{"id":"4a091774-48c9-4555-a8cc-afa91e32f3e8","arxiv_id":"2412.16450","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A new [[(w+1)(w+K), K]] code family approximately corrects up to w amplitude-damping errors, and concatenation with the dual-rail code adds collective-coherent-error immunity.","lead":"This paper builds a new family of quantum error correcting codes that correct up to w energy-loss (amplitude-damping) errors while encoding K logical qubits at a higher rate than before. Doubling the code with a dual-rail inner code makes the same family immune to collective coherent errors, the identical small rotations that hit every qubit.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Orthogonality in Theorem 1 is likely true, but the claimed Z-syndrome lookup table is unproven for w≥2: colliding error patterns with identical syndrome may require different recovery, threatening the local-decoding claim.","rationale":"I read the paper in good faith and checked the central AQEC claim directly. The reader's weakest assumption, the asserted orthogonality of error subspaces in Appendix B, is not a genuine correctness risk: it follows from a simple Hamming-distance argument because any two terms of a codeword differ in at least 2(w+1) qubits for same-parity terms and in at least (K+1)(w+1) qubits for opposite-parity terms, while each side can change at most w qubits under weight-≤w AD errors. The diagonal O(γ^{w+1}) estimate also survives a detailed check: each per-block difference α(0)−α(1) is O(γ), and the parity imbalance over w blocks is a product of such differences, giving O(γ^w), so the total difference is O(γ^{w+1}). Thus Theorem 1 is mathematically defensible. However, the paper's advertised advantage of efficient detection via local Z-stabilizer lookup tables is not established for w≥2: the Z-syndrome is not injective on error patterns, as shown by the {0,2} versus {1} collision in the first block of the [[9,1]] code. The formal recovery in Appendix D can distinguish orthogonal error states, but the paper does not explain how to implement that distinction with only local Z measurements, nor does it provide the promised lookup table for multi-error cases. This gap affects the practical decoding claim, not the existence of some recovery achieving the AQEC bound, so the reader's CONDITIONAL verdict remains appropriate: the formal theorem is sound but the paper is incomplete in its decoding analysis. I therefore recommend no verdict change.","tokens_in":22014,"tokens_out":32173,"duration_ms":290550,"concrete_test":"For the [[9,1]] w=2 code, enumerate all 37 error patterns of weight at most 2, compute their 6-bit Z-stabilizer syndromes, and identify collisions such as {0,2} versus {1} in the first block. For each collision, construct the optimal syndrome-conditional recovery (e.g., the best CPTP map on the syndrome subspace) and evaluate the worst-case output fidelity as γ→0. If the infidelity is Θ(γ^2), the local Z-syndrome lookup-table claim fails; if it is O(γ^3), the ambiguity is benign and the local-decoding claim survives.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The weakest point is not the Appendix B orthogonality assertion. That assertion can be justified by a Hamming-distance argument: distinct codeword terms differ in at least 2(w+1) qubits (same parity) or (K+1)(w+1) qubits (opposite parity), so with at most 2w damping events their supports cannot overlap, and distinct error patterns on the same term give disjoint supports. The real gap is the decoding claim in the main text: 'a lookup table that maps a length-w binary syndrome string to identify AD errors.' For w≥2 the Z-stabilizer syndrome is not injective on error patterns. In the [[9,1]] w=2 code, for the first block of length 3, the two error patterns {0,2} and {1} both flip the two Z stabilizers Z0Z1 and Z1Z2, so they share a syndrome, yet their error states on |0_L> are orthogonal (010 vs 101 in that block). A recovery that treats both as the single error at position 1 applies an artificial AD rotation to a qubit that is |1> in the {0,2} state, potentially damping it and leaving a residual error that scales as O(γ^2), spoiling the claimed O(γ^{w+1}) fidelity for syndrome-only decoding. The Appendix D recovery uses projectors onto individual error states and can distinguish these patterns, but implementing it requires more than the advertised local Z measurements. This does not refute Theorem 1, which is an existence statement, but it does make the unqualified 'efficient detection using only local operations' claim unsubstantiated for w≥2 unless a correct syndrome-conditioned recovery is supplied.