{"id":"e117bbfd-4334-4bec-8117-6b2bfea5c514","arxiv_id":"2509.09047","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Multi-qubit golden and super-golden gate sets, including 2-qubit Clifford+CS, are constructed with provably optimal covering rates up to polylog factors, giving 4.8x to 10x asymptotic reductions in expensive gate count.","lead":"The paper constructs new gate sets for approximating quantum operations on two and three qubits, including one built from the standard Clifford plus controlled-S gates, and proves their approximation rate is optimal up to log factors. If the proof holds, these sets would use roughly five to ten times fewer expensive non-Clifford gates than the standard Clifford+T approach.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Optimal covering rests on unpublished 'KMSb' companion to [KMSW14]; if that inner-form classification is unavailable, Theorem 1.2.5's 4.8x claim is unproven.","rationale":"The reader's weakest assumption correctly identifies the unpublished 'KMSb' companion to [KMSW14] as the load-bearing external premise for the optimal covering property. My reading of the paper confirms that Theorem 1.3.2 and its application in §8 depend on the inner-form endoscopic classification, and the paper itself flags the unpublished dependency in its Conditionality paragraph. There is no internal contradiction or fitting-as-derivation; the gap is explicit and external. The reader's CONDITIONAL verdict already accurately reflects this: the central claim is plausible and well supported except for the missing reference. I see no reason to move the verdict. A concrete dependency audit, as proposed, would either resolve the conditionality or firmly establish the unproven premise, so the recommendation remains UNCHANGED at this stage.","tokens_in":73049,"tokens_out":5873,"duration_ms":62316,"concrete_test":"Perform a dependency audit: list each statement from [KMSW14] used in Theorem 7.1.1 (via [DGG24, Thm 6.5.1]), Theorem 7.2.5, and Corollary 7.3.2, and determine which require the unpublished “KMSb.” Then check whether the specific n=4,8, F=Q, E=Q(i) or Q(√−3) cases can be derived using only the published quasi-split classification [Mok15] plus [AGI+24] (e.g., by verifying that the local packets at the relevant primes are either unramified or covered by Corollary 6.3.6). If every needed statement is available from published sources, the conditionality is resolved; if not, the paper should explicitly record the missing inner-form results as an unproven premise.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The optimal covering theorem (Thm 8.0.1/Prop 8.3.3) is the signature property of Theorem 1.2.5, and it is gated on the density hypothesis Theorem 1.3.2. The proof of Theorem 1.3.2 in §§5–7 imports the endoscopic classification of [KMSW14] for non-quasisplit (inner) forms of unitary groups—used in Theorem 5.2.2 and the input bound Theorem 7.1.1. The paper's Conditionality paragraph explicitly states that [KMSW14] depends on the unpublished weighted twisted fundamental lemma and, in addition, pushes technical details to the unpublished companion “KMSb.” The definite unitary groups used in Theorems 1.2.3–1.2.5 are inner forms, so the full strength of the inner-form classification is invoked. [AGI+24] resolves the dependence on the unitary analogues of Art13's A25–27, but it does not, per the paper's own note, remove the KMSb dependence. Thus, if KMSb is unavailable or incorrect, the bound on the automorphic family in Theorem 7.1.1 is unsupported, and the optimal covering property—the property that gives the 4.8×/10× asymptotic comparisons—is unproven. The other three properties (growth, navigation, heuristic approximation) do not rely on this input. This is an explicit external dependency, not an internal inconsistency.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper extends the golden/super-golden gate constructions of Sarnak and Parzanchevski from PU(2) and PU(3) to the multi-qubit cases PU(4) and PU(8). The authors define arithmetic 'golden adelic subgroups', give several explicit constructions over Q(√-3), Q(i), Q(√-7), and Q(√-2), and in particular show that the 2-qubit Clifford group together with the CS gate is a super-golden gate set. The optimal covering property is derived from a Sarnak-Xue density hypothesis in the weight aspect, proved via the endoscopic classification of automorphic representations for definite unitary groups. The paper claims an asymptotic 4.8x reduction in non-Clifford gates over Clifford+T and a 10x reduction for another example. The proof is explicitly conditional on the unpublished companion 'KMSb' of [KMSW14] for inner forms, and it also relies on several asserted finite computer checks that are not accompanied by code or verification data.","tokens_in":73247,"tokens_out":6183,"duration_ms":69580,"significance":"If the full proof chain is eventually completed, this would be a significant step: it provides the first optimal multi-qubit golden gate sets, matches the worst-case lower bound of [GRT21], and demonstrates a nontrivial use of modern Langlands machinery in quantum computation. The paper