{"id":"8d760b66-c1f3-4d28-a443-a9d8b4cd5034","arxiv_id":"2507.16985","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A complete classification of ω-categorical structures with unlabelled growth below 2^n/p(n), confirming Thomas' conjecture and giving optimal growth gaps for this class.","lead":"This paper classifies all countable structures whose number of distinct n-element subset types grows more slowly than 2^n divided by any fixed polynomial. The classification implies such structures are rare, all built from the rational order, and each has only finitely many first-order reducts.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 4.2 is applied in Theorem 4.4 to restrictions not yet known to be closed, making the stabilization/closedness argument circular.","rationale":"The reader's weakest assumption was the unproved import of Lemma 4.1 from [Sim18b]. After reading the manuscript closely, the most load-bearing spot is one step further inside Theorem 4.4: even granting Lemma 4.1, its proof applies Lemma 4.2 to groups whose closedness has not been established, and Lemma 4.2's proof relies on closedness to pass from isomorphism to equality. This is a concrete, local gap in the chain that produces finite fiber factors. It does not amount to a refutation of the central classification; rather, it identifies a missing proof obligation. Since the reader's verdict was CONDITIONAL and this concern is a specific instance of the same unverified external input, the appropriate verdict is unchanged: the paper should be accepted only conditionally on closing this gap, ideally by making the relevant part of [Sim18b] fully self-contained or by proving that pointwise-stabilizer restrictions of closed group covers are closed.","tokens_in":46266,"tokens_out":8734,"duration_ms":111169,"concrete_test":"Settle whether restrictions of the form (G_{B,{C}})|C arising in Theorem 4.4 are automatically closed. Concretely: using the normal form of Theorem 5.2 for closed finite covers, determine whether a finite cover of Aut(Q;<) with the same fibers can contain a proper dense subgroup that is also a finite cover of Aut(Q;<) with the same fibers. If such a subgroup exists, Lemma 4.2 cannot be applied to the non-closed restrictions and Theorem 4.4 has a genuine gap; if no such dense subgroup exists, the closedness assumption is automatic and the proof can be repaired by adding a one-line justification. Either outcome is verifiable from the explicit F/E construction in [Sim18b, Thm 5.1] and the strong-split normal forms in Section 5.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The classification enters through Theorem 4.4, which upgrades the cover given by Lemma 4.1 to one with finite fiber factors. The proof fixes finite sets A_i exhausting X \\ C and considers H_i := (G_{A_i,{C}})|C. It asserts that the H_i are all in F and hence stabilize by Lemma 4.2. But Lemma 4.2 is a statement about F, the class of closed finite covers of highly set-transitive groups. Directly after Lemma 4.1, the paper explicitly says that closedness of these restrictions does not follow from [Sim18b]. No argument is supplied before Theorem 4.4 proving that each H_i is closed. Moreover, the proof of Lemma 4.2 uses closedness essentially: after observing that there are finitely many isomorphism types, it concludes that an isomorphic proper inclusion cannot occur because two closed oligomorphic groups with the same n-orbits are equal. Without closedness, an infinite descending chain of pairwise isomorphic but distinct finite covers with the same fibers is not excluded. Thus the stabilization step in Theorem 4.4 has a missing premise. Since Theorem 4.4 is what supplies the finite fiber factors needed to invoke the L(\\tilde G, G*, D) reconstruction of Section 3, this gap is load-bearing: if some H_i can fail to be closed, the cover need not have finite fiber factors, and the classification in Theorem 7.10, the gap theorem, and the proof of Thomas' conjecture for S do not follow.