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a family of [[(w+1)(w+K), K]] amplitude-damping (AD) codes that generalize Shor codes, proves approximate Knill-Laflamme conditions for correcting up to w AD errors (Theorem 1), and concatenates with a dual-rail inner code to obtain [[2(w+1)(w+K), K]] constant-excitation codes immune to collective coherent errors (Corollary 2). The paper also gives explicit encoding circuits, logical operators, a two-dimensional layout, and a comparison with the Duan et al. construction. The main theorem is an existence result; the recovery described in Appendix D uses projectors onto individual error states rather than a syndrome-based decoder.","tokens_in":22323,"tokens_out":14651,"duration_ms":121743,"significance":"If the results are correct, the code family offers a rate improvement over prior AD codes (asymptotic rate 1/(w+1) versus 1/(w+1)^2 for the single-logical-qubit Shor code) and includes the Shor code and the Fletcher et al. codes as special cases. The proof of Theorem 1 is a genuine derivation with no fitted parameters, and the paper provides explicit encoding circuits, logical operator implementations, and a concrete qubit-count comparison (Table II). The dual-rail concatenation step is standard but applied cleanly. However, the advertised efficient local decoding based on Z-stabilizer syndromes is not established for w ≥ 2, which is a significant gap for the practical claims.","major_comments":[{"comment":"The claimed lookup table that maps a length-w Z-syndrome string to an AD error is not injective for w ≥ 2, so the advertised local syndrome-based decoding is unsupported. For the [[9,1]] w=2 code (the K=1 case), the error patterns {0,2} and {1} acting on the first three-qubit block both flip the two stabilizers Z0Z1 and Z1Z2, giving the same syndrome string 11; yet on |0_L> they produce orthogonal block-one states |010> and |101> (compare Table III). A decoder that assigns a single correction to syndrome 11 cannot correct both; for instance, applying the correction appropriate for the single error at position 1 to the {0,2} state leaves a logical error whose probability is O(γ^2), so the recovered fidelity would be 1 − O(γ^2) rather than the claimed 1 − O(γ^{w+1}) with w=2. The recovery in Appendix D (Eqs. D1–D2) uses projectors onto individual error states |i'^(k)> and can distinguish these patterns, but that recovery requires knowledge of the exact error pattern k and is not equivalent to the local Z-stabilizer lookup table described in the main text. The abstract's claim of 'efficient detection of AD errors using only local operations and ancillary qubits' therefore needs either a correct syndrome-conditional recovery for all patterns sharing a syndrome, an explicit extension of the syndrome with additional local measurements, or a qualified statement of what the stabilizer measurements achieve.","section":"Syndrome extraction and recovery operations (paragraphs after Eq. (8))"},{"comment":"The central estimate that the difference of diagonal terms is O(γ^{w+1}) is asserted rather than proved. The text states that each difference (α(a_w=0) − α(b_w=i)) is O(γ) and that 'by iterating this argument' the total difference is O(γ^{w+1}), but no induction is supplied. The same issue appears in Appendix A (Lemma 4), where the second step asserts orthogonality for k ≠ l from 'orthogonal error syndromes' without giving the Hamming-distance argument. Since these estimates are the core of the AQEC proof, the appendices should be expanded to include the missing formal details; as written, the phrases 'It can be shown' and 'By iterating' leave the central derivation incomplete.","section":"Appendix B, proof of Theorem 1 (paragraph beginning 'It can be shown that each difference term...')"}],"minor_comments":[{"comment":"The phrase 'Similarly, or the logical one state' should read 'Similarly, for the logical one state'.","section":"Appendix B, text near Eq. (B7)"},{"comment":"Several codeword expressions contain unmatched parentheses (e.g., the entries for (w,K)=(2,1)), which makes the definitions hard to parse.","section":"Tables III and IV"},{"comment":"The phrase 'a significant square-root improvement' is misleading: the asymptotic rate improves from 1/(w+1)^2 to 1/(w+1), which is a factor of (w+1), not a square-root scaling.","section":"Paragraph after Table I"},{"comment":"Theorem 3 is a restatement of a result from Duan et al. [19] and should be attributed to that work or proved in the appendix, since it is currently presented as a new theorem without proof.","section":"Statement of Theorem 3"},{"comment":"The weight-1 fidelity contribution appears to contain a typo: the term is written as 4[γ(1−γ)^3/2], but the standard expression is 4γ(1−γ)^3; please check the 1/2 factor and the analogous term in Eq. (C2).","section":"Appendix C, Eqs. (C1)–(C2)"}],"recommendation":"major_revision","confidential_remarks":"The manuscript's central existence result (Theorem 1) appears sound in outline, and the code construction is a useful contribution. However, the advertised local decoding claim is contradicted by the syndrome collision example for w=2, and the proof of the key O(γ^{w+1}) estimate is incomplete. Both points are fixable within the scope of the paper, so I recommend major revision rather than rejection. Please ensure the authors either provide a correct syndrome-conditional decoder or explicitly restrict the decoding claims, and complete the inductive proof in Appendices A and B."