{"id":"146eab2e-4678-4733-9774-913d78d2e393","arxiv_id":"2507.06770","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For finite-dimensional quantum relay channels, the paper proves achievable quantum-information and entanglement-generation rates using full and partial decode-forward coding.","lead":"The authors derive achievable rates for sending quantum information through a three-terminal relay channel, where the relay decodes either all or part of the message. This provides information-theoretic bounds that could inform how quantum repeaters and distributed quantum networks are designed.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 3's unassisted rate depends on recycling FQSW by-product entanglement across blocks, but no lemma establishes that the generated entanglement is both clean and reusable at no rate cost.","rationale":"The reader's weakest_assumption identified essentially the same step in the unassisted proof: the reuse of by-product entanglement as a clean resource at no rate cost. The reader also identified an additional concern about the decoupling constraint and data-processing inequality, but I find the by-product recycling to be the more concrete and more load-bearing issue, since it is presupposed in both Theorem 2 and Theorem 3. The paper gives plausible decoupling bounds (Appendix B-A) that support the existence of the by-product entanglement at the end of a block, but the proof jumps from 'the by-product exists' to 'the by-product is available for the next block at zero rate'. In a block-Markov protocol, the relay's decoding in block j-1 may correlate or consume the by-product systems. The paper does not present a lemma establishing that the by-product can be cleanly handed to the next block. This is a specific, fixable gap rather than a contradiction: the rate expressions may well be correct, but the proof as written does not establish them. I agree with the reader's verdict of CONDITIONAL; I disagree with the reader's emphasis on the data-processing inequality as the primary issue. The erasure-channel example provides helpful numerical illustration but does not test the recycling step, since it assumes the entangled inputs and orthogonal structure directly. No formal verification or code is present, which reinforces the need for the proposed check.","tokens_in":22657,"tokens_out":1779,"duration_ms":17977,"concrete_test":"Write out the block-Markov recursion explicitly for two consecutive blocks j-1 and j, tracking the quantum state of the systems G-hat'_1,A, G-hat'_2,A, G-hat''_A from the end of block j-1 through the relay's decoding and encoding maps to the start of block j. Then verify: (1) the by-product systems are uncorrelated with M' and R' after the relay applies its decoder; (2) the relay's re-encoding map can be chosen so that the by-product entanglement is not consumed or altered. If either condition fails, recompute the rate with the appropriate net entanglement-consumption term; if the resulting rate falls below QPD-F(N), the claimed unassisted rate is not established.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim of Theorem 3 is an achievable entanglement-generation rate for a fully quantum relay channel without assistance. The proof in Appendix C-B derives the rate QPD-F(N) by taking all entanglement-assistance rates L'E, L'B, L''B, and all by-product generation rates L-hat'E, L-hat'B, L-hat''B to zero in Proposition 1. This is only legitimate if the by-product entanglement generated during FQSW decoding in block j-1 is available, uncorrelated with the decoded messages and untouched by the decoding operations, as a free resource in block j. The paper asserts this recycling in the introduction ('we leverage the fact that communication occurs over multiple rounds, allowing the protocol to reuse entanglement generated as a by-product in block i-1 during block i') and in Section I, but the formal proof in Appendix B and C never isolates or proves a lemma for this reuse. The decoupling bounds in Eqs. (29), (33), and (36) ensure the by-product states are approximately maximally entangled with the relevant reference systems at the end of each block, but they do not address whether these by-product states survive the relay's subsequent decoding or re-encoding operations, nor whether they remain independent of the message M' decoded by the relay. If the by-product is correlated with the decoded message or is partially consumed by the decoding operation, the unassisted rates in Theorem 3 (and Theorem 2, which relies on the same scheme) would require an additional entanglement-consumption term, weakening the claimed result. This is the load-bearing step because the entire unassisted result is obtained by zeroing out all assistance and generation rates in Proposition 1.