertion in Theorem 3.6 is a genuine hole as written, but it appears repairable from the paper's own ingredients: trace-preservation forces the diagonal blocks of J_{Φ*} to be diagonal, making M(J_{Φ*}) a signature conjugate of J_{Φ*}.","rationale":"The reader identified exactly the right load-bearing weakness: the comparison-matrix PSD assertion is the sole bridge from DNN to CP in Theorem 3.6, it is unproved, and Theorem 3.7 depends on it. My independent check confirms the gap is real as the text stands — DNN alone does not imply a PSD comparison matrix, so the missing step must exploit the channel structure, which the paper never does explicitly. However, the gap looks fillable in a few lines (TP ⟹ unital dual ⟹ diagonal blocks of J_{Φ*} are diagonal ⟹ comparison matrix is a signature conjugate of J_{Φ*}), and Cha's independent proof means the claim itself is not in doubt. This supports keeping CONDITIONAL rather than escalating: the paper's \"alternative proof\" is currently incomplete but almost certainly completable, and the conditionality should be specifically about inserting the diagonal-block lemma. The other headline results hold up: I verified the §5 strong-monotonicity counterexample arithmetic (eigenvalues of ρ − I/5, the values N_T(|Υ₂⟩)=1, N_T(|Υ₃⟩)=4/3, and the interval (4/25, 2/5)), and the d≤4 characterization reduces correctly to standard copositivity tests once CP=DNN is invoked. Net: agree with the reader's weakest_assumption, agree with CONDITIONAL, with the refinement that the required fix is likely a short lemma rather than a new idea.","tokens_in":11524,"tokens_out":5534,"duration_ms":109840,"concrete_test":"Fill in the missing lemma and check it: for a TP CPDNN map Φ: M_n→M_2, prove each diagonal block Φ*(E_{mm}) of J_{Φ*} is diagonal (from Σ_m Φ*(E_{mm}) = I_n plus entrywise nonnegativity), then confirm M(J_{Φ*}) = diag(I,−I) J_{Φ*} diag(I,−I). Corroborate numerically: sample random CPDNN channels Φ: M_3→M_2 (random PSD entrywise-nonnegative Choi matrices rescaled to meet the TP marginal), compute the minimum eigenvalue of M(J_{Φ*}). If any sample yields a negative eigenvalue, the repair fails and the proof gap is substantive; if all pass and the diagonal-block argument goes through, Theorem 3.6 (and hence 3.7) stands with a one-paragraph patch.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The reader correctly located the weakest point. In Theorem 3.6 the entire DNN⇒CP reduction rests on one sentence: \"The comparison matrix of J_{Φ*} is positive semidefinite, so J_{Φ*} is CP.\" No argument is given, and the assertion does not follow from DNN-ness alone — for a general DNN matrix A, M(A) need not be PSD (otherwise DNN=CP in all dimensions, contradicting the 5×5 counterexamples the paper itself relies on in §4). So the bridge must use the extra structure of J_{Φ*}, and the paper never says what that structure buys.\n\nChecking whether the gap is fatal: it is not, apparently. Φ trace-preserving ⟹ Φ* unital, so Σ_m Φ*(E_{mm}) = I_n. Each Φ*(E_{mm}) is DNN (PSD because Φ* is a CP map; entrywise nonnegative because J_{Φ*} is). A sum of entrywise-nonnegative PSD matrices equals I_n, whose off-diagonal entries are zero; since every summand's off-diagonal entries are ≥ 0, each summand must be diagonal. Hence in J_{Φ*} = [[D, F],[F^T, I−D]], both D and I−D are diagonal, and therefore M(J_{Φ*}) = S J_{Φ*} S with S = diag(I_n, −I_n) — a congruence by a signature matrix, so M(J_{Φ*}) ≽ 0 follows immediately from J_{Φ*} ≽ 0. Theorem 3.5 then applies and the proof closes. So the assertion is true and provable in three lines from facts already in the text — but the paper as written does not prove it, and a reader cannot verify the central alternative proof without supplying this lemma themselves. Theorem 3.7 inherits the gap entirely (it is a one-line reduction to 3.6). The §5 counterexample, by contrast, I verified directly: the eigenvalue computation for ||ρ − I/5||_1 is correct, only an upper bound on N_T(ρ) is needed, and the inequality (4−p)/3 > (|1−5p|+|4−5p|+3)/5 indeed holds on (4/25, 2/5) — that result is sound.","agreement_with_reader":"agree"},"referee_report":{"model":"moonshotai/kimi-k3","summary":"The paper works within the Johnston–Sikora resource theory of non-negativity, in which CP matrices are free states and CP-preserving channels are free operations. It gives four main results: (i) necessary and sufficient conditions for a channel to be CP-preserving in dimensions d≤4 (Theorem 2.1 and Observations 2.3–2.8), phrased as copositivity of symmetrized dual-map images, recovering the known qubit conditions and yielding an SDP formulation; (ii) an