{"id":"cb809e52-dc54-4d17-8380-d0c725807ba3","arxiv_id":"2608.13528","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Even with arbitrary ambient unitaries and ancillas, sublinear-depth nearest-neighbour circuits remain far from approximate 2-designs over the matchgate, orthogonal, and symplectic groups and from a Clifford 4-design.","lead":"This paper proves that allowing circuits to use unitaries outside a target group does not help: any shallow nearest-neighbour circuit ensemble is still far from a design over the matchgate, orthogonal, symplectic, or Clifford group. The result rules out sublinear-depth implementations of several quantum tomography and benchmarking protocols, and shows known linear-depth constructions are optimal up to constants.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Ancilla extension in Remark 1 is asserted without proof for effective CPTP channels; because the abstract's 'possibly acting on ancilla qubits' depends on it, the central no-go is not yet fully established.","rationale":"The central theorem for ensembles on a fixed Hilbert space is correct: the two-POVM argument gives the stated lower bound, and the Clifford, matchgate, orthogonal, and symplectic applications follow from the OTOC values of Ref. [4]. The Clifford subsystem-indexing typo noted by the reader is harmless; the final acceptance probability (4^(L+1)−1)/(4^n−1) is correct on the corrected reading. The single load-bearing weakness is the ancilla extension. As written, Remark 1 does not define the effective k-th moment channel after tracing ancillas, nor does it prove the key operator inequality B†(Π⊗I)B≥Q⊗I for W on H⊗H_anc. Without that, the abstract's 'possibly acting on ancilla qubits' is unsupported. I do not see a counterexample: a depth-L nearest-neighbour W should keep the Heisenberg-evolved perturbation inside H_L⊗H_anc, leaving H_barL untouched, which should imply the operator inequality and make the remark correct. But this is exactly the kind of step that should be written out. Hence conditional acceptance: require the authors to supply the missing ancilla proof, or to restrict the abstract's claim to ambient unitaries on H.","tokens_in":11485,"tokens_out":46185,"duration_ms":499453,"concrete_test":"Provide the missing ancilla proof: define M_E on H^(⊗k) by M_E(ρ)=E_W Tr_anc[W^(⊗k)(ρ⊗ρ_anc^(⊗k))W†], and verify for each W that (W(A⊗I_anc)W†)(Q⊗I_anc)(W(A†⊗I_anc)W†)≤Π⊗I_anc, using the depth-L light-cone containment of A; then check the two POVM experiments yield δ≥1−Tr[ΠΦ_μ(Ad_A(ρ_Q))]. If the inequality can be established, Remark 1 is correct and the abstract's ancilla claim stands; if a shallow W with support inside the system light cone violates it, the ancilla part of the central claim fails.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"Theorem 1 compares ensembles of unitaries on H: Phi_nu^(k)(X)=E_U Ad_U^(⊗k)(X). In the ancilla setting the sampled objects are W∈U(H⊗H_anc) and after initializing ancillas in ρ_anc and tracing them out, the relevant object is the CPTP map M_E(ρ)=E_W Tr_anc[W^(⊗k)(ρ⊗ρ_anc^(⊗k))W†] on L(H^(⊗k)), not a unitary conjugation. Remark 1 only says to extend the POVMs by identity on H_anc^(⊗k); it gives no definition of the effective channel and no proof that the two-experiment inequality survives. Concretely, Theorem 1's second experiment needs, for each W with B=W^(⊗k)(A⊗I_anc)W†, the operator inequality B†(Π⊗I_anc)B≥Q⊗I_anc on the relevant input, so that the ν-acceptance in the perturbed experiment is at least the unperturbed Q-acceptance. This is not automatic: if W moves the perturbation A into ancilla degrees of freedom, the conjugation of Q can acquire ancilla support, and the comparison with Π, which is local on the system, has to be checked separately. The remark asserts rather than proves this. Since the abstract's 'possibly acting on ancilla qubits' is part of the central claim, this omitted derivation is load-bearing. The gap appears repairable: the light cone of A under W should imply support in H_L⊗H_anc, leaving H_barL intact. But as written, the paper proves the no-go only for unitaries on H.