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies a three-terminal quantum relay channel N_{AD→BE} and proposes block-Markov coding schemes based on the fully quantum Slepian-Wolf (FQSW) protocol. It states Proposition 1, a rate region for transmission with rate-limited entanglement assistance, and then derives unassisted lower bounds for full decode-forward (Theorem 2) and partial decode-forward (Theorem 3), an entanglement-assisted bound (Theorem 4), and an erasure-channel example in Section VI. The claimed unassisted bounds are Q_D-F = max min{I(A1>B), I(A1>E)} and Q_PD-F = max min{I(A1A2>B), I(A1>E)+I(A2>BA1)}. The proof is based on decoupling inequalities in Appendix B, Uhlmann's theorem, and taking all entanglement-consumption and by-product-generation rates to zero.","tokens_in":52,"tokens_out":53439,"duration_ms":835707,"significance":"The topic is timely: quantum relay channels are central to repeater architectures and distributed quantum computing, and a rigorous achievable-rate framework would be a useful contribution. The paper correctly identifies the importance of block-Markov coding and uses standard tools (FQSW, decoupling, typical subspaces). However, the main unassisted claims are not established. The proof that Eq. (30) is inactive uses a data-processing inequality that does not hold for the channel N_{AD→BE}, and the reuse of by-product entanglement across blocks is asserted but never proved. In addition, the full decode-forward bound in Theorem 2 is subject to a complementarity relation between coherent informations to the two outputs B and E, which makes the stated maximum zero for every channel under the paper's own nonnegativity constraints. These are load-bearing issues in the central results.","major_comments":[{"comment":"The proof that Eq. (30) is inactive relies on the identity H(A1|D)_σ = I(A1>A)_σ and on the statement that \"the data processing inequality for the coherent information\" gives I(A1>A)_σ ≥ I(A1>B)_ω. The identity is correct, but the data-processing step is not: the channel N_{AD→BE} acts jointly on A and D, so standard coherent-information data processing applies to the combined input AD, giving I(A1>AD)_σ ≥ I(A1>B)_ω, not I(A1>A)_σ ≥ I(A1>B)_ω. A concrete counterexample is σ_{A1 A D} = EPR(X,A) ⊗ EPR(Y,D) with A1=(X,Y), A=X, D=Y, and N the identity channel from AD to B with E trivial: then I(A1>A)_σ = 0 while I(A1>B)_ω = 2. Thus Eq. (30) may be active, and the derivation of Theorems 2 and 3 is incomplete.","section":"Appendix C-A, Eq. (30)"},{"comment":"There is a more basic obstruction to the stated full decode-forward bound. For any pure input σ_{A1AD} and any isometric extension V_{AD→BEJ} of the channel, the state on A1 B E J is pure. By weak monotonicity (equivalently strong subadditivity), H(B)+H(E) ≤ H(BJ)+H(EJ), which immediately gives I(A1>B)_ω + I(A1>E)_ω ≤ 0. Therefore, under the paper's own constraints I(A1>B)_ω ≥ 0 and I(A1>E)_ω ≥ 0, both quantities must be zero for every feasible σ. Hence Q_D-F(N) as defined in Eq. (10) is identically zero for every quantum relay channel, making Theorem 2 vacuous. A non-vanishing decode-forward rate must involve two different channel uses (first hop to the relay and second hop from the relay), with different input states, but the paper's formula and proof use the same ω = N(σ) for both terms.","section":"Section IV-B, Theorem 2"},{"comment":"The unassisted rates in Theorems 2 and 3 depend on reusing the by-product entanglement generated by FQSW decoding in block j-1 as a clean resource in block j. The paper asserts this in Section I (\"allow the protocol to reuse entanglement generated as a by-product in block i-1 during block i\") and in the proof it sets the by-product generation rates \\hat L to zero. However, no lemma in Appendix B or C proves that the by-product systems \\hat G'_1,A, \\hat G'_2,A, and \\hat G''_A are uncorrelated with the relay's decoded message M', survive the relay's subsequent decoding and re-encoding operations, and remain available at no rate cost. The decoupling bounds in Eqs. (29), (33), and (36) only establish closeness of reduced states; they do not address correlations with M' or consumption by the decoding operations. Without this lemma, taking the by-product rates to zero is not justified, and the claimed unassisted rates may require an additional entanglement cost.","section":"Section I and Appendix C"},{"comment":"The erasure example appears to model B' and E as two independent erasure outputs of the same input, with B' obtained by erasure α and E obtained by a further erasure β applied to B'. For a valid