alternative proof that every CPDNN channel Φ : M_n→M_2 is CPCP (Theorem 3.6, a result recently obtained by Cha) and a new dual statement that every unital CPDNN map M_2→M_n is CPCP (Theorem 3.7); (iii) symmetry reductions showing the nearest CP/DDN/DNN state to a permutation-invariant state can be assumed to share the symmetry, with evaluation of the trace-distance measure for the uniform states |Υ_d⟩; (iv) an explicit counterexample showing the trace-distance of non-negativity violates strong monotonicity under CPCP operations, using ρ = p|Υ₂⟩⟨Υ₂| ⊕ (1−p)|Υ₃⟩⟨Υ₃| and the block-projection channel, for p ∈ (4/25, 2/5). Items (i), (iii), and (iv) are sound on inspection; item (ii) contains an unproved load-bearing assertion.","tokens_in":11874,"tokens_out":5988,"duration_ms":210685,"significance":"If the results stand, the paper closes two concrete open problems from Johnston–Sikora (2022): the failure of strong monotonicity for the trace-distance of non-negativity is demonstrated by a fully explicit, falsifiable counterexample (state, channel, and elementary inequality all written out), and the CPCP/CPDNN equivalence is given an alternative, arguably more transparent route via the comparison-matrix criterion, plus a new dual direction (unital maps M_2→M_n). The d≤4 characterization is checkable via known copositivity criteria, and the symmetry reduction of §4 makes measure evaluations tractable. These are modest but solid and verifiable contributions to a small resource-theoretic literature; the explicit counterexample is the most durable item.","major_comments":[{"comment":"The proof asserts, without argument, that 'the comparison matrix of J_{Φ*} is positive semidefinite, so J_{Φ*} is CP' (via Theorem 3.5). This cannot follow from DNN-ness alone: for a general DNN matrix A, M(A) need not be PSD, otherwise DNN=CP in all dimensions, contradicting the d≥5 counterexamples the paper itself relies on in §4/Table 1. The bridge must therefore use the channel structure, and it can: Φ trace-preserving implies Φ* unital, so Σ_m Φ*(E_{mm}) = I; each summand Φ*(E_{mm}) is PSD and entrywise nonnegative, and since the sum has zero off-diagonal entries, each Φ*(E_{mm}) is diagonal. Hence in J_{Φ*} = [[D, F],[F^T, I−D]] both diagonal blocks are diagonal, so M(J_{Φ*}) = S J_{Φ*} S with S = diag(I_n, −I_n), a signature congruence, and PSD follows immediately. As written, however, the paper proves none of this, and the assertion is the sole load-bearing step. Please add this","section":"§3, proof of Theorem 3.6"},{"comment":"Theorem 3.7 is a one-line reduction to Theorem 3.6 (duality plus Lemmas 3.3–3.4) and therefore inherits the unproved comparison-matrix step in full. It cannot be considered established until the gap in Theorem 3.6 is closed. The repair indicated above suffices for both results, so this requires no new ideas — but the revision must be made before either theorem's proof is complete.","section":"§3, Theorem 3.7"}],"minor_comments":[{"comment":"This no-go claim ('there exists no unital CP-preserving channel...') rests entirely on a numerical SDP being 'found to be strictly infeasible' in MATLAB. No code, solver, tolerance, or dual (Farkas) certificate is given. A non-existence claim supported only by an unreported numerical run is not verifiable; please deposit the CVX script, report the infeasibility certificate, or give an analytic argument.","section":"§2.2, Proposition 2.9"},{"comment":"The set DDN appears in Eq. (5), Observation 4.2, and Table 1 without definition. Please define it (presumably diagonally dominant nonnegative, from [3]) at first use.","section":"§4, Eq. (5)"},{"comment":"The d=5 row reports N_T^DDN = N_T^CP = (17−√5)/10 and N_T^DNN = 1 + 1/√5 with no derivation, in a section that advertises that the nearest CP state 'can be found without any computation.' Please provide at least a sketch (which a_k in Eq. (9) achieve the optimum, and the DNN lower bound) or a reference.","section":"§4, Table 1"},{"comment":"The logic is correct but should be made explicit: I/5 is a CP state, so N_T(ρ) ≤ ‖ρ − I/5‖₁, and since the post-measurement average (4−p)/3 strictly exceeds this upper bound for p ∈ (4/25, 2/5), it exceeds N_T(ρ) itself. One line stating that an upper bound suffices would prevent misreading. Also worth noting explicitly that K_1, K_2 have nonnegative entries, so Φ is CPCP by Theorem 3.1(iii), not merely CP-preserving.","section":"§5"},{"comment":"The claim that 'every symmetric matrix [is a] difference of two CP matrices' is stated without justification. It is true (A + λI is strictly diagonally dominant with nonnegative entries for large λ, and diagonally dominant