is unusually transparent about its external dependencies, and the structural framework separating growth, navigation, approximation, and covering is useful. However, the signature property—optimal covering—is gated on an unpublished inner-form endoscopic classification, and several load-bearing numerical checks are asserted without reproducibility. For these reasons the central theorems should presently be regarded as conditional, and the manuscript requires substantive revision before the claims can be accepted as stated.","major_comments":[{"comment":"The optimal covering property, and hence the 4.8x/10x asymptotic claims, rests on Theorem 1.3.2, whose proof in §§5-7 imports the endoscopic classification of [KMSW14] for non-quasisplit inner forms. The paper's own Conditionality paragraph states that [KMSW14] depends on the unpublished weighted twisted fundamental lemma and that technical details are pushed to the unpublished companion 'KMSb'. Since the definite unitary groups used in Theorems 1.2.3-1.2.5 are inner forms, the full force of this input is invoked. The abstract and theorem statements nevertheless present the results as unconditional. This is a load-bearing external dependency, not a presentation issue: if KMSb is unavailable or incorrect, the density hypothesis, and with it the optimal covering property, is unproven. The paper should either supply the missing input, or explicitly restate the affected theorems as condition","section":"Conditionality paragraph, §1.2, Theorems 1.2.3/1.2.5, Theorem 1.3.2, Theorem 7.1.1, Theorem 5.2.2"},{"comment":"Several load-bearing claims are justified only by 'computer check' without code, scripts, or independent verification data. In particular, the class-number-one mass computations (e.g., |G(Z)| = 155520 in Proposition 4.5.4, |G(Z) ∩ K'| values in Propositions 4.5.5-4.5.14), the group-intersection checks, and the finite inequalities in the proof of Theorem 7.3.1 ('By a computer check ...') and Corollary 7.3.4 are essential to the explicit gate-set constructions and to the density bound. A referee cannot certify these claims from the manuscript alone. The authors should provide executable code or detailed certifiable verification, or at minimum specify the exact computations in a way that permits independent replication.","section":"Propositions 4.5.4, 4.5.5, 4.5.6, 4.5.10, 4.5.11, 4.5.13, 4.5.14; proof of Theorem 7.3.1; Corollary 7.3.4"},{"comment":"The density hypothesis Corollary 7.3.2 is stated as applying in a list of cases that includes 'Conjecture 6.3.4 holds for Arthur-type representations of G_v with a K'_v-fixed vector.' For n = 4 this is resolved by Corollary 6.3.6, and for n = 8 by the computer-assisted Corollary 7.3.4. However, the organization makes the logical dependence hard to track: Theorem 7.3.1 appears to prove the exponent inequality for all shapes, while the local exponent bound needed for non-split v is conditional on Conjecture 6.3.4. The authors should make the final unconditional range of validity explicit in the statement of Corollary 7.3.2, since as written the bullet list can be read as including an unproved conjecture as a hypothesis rather than as a resolved case.","section":"Theorem 7.3.1 and Corollary 7.3.2, with Conjecture 6.3.4"}],"minor_comments":[{"comment":"The symbol B is used both for a matrix in Theorem 1.2.3 and for the Bruhat-Tits building throughout the paper (per Notation 1.5.2). This is confusing; consider renaming the matrix.","section":"Theorem 1.2.3 and Notation 1.5.2"},{"comment":"The comparison in §4.6 writes 'B(A, ε^{15})' while also speaking of 'within distance ε'. Definition 1.2.1 defines B(x,ε) as the ball of volume ε, so the relation between volume ε^15 and metric distance ε should be stated explicitly to avoid ambiguity.","section":"§4.6 and Definition 1.2.1"},{"comment":"The table refers to added gates such as T'_G, T_E,2, T_E,3 without displaying their definitions in the table itself; the reader must return to the corresponding propositions. A pointer is fine, but the cost model used for 'R' and 'covering efficiency' should be stated once in the caption or preceding text.","section":"Table 4.1"}],"recommendation":"major_revision","confidential_remarks":"The KMSb dependence is a genuine scope and novelty issue: the main number-theoretic theorem, and therefore the headline quantum-computing comparison, is conditional on an unpublished companion paper. I would recommend that the editor require the authors either to make the missing input available or to recast the main theorems as conditional results. The computational reproducibility issue is also important for a paper of this level of technical ambition."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is the real thing. It constructs the first golden and super-golden gate sets for PU(4) and PU(8), including the practically relevant 2-qubit Clifford+CS set, and proves a weight-aspect Sarnak–Xue density hypothesis for definite unitary groups via the endoscopic classification. The framework—Cartan norms on Bruhat–Tits buildings, Mœglin exponents, Hecke operator bounds—is coherent, and the paper is unusually explicit about its own dependencies. That matters here, because the central claim is conditional.