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The genuinely new thing here is the [[(w+1)(w+K), K]] interpolation between Shor's AD codes and the Fletcher–Shor–Win family, with the multi-logical-qubit AQEC proof and the dual-rail concatenation giving CC immunity. That part is worth taking seriously. The rate improvement from O(1/(w+1)^2) to O(1/(w+1)) is real, the encoding circuits are explicit, and Theorem 1 is a derivation, not a fit. The block-wise orthogonality argument in Appendix A is sound: distinct codeword terms differ in enough qubits that at most w damping events cannot make their supports overlap, and the iteration to O(γ^(w+1)) works because each difference factor contributes one power of γ. The [[6,2]] example is worked out with enough care to check out. The paper also correctly stresses that X-stabilizers are ill-suited to true AD noise and that Pauli approximations miss the energy bias; those points are fair and useful.\n\nSoft spots, in proportion. First, Appendix B's 'first and second steps follow straightforwardly' is where the general-K orthogonality actually needs proof. A Hamming-distance argument can fill it, but it should be written out rather than asserted. Second, and more serious, the main text claims a lookup table that maps a length-w Z-syndrome string to AD errors, enabling 'efficient detection using only local operations.' The stress-test note is right that for w≥2 this is not injective. In the [[9,1]] w=2 code, error patterns {0,2} and {1} in the same block share the same two-bit syndrome yet require different recoveries. A syndrome-only recovery that treats both as the single error at position 1 will apply the wrong correction to the {0,2} case, leaving a residual error of order γ^2 rather than the claimed γ^(w+1)=γ^3. The Appendix D recovery uses projectors onto individual error states, which does distinguish the patterns, but that is not the advertised local Z-measurement lookup. So the efficient decoding claim is unsubstantiated as stated. This does not refute Theorem 1, which is an existence statement about the code's AQEC conditions and appears correct, but it does mean the paper overclaims the decoding overhead.\n\nNo simulations or code are provided, and the recovery analysis for general (w,K) is sketched rather than fully derived. Still, the central construction and the main theorem are solid enough to merit serious referee time. The authors should be asked to prove the orthogonality explicitly, give a correct syndrome-conditioned recovery or clearly weaken the efficient-decoding claim, and ideally provide numerical checks for a few small (w,K) cases. This is a genuinely useful family of codes, and with those revisions it would be a good contribution.","headline":"A real new AD code family with a solid AQEC theorem, but the efficient syndrome-only decoding claim is unproven for w≥2 due to non-injective syndromes.","tokens_in":22916,"tokens_out":3269,"would_cite":true,"duration_ms":30483,"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":"A new Shor-code family approximately corrects up to $w$ amplitude-damping errors, and its dual-rail concatenation is immune to collective coherent errors.","keywords":["amplitude-damping codes","approximate quantum error correction","Shor codes","dual-rail code","collective coherent errors","constant-excitation codes","stabilizer codes","high-rate quantum codes"],"falsifier":"Compute every inner product $\\langle i|A_k^\\dagger A_l|j\\rangle$ for the $[[6,2]]$ 1-code and the $[[12,2]]$ 2-code, with $i,j$ ranging over all logical codewords and $k,l$ over all error patterns of weight at most $w$. If any off-diagonal term with $k\\neq l$ or $i\\neq j$ fails to vanish, or if any diagonal difference between $\\langle 0|A_k^\\dagger A_k|0\\rangle$ and $\\langle i|A_k^\\dagger A_k|i\\rangle$ is not of order $\\gamma^{w+1}$ at small $\\gamma$, then Theorem 1 is false. This is a finite check that settles the unproved orthogonality step.","tokens_in":21774,"feed_emoji":"🛡️","tokens_out":12613,"duration_ms":101253,"temperature":0.7,"pith_summary":"This paper proposes a family of quantum error-correction codes, the $[[(w+1)(w+K), K]]$ AD Shor