DNN matrices are CP) but needs a line of argument or a citation to [2]/[13].","section":"§2, Note 2.7"},{"comment":"The Y_{r,s} defined here as (1/2)Σ Φ(|i⟩⟨j|+|j⟩⟨i|)_{r,s}|i⟩⟨j| differs from Observation 2.4's Y_{r,s} (which symmetrizes the matrix units); please check the factors and make the notation consistent. In the d=3 example after Observation 2.4, the (2,1) entry reads '⟨2⟩⟨11|' — a typo for |2⟩⟨1|.","section":"§2, Observation 2.8"},{"comment":"Numerous typos: 'CP-preservinng', 'positve', 'quantumk', 'atlest', 'arbitray', 'chaannel', 'qubut', 'simialr', 'satifies', 'Johnstonet al.' (missing space), J_ϕ for J_Φ in Lemmas 3.3–3.4. Theorem 3.5 should carry a theorem/page number in [13]. Corollary 2.5's phrase 'trace preserving condition of Φ*' should read 'Φ* is trace-preserving since Φ is unital.'","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"Theorem 3.6 is explicitly an alternative proof of a result already settled by Cha (arXiv:2603.16962), so the genuinely new contributions are the d≤4 characterization of §2, the dual statement Theorem 3.7, and the §5 strong-monotonicity counterexample. Given that the alternative proof currently contains an unjustified step, the editor may wish to ask the authors to delineate more precisely what is new here versus [5] once the proof is completed."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The paper cleanly answers two concrete questions left open by Johnston–Sikora: it gives necessary-and-sufficient copositivity criteria for CP-preserving channels when both dimensions are at most 4, and it supplies an explicit counter-example showing that the trace-distance measure of non-negativity fails strong monotonicity under CPCP (hence also under CP-preserving) operations. Both of those pieces are usable as written.\n\nThe counter-example in §5 is the cleanest contribution. The block-diagonal state ρ = p|Υ₂⟩⟨Υ₂| ⊕ (1-p)|Υ₃⟩⟨Υ₃| and the two-block projection channel are elementary; the upper bound on N_T(ρ) via ||ρ-I/5||_1 is correct, and the elementary inequality (4-p)/3 > (|1-5p|+|4-5p|+3)/5 holds on (4/25,2/5). That settles the question negatively and can be trusted immediately. The d≤4 characterization (Thm 2.1 and the dual/Choi/copositivity observations) is the natural extension once CP=DNN is granted; the qubit recovery matches the known conditions, and the SDP formulation is a practical bonus. The symmetry-reduction argument for N_T(|Υ_d⟩) is also neat and immediately gives the values in Table 1.\n\nThe soft spot is Theorem 3.6 (and the one-line dual statement 3.7). The entire DNN\to CP step rests on the bare assertion that the comparison matrix of J_{Φ*} is PSD. That does not follow from DNN-ness alone, so a reader cannot verify the alternative proof without supplying the missing three-line lemma: trace-preservation forces the diagonal blocks of J_{Φ*} to be diagonal, after which M(J_{Φ*}) is a signature congruence of J_{Φ*} and therefore PSD. The claim is true and uses only ingredients already in the text, but as written the proof has a hole. Cha’s independent argument remains the only complete published proof for the qubit-output direction. Writing quality is rough (typos, incomplete sentences), yet the logical skeleton is recoverable.\n\nThis is for people already working inside the Johnston–Sikora resource theory or anyone who needs low-dimensional tests or a concrete non-monotone example. It deserves a serious referee; the monotonicity failure and the d≤4 criteria should survive, and the comparison-matrix gap is easily fixed. I would accept it for peer review.","headline":"Solid niche follow-up that settles the strong-monotonicity question with an explicit counter-example and extends the d≤4 characterization; the claimed alternative CPDNN⇒CPCP proof has a one-sentence gap that is repairable from the paper’s own facts but is not written down.","tokens_in":12708,"tokens_out":737,"would_cite":true,"duration_ms":13866,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P40","15B48","90C22"],"pacs":[],"model":"grok-4.5","headline":"CP-preserving channels are fully characterized in dimensions up to 4, and the trace distance of non-negativity fails strong monotonicity.","keywords":["CP-preserving channels","CPCP maps","CPDNN maps","trace distance of non-negativity","copositive matrices","quantum resource theory","Choi matrix"],"falsifier":"Construct a concrete CPDNN channel from n-by-n matrices to 2-by-2 matrices whose dual Choi matrix has a comparison matrix that is not positive semidefinite; if such a channel exists the equivalence CPDNN = CPCP collapses.","tokens_in":12222,"feed_emoji":"⚛️","tokens_out":917,"duration_ms":19071,"temperature":0.7,"pith_summary":"This