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves no-go results for approximate designs over the matchgate, orthogonal, symplectic, and Clifford groups generated by shallow local circuits, even when the circuits are allowed to use unitaries outside the target group and, as claimed, when ancilla qubits are available. The main tool is a new lower bound, Theorem 1, on the diamond distance between the k-th moment channel of an arbitrary unitary ensemble and that of the Haar measure on a group G, obtained from a G-invariant projector Q and a lightcone-bounded POVM. Theorem 2 specialises this to the case of a maximally entangled invariant state, and Corollaries 1 and 2 follow from previously computed OTOC values for the matchgate, orthogonal, and symplectic groups. Corollary 3 treats the Clifford 4-design case using the Clifford commutant projector Q and a lightcone POVM. The authors conclude that sublinear depth is impossible for these approximate designs and that known linear-depth constructions are optimal up to constants.","tokens_in":11843,"tokens_out":17299,"duration_ms":173106,"significance":"If the claims hold, the paper closes a real loophole in earlier no-go results: prior work only ruled out ensembles whose unitaries belong to the target subgroup, leaving open constructions that use arbitrary ambient unitaries. The invariant-state-plus-lightcone framework is elegant and the main no-ancilla argument is cleanly executed. The paper is also honest about relying on prior OTOC computations in Ref. [4], and the final constant-factor optimality statements for linear-depth constructions are clearly stated. The ancilla claim, however, is currently unsupported by a proof, and the Clifford proof contains a variable typo, so the central claim is not yet fully established as written.","major_comments":[{"comment":"placeholder","section":"Section 2, Remark 1"}],"minor_comments":[],"recommendation":"major_revision","confidential_remarks":"The main no-ancilla results appear sound and publishable after revision. The ancilla gap is likely fixable with a moderate addition, and the Corollary 3 typo is clearly a slip, so I recommend a major revision rather than rejection. Please ask the authors to supply a complete proof for the ancilla case and to correct the subsystem labels in the Clifford calculation."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my take on 2608.13528.\n\nThe core result is real. Theorem 1 is a clean dichotomy: a shallow ensemble either stabilizes the invariant state, in which case the lightcone POVM catches it, or it doesn't, in which case the invariant-state POVM does. This genuinely extends Refs [4,5], which only handled ensembles supported inside the target group. The Clifford 4-design application (Corollary 3) is new and the calculation checks out; the matchgate, orthogonal, and symplectic corollaries follow quickly from OTOC values already computed in [4]. The dependency is explicit and those OTOCs aren't the contested point.\n\nWhere I'd push back: the ancilla claim. The abstract advertises 'possibly acting on ancilla qubits,' but Remark 1 is a single sentence. It doesn't define the effective channel after tracing out ancillas, and it doesn't show that the two-experiment comparison survives the trace. The stress-test note is right that this is not automatic: W^{⊗k} can move the perturbation into ancilla degrees of freedom, so the operator inequality Ad_{W^{⊗k}(A)}(Q) ≤ Π has to be checked on the extended space. I suspect the gap is repairable—the light cone of A under W should land in H_L ⊗ H_anc, leaving H_barL untouched—but as written the paper proves the no-go for unitaries on H and asserts it for ancillas. That needs fixing: either a real proof or a softened abstract.\n\nThere's also a small typo in Corollary 3's proof: the acceptance condition should be P_barL = 1_barL, not P_L = 1_L. The final bound (4^{L+1}-1)/(4^n-1) is correct, so it's an explanatory slip.\n\nBottom line: a solid, useful note for people working on matchgate shadows, fermionic tomography, and randomized benchmarking. The main theorem deserves to be in the literature. I'd send it to peer review, with a request to properly derive or explicitly caveat the ancilla extension.","headline":"Solid short note closing the ambient-unitary loophole for shallow group designs; the ancilla extension is asserted rather than proved and looks repairable.","tokens_in":12350,"tokens_out":7193,"would_cite":true,"duration_ms":58480,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P68"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that no local sublinear-depth circuit ensemble can form approximate designs over the matchgate, orthogonal, symplectic, or Clifford groups, even using ambient unitaries or ancilla qubits.","keywords":["approximate unitary designs","group designs","matchgate group","Clifford 4-design","diamond distance","lightcone argument","circuit depth lower bounds","classical shadows"],"falsifier":"Construct an explicit family of depth-L nearest-neighbour circuits on n+k qubits whose unitaries are ambient (outside the target group), initialize the ancillas in any fixed state, and compute the induced moment channel on the original n qubits; if for some L below the stated threshold the diamond distance to the group's Haar channel falls below 3/4 (or below $1 - (4^{L+1}-1)/(4^n-1)$ for the Clifford case), the paper's central claim would be false.","tokens_in":11302,"feed_emoji":"⚛️","tokens_out":9769,"duration_ms":86985,"temperature":0.7,"pith_summary":"This