quantum channel, two outputs of the same transmission cannot both carry positive coherent information about the same quantum input; the complementarity relation I(A1>B')+I(A1>E) ≤ 0 holds for any isometric channel. A channel that outputs a noiseless copy to B' and a degraded copy to E would violate no-cloning unless B' is correspondingly disturbed. The rate computations in Eqs. (18)-(22), which give positive values for both I(A1>B') and I(A1>E), are therefore not achieved by a physically realizable quantum channel, and the example needs to be replaced by a model with a genuine quantum tradeoff between the two outputs.","section":"Section VI, Eqs. (18)-(22)"}],"minor_comments":[{"comment":"There are several typos and grammatical issues, including \"strictly casual\" instead of \"strictly causal\" in Section III, \"fideluty\" in Appendix A, \"intepretation\" in Section IV, and \"a the US National Security Agency\" in the Introduction; these should be corrected.","section":"Throughout"},{"comment":"The expression in Eq. (23) is missing a trace-distance norm symbol on the left-hand side; as printed, it is not a well-formed inequality.","section":"Eq. (23)"},{"comment":"The definition of the rate region in Proposition 1 would benefit from explicitly stating that all rates are nonnegative and from defining the notational convention for the by-product systems in one place rather than only in Appendix B.","section":"Section IV-A"}],"recommendation":"reject","confidential_remarks":"The manuscript's main unassisted results appear to be invalid as stated: Theorem 2 is vacuous due to the complementarity of coherent informations to the two outputs, and the proof of both Theorems 2 and 3 contains an invalid data-processing step and an unproved by-product-reuse lemma. These are not local fixable presentation issues; they affect the central claims. The Section VI example also seems to rely on an unphysical channel model. I would recommend rejection, though the authors might be able to salvage a corrected version by reformulating the bounds with two different channel uses and proving the by-product reuse lemma."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper does something genuinely new: it proposes a partial decode-forward strategy for fully quantum messages over a three-terminal quantum relay channel, with explicit achievable-rate formulas for both entanglement-assisted and unassisted settings. The idea of reusing FQSW by-product entanglement across blocks is interesting and could matter for quantum repeater analysis. The erasure-channel example is a nice sanity check, and the framework generalizes earlier c-q relay results. That is real content, and it deserves a serious look.\n\nBut there are two load-bearing gaps. The first, and more serious, is the recycling assumption. The paper claims in Section I that entanglement generated as a by-product in block i-1 can be reused in block i at no rate cost. The decoupling bounds in Appendix B show the by-product states are close to maximally entangled at the end of each block, but nothing proves those states survive the relay's decoding and re-encoding operations, or that they remain uncorrelated with the decoded message. Since the unassisted rates are obtained by zeroing all assistance and generation rates in Proposition 1, this missing lemma is doing real work. Without it, the claimed unassisted rates might need an extra entanglement-consumption term.\n\nThe second gap is in the proofs of Theorems 2 and 3, where the first constraint is dropped using H(A1|D)_sigma = I(A1>A)_sigma and a data-processing inequality from A to B. That identity is fine for pure states, but the data-processing step only applies if the map from A to B is a channel independent of D. Here D is entangled with A and A1, so the effective map is not obviously a CPTP map from A alone. The paper cites Theorem 11.9.3 but does not justify its applicability. This needs a direct argument or a different way to show the first constraint is inactive.\n\nThese are specific, fixable issues rather than signs of a broken approach. The paper is otherwise coherent, and the authors engage with the relevant literature. The exposition has typos and rough edges, but that is minor. There is no code or formal verification, so the proof is what carries the result.\n\nMy recommendation: send it to peer review. The problem is important, the proposed rates are plausible, and the gaps can be addressed with a careful proof or a counterexample. But the current version is not ready for acceptance. I would not cite it as a proven result until the recycling lemma and the data-processing step are resolved.","headline":"Promising new lower bounds for fully quantum relay channels, but the unassisted proofs rest on a missing recycling lemma and a questionable data-processing step; send to review with major revision.","tokens_in":23476,"tokens_out":3513,"would_cite":false,"duration_ms":39692,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P45","94A40"],"pacs":["03.67.Hk","03.67.