paper works inside the resource theory of non-negativity of quantum amplitudes, where completely positive (CP) states are free and the free operations are channels that map CP states to CP states. It supplies necessary and sufficient conditions for a channel to be CP-preserving when the local dimensions are at most 4, recovers the known qubit criteria as a special case, and notes that the same conditions remain necessary in higher dimensions. It also proves that every CPDNN channel from n-by-n matrices into 2-by-2 matrices is automatically CPCP, and that every unital CPDNN map in the opposite direction is likewise CPCP. Finally it exhibits an explicit counter-example showing that the natural trace-distance measure of non-negativity is not strongly monotonic under free operations. A sympathetic reader cares because these results close several concrete open questions left by the original resource-theory paper and clarify which free operations and which monotones can safely be used.","feed_headline":"Trace distance of non-negativity is not strongly monotonic","feed_subtitle":"CP-preserving channels get full criteria up to dimension 4, and CPDNN maps into qubits are CPCP","key_machinery":"The dual-map copositivity criterion (Theorem 2.1 and Observation 2.4) together with the comparison-matrix argument that turns a DNN dual Choi matrix into a CP matrix (Theorem 3.6).","core_discovery":"For dimensions 2 through 4 a quantum channel is CP-preserving if and only if the symmetric real parts of the dual images of the matrix units are copositive; the same statements become only necessary once the dimension exceeds 4. Independently, every CPDNN channel into qubits is CPCP, every unital CPDNN map out of qubits is CPCP, and the trace distance of non-negativity fails strong monotonicity under CPCP channels.","pith_inferences":["The same block-projection counter-example that kills strong monotonicity of the trace distance will likely work for any other distance-based monotone that does not automatically average under direct sums.","If the comparison-matrix step can be repaired or replaced, the equivalence CPDNN = CPCP may extend to a larger class of output dimensions.","The SDP feasibility program already written for state conversion under CP-preserving channels can be reused to search systematically for a golden resource state in dimension 3 or higher."],"forward_implications":["CP-preserving channels on systems of size at most 4 can be certified by a finite list of copositivity checks or by a single SDP.","The free operations of the resource theory may safely be taken to be the CPCP maps whenever the output is a qubit.","Unital CPDNN maps from qubits to higher dimensions are automatically free operations in the strongest (CPCP) sense.","The trace-distance monotone cannot be used for asymptotic or multi-copy resource conversion arguments.","Symmetry reduction under permutations yields closed-form values of the trace-distance measure for the uniform pure states up to dimension 4."],"fun_headline_variants":["CP-preserving channels: full criteria in dims ≤4","Trace distance of non-negativity fails strong monotonicity","Every CPDNN channel into qubits is CPCP","Unital CPDNN maps from qubits are CPCP","Copositivity characterizes CP-preservers up to dim 4"],"cache_read_input_tokens":128,"weakest_assumption_plain":"The proof that a CPDNN channel into qubits is CPCP rests on an unproved claim that the comparison matrix of the dual Choi operator is positive semidefinite.","fun_headline_variants_meta":{"raw":{"variants":["CP-preserving channels: full criteria in dims ≤4","Trace distance of non-negativity fails strong monotonicity","Every CPDNN channel into qubits is CPCP","Unital CPDNN maps from qubits are CPCP","Copositivity characterizes CP-preservers up to dim 4"]},"model":"grok-4.5","effort":"low","cost_usd":0.003889,"raw_usage":{"total_tokens":1212,"prompt_tokens":737,"num_sources_used":0,"completion_tokens":64,"cost_in_usd_ticks":38888000,"prompt_tokens_details":{"text_tokens":737,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":411,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":737,"tokens_out":64,"duration_ms":6359,"temperature":1.0,"reasoning_tokens":411,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-31T07:59:57.609875+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Construct a concrete CPDNN channel from n-by-n matrices to 2-by-2 matrices whose dual Choi matrix has a comparison matrix that is not positive semidefinite; if such a channel exists the equivalence CPDNN = CPCP collapses.","supporting_citations":[],"review_version":1}