codes, and claims that they approximately correct up to $w$ amplitude-damping (energy-loss) errors while achieving a code rate approaching $1/(w+1)$. The supporting proof shows that the approximate quantum error-correction conditions hold: for any two error patterns of weight at most $w$, the overlap between error-affected codewords is a logical-state-independent coefficient plus a correction of order $O(\\gamma^{w+1})$. Concatenating the same family with the dual-rail code produces $[[2(w+1)(w+K), K]]$ codes whose codewords have constant excitation, making them exactly immune to collective coherent errors while still correcting $w$ amplitude-damping events. A reader should care because realistic quantum channels suffer both energy relaxation and Hamiltonian drift, and this family addresses both with explicit local encoding, syndrome extraction, and recovery circuits.","feed_headline":"Quantum code family corrects w amplitude-damping errors","feed_subtitle":"Shor-style construction reaches rate 1/(w+1) and stays immune to collective coherent errors.","key_machinery":"The engine of the argument is a CSS (Calderbank-Shor-Steane) stabilizer code, meaning a stabilizer code whose generators are each products of only $X$-type or only $Z$-type Pauli operators. Its codewords are parity-constrained superpositions of repetition-code blocks: each of the first $w$ blocks is either $|0\\rangle^{\\otimes(w+1)}$ or $|1\\rangle^{\\otimes(w+1)}$, and the last block encodes the $K$ logical bits as $|i\\rangle^{\\otimes(w+1)}$ or $|i'\\rangle^{\\otimes(w+1)}$. The inner $(w+1)$-qubit repetition code supplies two orthogonal block states, and the operator $A_k^\\dagger A_k$ acts on each qubit as a diagonal operator with eigenvalue $1$ or $1-\\gamma$, so any block where two logical words differ contributes a factor of order $\\gamma$. With $w$ parity-constrained blocks, the accumulated difference between logical expectations is $O(\\gamma^{w+1})$. Syndrome extraction measures only the local $Z\\otimes Z$ stabilizers, yielding a single-bit syndrome adapted to amplitude damping, and recovery uses a lookup table followed by controlled-$Y$ rotations that implement an artificial damping channel.","core_discovery":"The paper's central discovery is that blockwise orthogonality inside a Shor-code structure suffices to make a whole family of high-rate codes approximate amplitude-damping correctors. Theorem 1 states that for the codewords $|i\\rangle^{(w,K)}_{\\mathrm{AD}}$ and for any two Kraus operators $A_k, A_l$ acting on at most $w$ qubits, the inner product satisfies $\\langle i|A_k^\\dagger A_l|j\\rangle = \\delta_{ij} C_{kl} + O(\\gamma^{w+1})$; this is exactly the approximate Knill-Laflamme condition needed to improve the raw transmission fidelity from $1-O(\\gamma)$ to $1-O(\\gamma^{w+1})$. The proof decomposes $A_k^\\dagger A_k$ into a tensor product of single-qubit diagonal operators and shows that each parity-constrained repetition block where two logical words differ contributes one factor of order $\\gamma$, so the full difference is $O(\\gamma^{w+1})$. Corollary 2 adds that the dual-rail concatenation is constant-excitation and hence immune to collective coherent errors, and the paper supplies local $Z$-stabilizer syndrome extraction together with a recovery operation built from controlled-$Y$ artificial damping rotations, worked out in detail for the $[[6,2]]$ code.","pith_inferences":["Inference: The unproved orthogonality step — that distinct weight-at-most-$w$ error patterns land in orthogonal subspaces inside each codeword — is the point to probe first; a direct computation of all off-diagonal inner products for the $[[12,2]]$ code would confirm or refute the family's guarantee beyond the worked examples.","Inference: Since the proof relies only on the diagonal form of $A_k^\\dagger A_k$ and on repetition-code block parity, the same construction may transfer to qudit amplitude damping or to bosonic loss channels, where the excited-state projector is replaced by the number operator.","Inference: The corrected infidelity estimate for the $[[8,1]]$ code ($6\\gamma^2$ instead of $28\\gamma^2$) suggests that adding collective-coherent-error immunity through the dual-rail code is far cheaper than earlier analyses indicated; repeated-channel numerical simulations could test this for larger members of the family."],"forward_implications":["For fixed $w$, the asymptotic code rate $1/(w+1)$ is a square-root improvement over the single-logical-qubit Shor-code rate $1/(w+1)^2$, so the family protects multiple logical qubits against weight-$w$ energy loss with proportionally fewer physical qubits.","The $[[2(w+1)(w+K), K]]$ dual-rail concatenation corrects $w$ amplitude-damping events and is immune to collective coherent rotations, so