{"id":"c10a91cc-f6a7-45e6-86ab-2f23d6c5367d","arxiv_id":"2607.20135","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"full","parameter_count":4,"one_line_summary":"The lamplighter group has stability radius growth ⪯ r^21 and stability rate ≥ cκ^70, giving the first explicit polynomial Hilbert–Schmidt stability bounds for an infinitely presented group.","lead":"This paper proves that approximate matrix symmetries of the lamplighter group can be corrected to exact symmetries after checking only polynomially many relations to polynomial precision. It gives the first explicit stability-radius bound for an infinitely presented group, using a new effective tower decomposition for approximately invariant measures.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified; central argument is coherent and formally supported. The Section 7 exponent mismatch is a presentation issue, patchable by taking M=C*κ^{-21}.","rationale":"The paper's main theorem is established through a clear reduction: approximate representations -> approximately equivariant PVMs -> scalar approximately invariant measures -> polynomial tower decomposition -> rounded projection towers. Proposition 4.3 is indeed the load-bearing dynamical input, as the reader says. I checked the main steps of its proof in Sections 5-6: the periodic covering (Lemma 6.6) uses the minimal-period extension and a geometric measure decay argument that is sound; the marker construction (Proposition 6.8) is a specialization of a known efficient LOCAL/descriptive combinatorics result and is plausible; the transition from markers to Kakutani-Rokhlin towers and the final splitting of bases to obtain singleton projections are consistent. Lemma 6.4 leaves two proofs to the reader, but the statements are elementary and the formalization claim provides independent support. The Section 7 exponent mismatch is real but not load-bearing: the theorem's stated radius/rate can be recovered by choosing the theorem's own M=⌈Cκ^{-21}⌉ instead of the minimal M=⌈Cκ^{-20}log⌉ used in the proof. Since the qualitative central claim - first explicit polynomial stability radius for an infinitely presented group - is unaffected and the proof is internally coherent, I do not see a reason to change the reader's CONDITIONAL verdict; the conditionality from the unverified Lean recompile and the textual exponent inconsistencies remains appropriate.","tokens_in":31946,"tokens_out":36123,"duration_ms":323193,"concrete_test":"Recompile the Lean repository at [20] and check that the proof of Proposition 4.3 (including Lemma 6.4 and Proposition 6.8) compiles with no 'sorry' or admitted axioms; then re-run the Section 7 calculation with M=⌈Cκ^{-21}⌉ to confirm the stated SRate ≥ cκ^{70} at r_κ=8M+4.","verdict_should_be":"UNCHANGED","load_bearing_attack":"No load-bearing concern. The central claim rests on Proposition 4.3; its proof is intricate but internally consistent: Lemma 6.6 covers approximately periodic points with δ-closed towers, Proposition 6.8 yields polynomial marker sets on the aperiodic part, and the final high/short tower dichotomy supplies the required height/closedness. The omitted proofs in Lemma 6.4 are routine and the Lean formalization is claimed to cover the theorem. The only real defect I find is textual: Section 7's proof chooses M_κ=C κ^{-20} log and derives SRate ≥ c κ^{67}/log^3, while Theorem 1.2 states r_κ=C κ^{-21} and SRate ≥ c κ^{70}. This does not threaten the polynomial claim: applying Theorem 1.1 with M=⌈Cκ^{-21}⌉ and Lemma 7.5 with r=8M+4 gives commutator defect ≤cκ^7/M^3 = cκ^{70}, so the stated exponents are recoverable. The manuscript should be corrected to make this parameter choice explicit.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper establishes explicit polynomial bounds for Hilbert–Schmidt stability of the lamplighter group Z/2∩Z. Theorem 1.1 asserts that unitaries A,T whose defects on a^2=1 and on [A,T^{-i}AT^i], |i|≤2M, are O(κ^7/M^2), with M=⌈Cκ^{-21}⌉, are κ-close to a genuine lamplighter representation. Theorem 1.2 translates this into SRad(r)⪙r^{21} and, at radius r_κ=Cκ^{-21}, SRate≥cκ^{70}. The proof is a transparent chain: preprocessing to commuting B_i (Lemma 4.1), passage to an approximately equivariant PVM and an approximately invariant measure on the full shift (Section 4.1), a polynomial Kakutani–Rokhlin tower decomposition (Proposition 4.3) proved via periodic extensions (Lemma 6.6) and efficient marker sets from LOCAL algorithms (Propositions 5.4 and 6.8), and finally rounding of approximate projection towers (Lemma 3.7). Appendix A gives a linear lower bound on the stability radius.","tokens_in":32166,"tokens_out":26218,"duration_ms":226175,"significance":"If