\n\nWhat's genuinely new: Theorem 1.2.5 identifies Clifford+CS as super-golden, giving a tight upper bound on CS-count matching the GRT21 lower bound, with a claimed 4.8x asymptotic reduction in non-Clifford gates versus Clifford+T. The PGSp4(F3) example and the PU(8) constructions are also new. The proof chain for growth, navigation, and approximation is self-contained and sound; soft-O notation checks out, and the growth exponents come from root-system data rather than fitted parameters.\n\nThe soft spot is real and the stress-test note is correct: the optimal covering property—the property that gives the 4.8x/10x constants—is gated on Theorem 1.3.2, whose proof imports the inner-form endoscopic classification of [KMSW14]. The paper's own Conditionality paragraph states that [KMSW14] depends on the unpublished weighted twisted fundamental lemma and pushes details to an unpublished companion 'KMSb'. If that input is unavailable or wrong, the covering theorem is unsupported. This is an explicit external dependency, not an internal inconsistency, but it is load-bearing. The other three properties (growth, navigation, approximation) do not depend on it.\n\nTwo smaller issues: several decisive finite checks (Props 4.5.4–4.5.16, Cor 7.3.4) are asserted as computer runs without code or certificates, which makes the n=8 extension harder to verify independently; and the approximation property is heuristic by design, though the paper says so plainly. None of this is hidden.\n\nWho should read it: quantum computing people who care about gate-set efficiency, and automorphic forms people interested in Sarnak–Xue density hypotheses. It deserves a serious referee. My recommendation: send it to peer review, and in the report insist that the KMSb dependency either be resolved, made public, or clearly separated into a conditional theorem—and that the computer checks be released. The n=4 results are strong enough that the paper should not be desk-rejected even if the unpublished reference is slow to appear.","headline":"First real multi-qubit golden/super-golden gate sets, including Clifford+CS, with a coherent automorphic proof chain—but the signature covering claim rests on an unpublished companion paper (KMSb).","tokens_in":73939,"tokens_out":1552,"would_cite":true,"duration_ms":21252,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11F70","11F72","22E55","81P68"],"pacs":[],"model":"deepseek-v4-flash","headline":"Two-qubit Clifford+CS is a super-golden gate set, cutting non-Clifford cost 4.8x.","keywords":["golden gates","super-golden gates","two-qubit quantum computation","Clifford+CS","optimal covering","automorphic density hypothesis","endoscopic classification","Bruhat-Tits buildings"],"falsifier":"Check the unpublished companion reference for the weighted twisted fundamental lemma: if that lemma fails, or if an explicit automorphic representation of a definite U(4) with regular integral infinitesimal character is found whose local matrix-coefficient decay at a prime violates the bound of Theorem 6.4.2, the density hypothesis collapses and with it the optimal covering theorem. Concretely, one could compute the covering radius of the Clifford+CS set at increasing lengths and look for a missed ball of volume larger than (log|S^(l)|)^c/|S^(l)|.","tokens_in":1586,"feed_emoji":"🎯","tokens_out":4465,"duration_ms":116970,"temperature":0.7,"pith_summary":"This paper constructs golden and super-golden gate sets for two- and three-qubit unitary groups, extending earlier one-qubit constructions. Its central theorem is that the controlled-S gate together with the two-qubit Clifford group is a super-golden gate set for PU(4): words of length l cover the group up to an epsilon_l = (log|S^(l)|)^c/|S^(l)| volume, the number of words grows exponentially, shortest words can be found efficiently, and a heuristic algorithm approximates any unitary. If correct, arbitrary two-qubit unitaries are approximable with CS-count within a polylog factor of the optimal covering rate, matching the known worst-case lower bound and using 4.8x fewer non-Clifford gates asymptotically than Clifford+T; another explicit example heuristically saves about 10x on T-type gates. The construction works through golden adelic subgroups of definite arithmetic unitary groups, with optimal covering supplied by a weight-aspect density hypothesis for automorphic representations proven via endoscopic classification. The paper flags that both the density hypothesis and its applications depend on an unpublished weighted twisted fundamental lemma and a companion reference that is not yet publicly available.","feed_headline":"Clifford+CS cuts two-qubit non-Clifford cost by 4.8x","feed_subtitle":"Two-qubit unitary approximations using CS gates hit the optimal covering rate up to a log factor, beating Clifford+T asymptotically.","key_machinery":"The load-bearing object is the golden adelic subgroup: a finite-index subgroup K' of the integral points of a definite unitary group G satisfying G(A)=G(F)G_infinity K' and G(F) intersection K' = {1} (or the weaker 