it applies to channels where Hamiltonian drift and $T_1$ decay act simultaneously.","Because only $Z$-type stabilizers are measured, the scheme avoids the fidelity loss that $X$-based measurements cause under genuine amplitude damping, where a damped state is not an eigenstate of the corresponding stabilizer.","The CSS structure and transversal logical Pauli gates make the codes compatible with fault-tolerant protocols that prevent logical operations from introducing new correlated errors.","In 12 of the 18 small parameter cases compared in the paper, the new codes require fewer physical qubits than concatenating optimal stabilizer codes with the dual-rail code."],"supporting_citations":[{"why":"Defines amplitude-damping w-codes and the dual-rail construction that Corollary 2 and Theorem 3 extend; the new family is compared against it in Table II.","marker":"[19]"},{"why":"Supplies the asymmetric high-rate [[2(K+1),K]] 1-code idea that the weight-w family generalizes.","marker":"[23]"},{"why":"Provides the original Shor-code construction and the CSS framework from which the [[(w+1)(w+K),K]] codes are built.","marker":"[21]"},{"why":"Supplies the stabilizer-code formalism and the error-correction conditions used to phrase and verify the approximate QEC statement.","marker":"[22]"},{"why":"Gives the smallest [[4,1]] AD 1-code and the recovery method, including the artificial amplitude-damping channel used in decoding.","marker":"[29]"},{"why":"States the Knill-Laflamme quantum error-correction conditions that Theorem 1 verifies in approximate form.","marker":"[15]"},{"why":"Provides the constant-excitation concatenation example and the prior infidelity estimate that Appendix C corrects.","marker":"[17]"},{"why":"Supplies an earlier amplitude-damping recovery scheme and the rate comparison against which the new family's asymptotic rate is measured.","marker":"[6]"}],"fun_headline_variants":["High-rate Shor codes fix amplitude damping and collective noise","New Shor codes: high rate, immune to collective coherent errors","Amplitude-damping codes achieve rate 1/(w+1) with dual immunity","Shor-style codes correct w AD errors at high rate, resist coherent noise","High-rate AD Shor codes with built-in immunity to coherent errors"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof assumes that distinct error patterns of weight at most $w$ map to exactly orthogonal subspaces inside each logical codeword, so that all off-diagonal Knill-Laflamme inner products vanish; this is stated as 'straightforward' in the appendix and is verified only in examples, not proved in general.","fun_headline_variants_meta":{"raw":{"variants":["High-rate Shor codes fix amplitude damping and collective noise","New Shor codes: high rate, immune to collective coherent errors","Amplitude-damping codes achieve rate 1/(w+1) with dual immunity","Shor-style codes correct w AD errors at high rate, resist coherent noise","High-rate AD Shor codes with built-in immunity to coherent errors"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000656,"raw_usage":{"total_tokens":2998,"prompt_tokens":931,"completion_tokens":2067,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":547,"completion_tokens_details":{"reasoning_tokens":1971}},"tokens_in":547,"tokens_out":2067,"duration_ms":14180,"temperature":1.0,"reasoning_tokens":1971,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T10:35:35.071266+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute every inner product $\\langle i|A_k^\\dagger A_l|j\\rangle$ for the $[[6,2]]$ 1-code and the $[[12,2]]$ 2-code, with $i,j$ ranging over all logical codewords and $k,l$ over all error patterns of weight at most $w$. If any off-diagonal term with $k\\neq l$ or $i\\neq j$ fails to vanish, or if any diagonal difference between $\\langle 0|A_k^\\dagger A_k|0\\rangle$ and $\\langle i|A_k^\\dagger A_k|i\\rangle$ is not of order $\\gamma^{w+1}$ at small $\\gamma$, then Theorem 1 is false. This is a finite check that settles the unproved orthogonality step.","supporting_citations":[{"cited_title":"Multi-error-correcting amplitude damping codes","cited_arxiv_id":null,"evidence_quote":"Supplies the asymmetric high-rate [[2(K+1),K]] 1-code idea that the weight-w family generalizes."},{"cited_title":"Assessing the performance of quantum repeaters for all phase- insensitive gaussian bosonic channels","cited_arxiv_id":null,"evidence_quote":"States the Knill-Laflamme quantum error-correction conditions that Theorem 1 verifies in approximate form."},{"cited_title":"Debroy, Muyuan Li, Michael Newman, and Kenneth R","cited_arxiv_id":null,"evidence_quote":"Supplies an earlier amplitude-damping recovery scheme and the rate comparison against which the new family's asymptotic rate is measured."}],"review_version":1}