correct, the paper gives the first explicit stability-radius upper bound for an infinitely presented group, answering a question from [21]. The new dynamical tools — approximately invariant measures, polynomial tower decompositions in the presence of periodic points, and efficient marker sets — are likely to be of independent interest. The main theorem is supported by an unusually detailed error budget, and the authors also provide a claimed Lean formalization of Theorem 1.1 with a comparator file [19,20]; if the formalization is independently confirmed, this is a substantial strength. I found no circularity: the external black boxes (finite abelian stability [12,13], Linial [35], Bernshteyn [10]) do not presuppose the lamplighter bound. The issues I identified below are local and do not undermine the central argument.","major_comments":[],"minor_comments":[{"comment":"The proof chooses M_κ=Cκ^{-20}∙log(2/κ) and derives SRate≥cκ^{67}/log^3, while the theorem statement claims r_κ=Cκ^{-21} and SRate≥cκ^{70}. The two are compatible (the latter radius is larger and κ^{67}/log^3≥κ^{70} for small κ), but the proof should either apply Theorem 1.1 with M∼κ^{-21} to obtain the stated exponents directly, or explicitly invoke monotonicity in r and the exponent comparison. This is a bookkeeping fix, not a substantive gap.","section":"Section 7, proof of Theorem 1.2"},{"comment":"The text claims “W_j ≠ ∅ for each j≤t” and then concludes that all towers have height at least t. The F_t-independence of Z actually gives W_j=∅ for 1≤j≤t; otherwise L^t x and L^{t+j}x are two points of Z separated by a shift j≤t. The subsequent decomposition over j=t+1,…,2t+1 confirms that this is the intended statement. Please correct the displayed claim and its conclusion.","section":"Section 6.2.3, after definition of W_j"},{"comment":"Items (i) and (ii) of Lemma 6.4 are left “for the reader.” These statements are elementary, but they are used in the load-bearing covering argument of Lemma 6.6. Since the manuscript is otherwise careful about details, include short proofs or a precise citation to keep the tower decomposition fully self-contained.","section":"Lemma 6.4"},{"comment":"In the Pythagoras estimate for ‖ρ(t^{-1})-T‖^2, the displayed formula writes the norm as a sum over tower blocks plus 2μ(e), but does not explicitly account for cross terms between P_τH and E_eH. These cross terms are controlled by Claim 4.10 (the off-diagonal blocks of T have squared norm bounded by the same tower defect sums), but the manuscript should say this explicitly rather than leaving it implicit.","section":"Proof of Theorem 1.1"},{"comment":"The computation ‖[X,Z]‖^2_HS=2 and the normalization (1/(4m))·2m appear inconsistent with the paper’s definition of normalized Hilbert–Schmidt norm on H=C^{4m}⊗C^2: with the paper’s convention one obtains a different constant (the trace of [X,Z]^*[X,Z] on C^2 is 8, so the correct constant is 2 after normalizing by dim H=8m, but the displayed intermediate formula is not the one used). The lower-bound argument itself survives with a corrected constant, but the displayed calculation should be fixed.","section":"Appendix A"},{"comment":"Lemma 4.8 refers to “Lemma 6.9 (proved in Section 5)”, but Lemma 6.9 is stated and proved in Section 6. Update the cross-reference.","section":"Cross-reference"},{"comment":"Proposition 5.4 is attributed to P. Naryshkin via a personal-communication reference. Since the proof is included in the paper, this is not a correctness issue, but the authors may wish to replace [37] with a public source or state explicitly that the proof is self-contained, so that readers are not dependent on a personal communication.","section":"Reference [37]"}],"recommendation":"minor_revision","confidential_remarks":"The paper is in good shape. The central reduction is coherent, the claimed formalization is a notable strength, and the issues I found are local: a parameter mismatch in Section 7, a couple of typos in Section 6.2.3 and Appendix A, and an omitted sentence in the final Pythagoras estimate. I saw no circularity or load-bearing error. The only bibliographic weakness is the personal-communication reference [37]; the proof is included, so this is not blocking. The A.I. disclosure is transparent and, in my view, does not raise concerns about the mathematical content."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The headline: this paper gives the first explicit polynomial bound on Hilbert-Schmidt stability radius and rate for an infinitely presented group, and the main argument holds up. I went in looking for a gap around Proposition 4.3, but the tower decomposition is dense, coherent, and the dependency chain from approximately invariant measures to corrected representations is sound.