paper closes a loophole in recent no-go results on shallow quantum designs. Earlier work showed that sublinear-depth circuits made only of matchgate, orthogonal, symplectic, or Clifford gates cannot form approximate designs over those groups, but left open the possibility that mixing in arbitrary 'ambient' unitaries from outside the group could still produce a good design. The paper shows this does not help: no ensemble of nearest-neighbour circuits of sublinear depth can form an approximate matchgate, orthogonal, or symplectic 2-design, or a Clifford 4-design, even when the sampled unitaries are arbitrary and may act on ancilla qubits. The proof is a single invariant-state argument that covers both a shallow unitary that approximately preserves the group-invariant state and one that does not. If correct, this means protocols that sample from these groups, such as matchgate shadow tomography or benchmarking, must pay linear circuit depth rather than the logarithmic depth available for the full unitary group.","feed_headline":"Shallow circuits can't fake matchgate or Clifford designs","feed_subtitle":"Even with non-group unitaries and ancillas, matchgate, orthogonal, symplectic 2-designs and Clifford 4-designs need linear depth.","key_machinery":"The carrying object is an invariant projector $Q$: the projector onto a state or subspace fixed by the $k$-fold tensor action of the target group. For $k=2$ this is a maximally entangled state $|\\Psi\\rangle\\langle\\Psi|$ invariant under the group; for the Clifford 4-design it is the Clifford commutant projector $Q=\\frac{1}{d^2}\\sum_P P^{\\otimes 4}$. The mechanism is the dichotomy of Theorem 1: a sampled shallow unitary either leaves $Q$ approximately invariant, in which case the Heisenberg lightcone of a probe $A$ stays within a local projection $\\Pi$ and the lightcone POVM separates it from the Haar measure, or it disturbs $Q$, in which case the POVM $\\{Q,\\mathbf{1}-Q\\}$ separates it from the group's Haar measure. The out-of-time-order correlator expression of Eq. (18) turns this bound into explicit constants for each group.","core_discovery":"The central claim is Theorem 1: for any ensembles $\\mu$ and $\\nu$, any $\\mu$-invariant orthogonal projector $Q$, and any probe unitary $A$, the diamond distance between the $k$th moment channels satisfies $\\|\\Phi_\\mu^{(k)}-\\Phi_\\nu^{(k)}\\|_\\diamond \\geq 1 - \\mathrm{Tr}[\\Pi\\, \\Phi_\\mu^{(k)}(\\mathrm{Ad}_A(\\rho_Q))]$ whenever $\\mathrm{Ad}_{\\mathrm{Ad}_{U^{\\otimes k}}(A)}(Q) \\leq \\Pi$ for every $U$ in the support of $\\nu$. The proof combines two distinguishing experiments: if a sampled $U$ fails to approximately stabilize $Q$, the POVM $\\{Q,\\mathbf{1}-Q\\}$ already separates it from Haar-distributed $\\mu$; if it does stabilize $Q$, the lightcone POVM $\\Pi$ does the work, because a shallow $U$ cannot move the probe $A$ outside its Heisenberg lightcone. For the matchgate, orthogonal, and symplectic groups, with $Q$ the maximally entangled invariant state, this gives lower bounds of about $3/4$ on the diamond distance to the corresponding 2-design for depths $L\\leq n/2-1$ (matchgates) or $L\\leq n-2$ (orthogonal and symplectic). For the Clifford group, $Q = \\frac{1}{d^2}\\sum_P P^{\\otimes 4}$ is the fourth-order commutant projector and the bound is $\\|\\Phi_{E_L}^{(4)}-\\Phi_{\\mu_{Cl_n}}^{(4)}\\|_\\diamond \\geq 1 - \\frac{4^{L+1}-1}{4^n-1}$, so depth $L\\leq n-2$ is at distance at least $3/4$ from a Clifford 4-design. The paper further asserts, in Remark 1, that extending the POVMs by the identity on the ancilla registers makes the argument go through unchanged when the sampled unitaries act on an extended Hilbert space containing ancillas.","pith_inferences":["Inference: the invariant-projector dichotomy is a general template, so any compact group with a nontrivial invariant projector in a low tensor power should inherit an analogous depth obstruction against ambient shallow ensembles, not just the four groups analyzed here.","Inference: the ancilla claim in Remark 1 would be worth making explicit; without an explicit effective-channel calculation there is a small logical gap between the diamond distance on the extended space and the design error on the original system.","Inference: a natural testable extension is to probe the third moment of the Clifford group with the same lightcone argument, which could sharpen the known 'narrow failure' of the Clifford 4-design into a quantitative depth bound."],"forward_implications":["Any tomography or benchmarking protocol that assumes an approximate matchgate, orthogonal, or symplectic 2-design, or a Clifford 4-design, must use circuits of linear depth; logarithmic-depth implementations are impossible even when the sampled unitaries are not restricted to the group.","The