-a"],"model":"deepseek-v4-flash","headline":"This paper proves that a finite-dimensional quantum relay channel can generate entanglement at any rate up to $Q_{\\rm PD-F}(N)=\\max_\\sigma\\min\\{I(A_1A_2\\rangle B)_\\omega, I(A_1\\rangle E)_\\omega+I(A_2\\rangle BA_1)_\\omega\\}$ using a partial…","keywords":["quantum relay channel","partial decode-forward","quantum capacity","entanglement generation","coherent information","block Markov coding","FQSW protocol","entanglement assistance"],"falsifier":"Run the two-block protocol on the orthogonal erasure relay channel of Sec. VI and compute the joint state of the fed-forward by-product EPR pairs and the relay's decoded message after FQSW decoding; if the trace distance from a product state does not vanish exponentially in the block length, or if the predicted rate $2(1-2q)+(1-2\\gamma)$ is not attained, the reuse step in the proof of Theorem 3 is broken.","tokens_in":22406,"feed_emoji":"📡","tokens_out":9284,"duration_ms":89055,"temperature":0.7,"pith_summary":"Quantum relays are a proposed remedy for photon loss and decoherence in long-distance quantum communication, but their information-theoretic limits are only partially understood. This paper establishes achievable rates for a three-terminal relay channel in which the relay decodes only part of the quantum message and forwards it, while the receiver decodes the other part directly. The main unassisted result says the entanglement-generation rate can be at least the minimum of two coherent-information expressions; the full decode-forward special case gives a lower bound on the ordinary quantum capacity, and an entanglement-assisted version is derived as well. These bounds matter because they quantify when a relay genuinely helps, and because the unassisted protocol achieves them by recycling the entanglement that the relay's own decoding produces as a by-product.","feed_headline":"Split-message code lets a quantum relay beat direct links","feed_subtitle":"The relay decodes one part; the receiver handles the other, boosting entanglement rates.","key_machinery":"The paper builds on the fully quantum Slepian–Wolf (FQSW) protocol, a procedure that sends quantum information while also outputting EPR pairs as a by-product, and embeds it in a block-Markov code over many blocks. At the relay, FQSW decoding of one block produces an entangled resource that the protocol feeds forward into the next block, so the unassisted rates do not need to consume preshared entanglement. Three decoupling steps (encoder, relay, destination) convert the protocol's error constraints into the coherent-information terms of the rate formulas, and rate splitting into $A_1$ and $A_2$ realizes the partial decode-forward tradeoff.","core_discovery":"The central claim, on the authors' own terms, is an achievability theorem: for a memoryless finite-dimensional quantum relay channel $N_{AD\\to BE}$, the entanglement-generation capacity without assistance is at least $Q_{\\rm PD-F}(N)$, the maximum over pure input states $\\sigma_{A_1A_2AD}$ of $\\min\\{I(A_1A_2\\rangle B)_\\omega, I(A_1\\rangle E)_\\omega + I(A_2\\rangle BA_1)_\\omega\\}$, where $\\omega = N_{AD\\to BE}(\\sigma_{A_1A_2AD})$ and $A_1,A_2$ are auxiliary systems for the relay path and the direct path. When the relay decodes everything, the unassisted quantum capacity is at least $\\max_\\sigma\\min\\{I(A_1\\rangle B)_\\omega, I(A_1\\rangle E)_\\omega\\}$; with unlimited entanglement assistance, the quantum capacity is at least $\\max_\\sigma[\\frac{1}{2}I(A_1;B)_\\omega-\\frac{1}{2}I(A_1;D)_\\sigma]$. All three are lower bounds on capacity, not exact characterizations.","pith_inferences":["If the by-product entanglement reuse is made fully rigorous with its own lemma, the same block-Markov and FQSW construction should yield rate formulas for assist-forward and compress-forward relaying, with the compression cost appearing as a conditional entropy term alongside the coherent-information minima.","Applied to bosonic channels, the min-of-coherent-informations structure predicts a rate-distance curve for relays that interpolates between the repeaterless rate-distance limit and an ideal repeater bound; the split between $A_1$ and $A_2$ would become a continuous optimization over how much entanglement to allocate to each segment.","The orthogonal erasure example suggests a design rule for