\n\nWhat is genuinely new: the quantitative stability theorem for Z/2≀Z, with stability radius O(r^21) and stability rate at least cκ^70. Prior work had only qualitative stability for the lamplighter. The machinery—approximately equivariant PVMs, polynomial Kakutani-Rokhlin towers for approximately invariant measures, and the use of Linial and Bernshteyn to get efficient marker sets—is a real addition, and it should transfer to other metabelian groups of the form N⋊Z. The reported Lean formalization of Theorem 1.1 is also strong evidence; I have not recompiled it, but the claim is concrete and the repositories are referenced.\n\nSoft spots are minor. Section 7's proof derives the rate with M_κ=⌈Cκ^{-20} log⌉ and obtains κ^{67}/log^3, while Theorem 1.2 states κ^{-21} and κ^{70}. The stress-test is right that this is patchable by choosing M=⌈Cκ^{-21}⌉, but the manuscript should be corrected so the stated exponents match the proof. Lemma 6.4(i),(ii) are left to the reader and are used in the periodic-tower construction; they look routine, but in a proof this intricate, leaving them as exercises is mildly annoying. The AI disclosure is transparent and does not bother me; the authors say they verified the LLM's contribution.\n\nNo circularity or data-fitting red flags: the parameters are chosen from the proof, not fitted to the answer, and the authors supply a linear lower bound in Appendix A, so the polynomial upper bound is not overclaimed.\n\nWho should read it: anyone working on quantitative stability, and the descriptive combinatorics community will be interested in the efficient marker lemma. It deserves a serious referee—not a desk reject. I would send it to review, asking the referee to read Section 6 carefully and to check parameter choices in Sections 4 and 7.","headline":"First explicit polynomial Hilbert-Schmidt stability bounds for the lamplighter group; the proof is sound, with a patchable exponent typo in Section 7.","tokens_in":32707,"tokens_out":2469,"would_cite":true,"duration_ms":24340,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["22D10","20E22","37B10"],"pacs":[],"model":"deepseek-v4-flash","headline":"The lamplighter group Z/2 ≀ Z is polynomially Hilbert–Schmidt stable: checking O(κ^{-21}) relations at precision O(κ^{7}/M^2) forces any approximate representation to be κ-close to a true one.","keywords":["Hilbert–Schmidt stability","lamplighter group","stability radius","stability rate","quantitative stability","approximately invariant measures","Kakutani–Rokhlin towers","descriptive combinatorics"],"falsifier":"Exhibit a family of unitaries (A, T) on growing Hilbert spaces with ∥A²−1∥_{HS} ≤ cκ^7/M² and ∥[A, T^{-i}AT^i]_{HS} ≤ cκ^7/M² for M = Cκ^{-21}, yet kept at HS distance ≥ cκ from any true lamplighter representation — such a family would disprove Theorem 1.1. A cheaper witness: construct an (M,η)-invariant measure on the shift with η ≤ cκ^{14} for which every clopen tower decomposition has a base of superpolynomial complexity in 1/κ, directly contradicting Proposition 4.3; the uniform measure on a long periodic orbit is a natural candidate to test.","tokens_in":31744,"feed_emoji":"🏮","tokens_out":9189,"duration_ms":76484,"temperature":0.7,"pith_summary":"The paper proves the first explicit polynomial bounds on the Hilbert–Schmidt stability of an infinitely presented group, specifically the lamplighter group Γ = Z/2 ≀ Z. It shows that if unitaries A, T satisfy the lamplighter relations to precision ε — with the relations tested only over a window of radius M = O(κ^{-21}) — then A and T are κ-close to genuine lamplighter operators. Equivalently, the stability radius grows at most like r^{21} and the stability rate is at least cκ^{70}. The proof builds a new quantitative bridge from approximate representations to approximate invariant measures on the two-sided shift, decomposes such measures into polynomial-complexity towers, rounds the induced approximate projection towers to exact ones, and reassembles a true representation.","feed_headline":"κ^{-21} checks: lamplighter group is stable","feed_subtitle":"First explicit stability-radius bound for an infinitely presented group: check O(κ^{-21}) relations to get κ-close to a true representation.","key_machinery":"The load-bearing mechanism is an effective tower decomposition for approximately invariant measures (Proposition 4.3). Working on the full shift X={0,1}^Z, it takes any (M,η)-invariant probability measure and returns a partition of X, up to a clopen error set of measure O(t^6(ν+δ+η)), into clopen towers of polynomial complexity; each tower is either δ-closed of