known linear-depth constructions for Clifford circuits are optimal up to constant factors for the 4-design property, matching the new lower bound.","For the matchgate group, approximate 2-designs require depth at least roughly $n/2$, while exact Haar-random matchgates can be sampled in depth $3n$, leaving only a constant-factor window.","The result transfers to arbitrary representations: any unitary that maps a representation of one of these groups to another must pay a circuit-depth overhead governed by the design's depth in the target representation.","Approximate designs over these four groups are exponentially harder in depth than approximate designs over the full unitary group, so the naive expectation that simpler groups are easier to sample is inverted."],"supporting_citations":[{"why":"supplies the invariant-state lightcone argument and the OTOC evaluations that this paper generalizes to ambient unitaries","marker":"[4]"},{"why":"independent contemporary no-go results that this paper extends by removing the restriction to unitaries inside the target group","marker":"[5]"},{"why":"establishes that the Clifford group is a unitary 3-design, the fact used to show coset ensembles remain approximate Clifford designs","marker":"[16]"},{"why":"confirms the Clifford 3-design property used in the same coset argument","marker":"[17]"},{"why":"defines the matchgate group and its relation to fermionic linear optics, the setting of the first application","marker":"[23]"},{"why":"gives the exact depth-3n construction for Haar-random matchgates that makes the matchgate lower bound tight up to a constant","marker":"[24]"},{"why":"supplies the Weingarten calculus used to evaluate the OTOC averages for the orthogonal and symplectic groups","marker":"[25]"},{"why":"identifies the Clifford commutant projector Q and establishes that the Clifford group is not a unitary 4-design","marker":"[26]"},{"why":"gives linear-depth synthesis of Clifford circuits, establishing tightness of the Clifford lower bound","marker":"[28]"}],"fun_headline_variants":["Shallow circuits still can't make group designs","Ambient unitaries don't help shallow designs","Ancillas don't solve shallow design problem","No shallow shortcut to matchgate or Clifford designs"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that adding ancilla qubits cannot change the conclusion: the paper assumes in Remark 1 that extending the POVMs by the identity on ancilla registers makes the proof of Theorem 1 go through unchanged, without an explicit channel-level check that the diamond distance in the extended space bounds the design error on the original system.","fun_headline_variants_meta":{"raw":{"variants":["Shallow circuits still can't make group designs","Ambient unitaries don't help shallow designs","Ancillas don't solve shallow design problem","No shallow shortcut to matchgate or Clifford designs"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000294,"raw_usage":{"total_tokens":1842,"prompt_tokens":1207,"completion_tokens":635,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":823,"completion_tokens_details":{"reasoning_tokens":585}},"tokens_in":823,"tokens_out":635,"duration_ms":6515,"temperature":1.0,"reasoning_tokens":585,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T04:59:41.509799+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Construct an explicit family of depth-L nearest-neighbour circuits on n+k qubits whose unitaries are ambient (outside the target group), initialize the ancillas in any fixed state, and compute the induced moment channel on the original n qubits; if for some L below the stated threshold the diamond distance to the group's Haar channel falls below 3/4 (or below $1 - (4^{L+1}-1)/(4^n-1)$ for the Clifford case), the paper's central claim would be false.","supporting_citations":[{"cited_title":"The clifford group forms a unitary 3-design,","cited_arxiv_id":null,"evidence_quote":"establishes that the Clifford group is a unitary 3-design, the fact used to show coset ensembles remain approximate Clifford designs"},{"cited_title":"Multiqubit clifford groups are unitary 3-designs,","cited_arxiv_id":null,"evidence_quote":"confirms the Clifford 3-design property used in the same coset argument"},{"cited_title":"Integration with respect to the haar measure on unitary, orthogonal and symplectic group,","cited_arxiv_id":null,"evidence_quote":"supplies the Weingarten calculus used to evaluate the OTOC averages for the orthogonal and symplectic groups"},{"cited_title":"Linear depth stabilizer and quantum fourier transformation circuits with no auxiliary qubits in finite-neighbor quantum architectures,","cited_arxiv_id":null,"evidence_quote":"gives linear-depth synthesis of Clifford circuits, establishing tightness of the Clifford lower bound"}],"review_version":1}