practical repeaters: allocate the sender's entanglement budget between the relay and direct paths according to the erasure probabilities, since the achievable rate is a sum of two independent terms."],"forward_implications":["For orthogonal links, the full decode-forward bound becomes $\\min\\{I_c(M), I_c(P)\\}$, so a non-orthogonal channel with entangled sender–relay inputs can beat a repeater chain built from independent links.","When the receiver's channel is a degraded version of the relay's channel, the full decode-forward bound reduces to $I(A_1\\rangle B)_\\omega$: the last hop is the bottleneck.","On a quantum erasure relay channel with erasure probabilities $\\alpha,\\beta,\\gamma$, the partial decode-forward rate is $2(1-2q)+(1-2\\gamma)$ with $q=\\alpha+(1-\\alpha)\\beta$, which exceeds either the relay-only or direct-only rate.","With entanglement assistance, the same relay setup supports classical communication at rate at least $2Q_{\\rm EA,D-F}$ by superdense coding."],"supporting_citations":[{"why":"Supplies the FQSW theorem and the side-information capacity formulas that the entanglement-assisted bound and the decoupling arguments are built on.","marker":"[51]"},{"why":"Gives the FQSW protocol as the mother-of-all-protocols result, the source of the EPR by-product that the unassisted scheme reuses.","marker":"[74]"},{"why":"Earlier partial decode-forward for classical-quantum relay channels, the baseline this work extends to fully quantum messages.","marker":"[37]"},{"why":"Prior framework for fully quantum relay channels with several coding strategies, the model and comparison class for the new decode-forward rates.","marker":"[42]"},{"why":"Data processing inequality for coherent information, used to drop the input-entropy constraint in the proofs of Theorems 2 and 3.","marker":"[63]"},{"why":"Generalized Uhlmann theorem with non-normalized purifications, used to construct the encoder and decoder partial isometries.","marker":"[62]"}],"fun_headline_variants":["Quantum relay splits message to boost rates","Partial decode-forward lifts quantum capacity","Relay decodes part, receiver rest, rate gains","Quantum relay beats direct links with split coding","Decode-forward strategy upgrades quantum relay"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The unassisted rates assume that entanglement produced as a by-product of the relay's decoding in one block can be carried into the next block as a clean, uncorrelated resource at no rate cost, a reuse the paper states but does not prove as a separate lemma.","fun_headline_variants_meta":{"raw":{"variants":["Quantum relay splits message to boost rates","Partial decode-forward lifts quantum capacity","Relay decodes part, receiver rest, rate gains","Quantum relay beats direct links with split coding","Decode-forward strategy upgrades quantum relay"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000156,"raw_usage":{"total_tokens":1250,"prompt_tokens":1006,"completion_tokens":244,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":622,"completion_tokens_details":{"reasoning_tokens":179}},"tokens_in":622,"tokens_out":244,"duration_ms":3332,"temperature":1.0,"reasoning_tokens":179,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T18:55:58.798734+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the two-block protocol on the orthogonal erasure relay channel of Sec. VI and compute the joint state of the fed-forward by-product EPR pairs and the relay's decoded message after FQSW decoding; if the trace distance from a product state does not vanish exponentially in the block length, or if the predicted rate $2(1-2q)+(1-2\\gamma)$ is not attained, the reuse step in the proof of Theorem 3 is broken.","supporting_citations":[{"cited_title":"The capacity of quantum channels with side information at the transmitter,","cited_arxiv_id":null,"evidence_quote":"Supplies the FQSW theorem and the side-information capacity formulas that the entanglement-assisted bound and the decoupling arguments are built on."},{"cited_title":"Partial decode-forward for quantum relay channels,","cited_arxiv_id":null,"evidence_quote":"Earlier partial decode-forward for classical-quantum relay channels, the baseline this work extends to fully quantum messages."},{"cited_title":"Quantum Relay Channels","cited_arxiv_id":"2411.16263","evidence_quote":"Prior framework for fully quantum relay channels with several coding strategies, the model and comparison class for the new decode-forward rates."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Data processing inequality for coherent information, used to drop the input-entropy constraint in the proofs of Theorems 2 and 3."}],"review_version":1}