height less than t or has height between t and 6t+1, with a singleton projection on its base. Around this hinge the paper uses four tools: Fourier passage from approximate representations to approximately equivariant projection-valued measures; a rounding lemma (Lemma 3.7) that turns an approximate closed projection tower into an exac","core_discovery":"On the authors' own terms, the paper establishes Theorem 1.1: for κ > 0, with M = ⌈Cκ^{-21}⌉ and ε = cκ^7/M², any unitaries A, T on a finite-dimensional Hilbert space satisfying ∥A²−1∥_{HS} ≤ ε and ∥[A, T^{-i}AT^i]_{HS} ≤ ε for all 0 ≤ i ≤ 2M can be moved within HS distance κ of operators that form a true unitary representation of the lamplighter. Theorem 1.2 restates this as stability radius SRad(r) ⪯ r^{21} and stability rate SRate(κ) ≥ cκ^{70}, which the authors present as the first explicit upper bound on the stability radius of an infinitely presented group, answering a question in [21, Question 12.10], and note the proof also works for A ≀ Z with finite abelian A.","pith_inferences":["Editorial: the polynomial tower decomposition is likely the reusable core of the method; other metabelian groups N ⋊ G with a well-behaved shift action may yield quantitative stability by constructing their own tower decompositions.","Editorial: the exponents in Theorem 1.2 are almost certainly not optimal; a tighter marker lemma or a sharper height-versus-closedness tradeoff could lower r^{21} toward the linear lower bound shown in the appendix.","Editorial: the same machinery might yield effective bounds for permutation stability of the lamplighter if a Hamming-norm analogue of the projection-tower rounding lemma can be found — a direction the paper explicitly leaves open.","Editorial: the explicit error stack (ε = cκ^7/M², M = Cκ^{-21}) suggests a testable check of Proposition 4.3 by direct computation with the Bernoulli measure on the shift, where the error bound t^6(ν+δ+η) should be verifiable explicitly."],"forward_implications":["The lamplighter group becomes the first known infinitely presented group whose stability radius admits an explicit upper bound, sharpening the earlier qualitative stability result.","Quantitative Hilbert–Schmidt stability, previously known only for finite and abelian groups, now extends to a nilpotent-wreath family; the proof carries over to A ≀ Z for any finite abelian A.","The proof supplies a template: any group action whose approximately invariant measures admit polynomial-complexity tower decompositions inherits polynomial HS stability bounds.","The linear lower bound on the stability radius recorded in Appendix A shows that the true growth exponent, whatever it is, lies between 1 and 21."],"fun_headline_variants":["κ^{-21} checks yield lamplighter stability","First explicit stability-radius bound for lamplighter","Lamplighter group: polynomial HS stability","Explicit bound: lamplighter is κ^{-21}-stable","Lamplighter stability: answering the open question"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"Everything rests on Proposition 4.3: that every approximately invariant measure on {0,1}^Z admits a polynomial-complexity partition into clopen towers with the stated height-versus-closedness guarantee; if any (M,η)-invariant measure escapes such a decomposition, the proof cannot deliver the corrected representation.","fun_headline_variants_meta":{"raw":{"variants":["κ^{-21} checks yield lamplighter stability","First explicit stability-radius bound for lamplighter","Lamplighter group: polynomial HS stability","Explicit bound: lamplighter is κ^{-21}-stable","Lamplighter stability: answering the open question"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000134,"raw_usage":{"total_tokens":939,"prompt_tokens":673,"completion_tokens":266,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":417,"completion_tokens_details":{"reasoning_tokens":201}},"tokens_in":417,"tokens_out":266,"duration_ms":3304,"temperature":1.0,"reasoning_tokens":201,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T10:42:36.108642+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Exhibit a family of unitaries (A, T) on growing Hilbert spaces with ∥A²−1∥_{HS} ≤ cκ^7/M² and ∥[A, T^{-i}AT^i]_{HS} ≤ cκ^7/M² for M = Cκ^{-21}, yet kept at HS distance ≥ cκ from any true lamplighter representation — such a family would disprove Theorem 1.1. A cheaper witness: construct an (M,η)-invariant measure on the shift with η ≤ cκ^{14} for which every clopen tower decomposition has a base of superpolynomial complexity in 1/κ, directly contradicting Proposition 4.3; the uniform measure on a long periodic orbit is a natural candidate to test.","supporting_citations":[],"review_version":1}