'almost' versions). This class-number-one structure makes the p-arithmetic subgroup Lambda_p act simply transitively on the vertices (or chosen edges) of the Bruhat-Tits building of G at p, so the gates can be taken as the building's nearest neighbors measured by a modified Cartan norm and words of length l are exactly points at distance l. Growth, navigation, and approximation follow from this building picture together with a higher-dimensional version of the efficient one-qubit","core_discovery":"On the paper's own terms, the discovery is that optimal gate sets for multi-qubit computation can be built from arithmetic groups with the same 'golden' property that made one-qubit gate sets work. Specifically, CS plus the two-qubit Clifford group is super-golden in PU(4): it has optimal covering rate up to polylog, exponential growth, navigation, and heuristic approximation, and the same template yields a PU(4) set whose added gate generates PGSp4(F3) with a heuristic 10x T-gate saving, and a PU(8) golden set for three qubits. The optimal covering bound is a theorem conditional on the paper's density hypothesis, which is argued through the endoscopic classification of automorphic represent","pith_inferences":["Once the referenced companion proof is public, the asymptotic savings become unconditional; until then the 4.8x/10x constants should be read as conditional on the density hypothesis.","The construction suggests a general recipe: search class-number-one definite unitary groups for special or edge stabilizers; each gives a candidate super-golden set with growth base determined by the building's geometry, and the main bottleneck for new examples is the automorphic density bound, not the building theory.","A practical testable extension is to optimize the approximation algorithm's constant factors; the paper itself notes these are far from optimal, so the theoretical 4.8x/10x savings may or may not survive in actual circuit synthesis.","The framework likely transfers to other Clifford-hierarchy gates or cyclotomic gate sets whenever the associated arithmetic group has class number one; the examples suggest lattices related to the E8 lattice in Q(sqrt(-3)) and Q(i) are a rich source of such sets."],"forward_implications":["Every two-qubit unitary can be approximated to accuracy epsilon by a product of CS gates with length O(log(1/epsilon)) up to a polylog factor, matching the worst-case lower bound.","The Clifford+CS set uses asymptotically 4.8x fewer non-Clifford gates than Clifford+T; the PGSp4(F3) example gives a heuristic ~10x saving.","The same arithmetic-group construction yields golden gate sets on PU(8), giving three-qubit gate sets with optimal covering.","Gate sets come with a polynomial-time navigation algorithm that finds shortest words from the Cartan decomposition, and with a heuristic approximation algorithm valid in higher dimension.","Any finite group plus extra finite-order element arising from a golden or super-golden adelic group inherits the same optimal covering, so the framework applies to other fault-tolerance-friendly gates at the third level of the Clifford hierarchy."],"fun_headline_variants":["Golden gates go multi-qubit: CS cuts non-Clifford cost 4.8x","Multi-qubit golden gates: 4.8x fewer non-Clifford gates","Super-golden Clifford+CS: tight 4.8x bound for two qubits","Two-qubit golden gates: 4.8x fewer expensive gates than T"],"cache_read_input_tokens":75008,"weakest_assumption_plain":"The optimal covering property, and with it the asymptotic 4.8x/10x gate-count savings, rests on the density hypothesis, whose proof depends on an unpublished weighted twisted fundamental lemma and a companion reference that is not yet publicly available; without that input, the remaining properties still hold but the covering claim is unproven.","fun_headline_variants_meta":{"raw":{"variants":["Golden gates go multi-qubit: CS cuts non-Clifford cost 4.8x","Multi-qubit golden gates: 4.8x fewer non-Clifford gates","Super-golden Clifford+CS: tight 4.8x bound for two qubits","Two-qubit golden gates: 4.8x fewer expensive gates than T"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001184,"raw_usage":{"total_tokens":4723,"prompt_tokens":735,"completion_tokens":3988,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":479,"completion_tokens_details":{"reasoning_tokens":3894}},"tokens_in":479,"tokens_out":3988,"duration_ms":32280,"temperature":1.0,"reasoning_tokens":3894,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T19:48:25.630839+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Check the unpublished companion reference for the weighted twisted fundamental lemma: if that lemma fails, or if an explicit automorphic representation of a definite U(4) with regular integral infinitesimal character is found whose local matrix-coefficient decay at a prime violates the bound of Theorem 6.4.2, the density hypothesis collapses and with it the optimal covering theorem. Concretely, one could compute the covering radius of the Clifford+CS set at increasing lengths and look for a missed ball of volume larger than (log|S^(l)|)^c/|S^(l)|.","supporting_citations":[],"review_version":1}