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper classifies the class S of countable structures whose unlabelled orbit growth is not at least 2^n/p(n) for any polynomial p. The main theorem, Theorem 7.10, characterizes automorphism groups in S as builds from finite highly set-transitive groups, finite covers of reducts of (Q;<), and hereditarily cellular groups, using finite covers, wreath products with Sym(omega), finite direct products, and finite-index or closed supergroups. From this classification the paper derives a gap theorem for growth rates (Theorem 1.5), finite homogenizability and finite boundedness, interpretability in (Q;<), countability up to bidefinability, and Thomas' conjecture for S (Theorem 9.11).","tokens_in":46549,"tokens_out":28271,"duration_ms":316255,"significance":"If the proof is completed, this is a substantial contribution: it supplies a full structural description in the c<2 range of unlabelled growth, confirming two conjectures from Braunfeld and generalizing the classifications in [FT20] and [Bod24]. The paper is careful about what is imported from earlier work, and the main classification is concrete and falsifiable rather than an existence result. The theorem gives explicit sufficient conditions for membership in S_d, and the growth gap theorem is a sharp quantitative consequence. These strengths make the paper worth publishing, provided the load-bearing gap identified below is repaired.","major_comments":[{"comment":"The stabilization argument applies Lemma 4.2 (= Corollary 5.3) to the sequence (G_{A_i,{C}})|C. Lemma 4.2 is a statement about the class F of closed finite covers of highly set-transitive groups. Immediately after Lemma 4.1 the paper explicitly notes that closedness of these restrictions does not follow from [Sim18b], and no argument is supplied before Theorem 4.4 showing that each (G_{A_i,{C}})|C is closed. Without closedness, an infinite descending chain of pairwise isomorphic non-closed groups with the same orbit partitions need not stabilize, since their closures can be equal while the groups themselves descend properly. Because Theorem 4.4 is what converts the cover from Lemma 4.1 into one with finite fiber factors, and Theorem 7.10, Theorem 1.5, and Theorem 9.11 all rely on that conversion, this is a load-bearing gap. The proof needs either a proof that the restrictions are closed, or a version of Lemma 4.2 for finite covers not assumed closed.","section":"Section 4, Theorem 4.4"},{"comment":"The proof asserts that the lifted triple (K,∇,∆) is an omega-partition of G and says this is clear from the definition, but condition (5) of Definition 2.18 — that G((C))/∆ = Sym(C/∆) for every ∇-class C — is not verified. This condition is needed for the induction step and hence for Lemma 6.3 and for the description of finite covers in Theorem 7.10(2). The gap is likely repairable by observing that the quotient G/∆ maps onto G*/∆*, but as written the verification is omitted and should be supplied explicitly.","section":"Section 6, Lemma 6.1"}],"minor_comments":[{"comment":"The recursive definition appears internally inconsistent: H_{-1}=∅ makes H_0 empty by the recursion, while Remark 2.17 says H_0 is exactly the class of finite-degree groups. The base case of the recursion should be corrected.","section":"Definition 2.16"},{"comment":"The proof cites 'Lemma 5.9' before that lemma is stated; it should cite Lemma 4.3, which is the version stated earlier.","section":"Theorem 4.4 proof"},{"comment":"Both proofs refer to 'Lemma 8', which does not exist in the manuscript; presumably Lemma 5.21 is intended.","section":"Lemma 7.7 and Theorem 7.10"},{"comment":"The polynomial is written f(x)=x^n - Σ_{i=1}^{k-1} c_i x^i, which does not match the displayed recursion an+k = Σ_{i=0}^{k-1} c_i an+i; the exponents in the polynomial should match the order k of the recurrence.","section":"Lemma 5.20"},{"comment":"The title contains a typo ('f ast'), and the heading 'F acts 2.44' should read 'Facts 2.44'.","section":"Title and heading"}],"recommendation":"major_revision","confidential_remarks":"The main obstacle is the missing closedness premise in the proof of Theorem 4.4. If the author can close that gap, the paper is likely acceptable; the rest of the architecture is coherent, and the classification would be a significant result. I would not recommend rejection, since the gap appears repairable and the surrounding lemmas provide a plausible route."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the short version: this is a major classification paper, probably correct, and worth refereeing seriously. The headline result — a complete classification of the class S, yielding the gap theorem, Thomas' conjecture, interpretability in (Q;<), and finite homogenizability — is genuinely new and appears sound. The proof strategy is sensible: take Simon's decomposition of S into a hereditarily cellular base plus finite covers of highly set-transitive groups, classify those covers in detail, then reassemble using the L(·,·,·) reconstruction. The finite cover classification in Sections 5–7 is careful, and the orbit-growth calculations give real content rather than just formal closure under constructions.\n\nThe soft spots are real but not fatal. Most importantly, Theorem 4.4's stabilization step applies Lemma 4.2 to the groups (G_{A_i,{C}})|C. Lemma 4.2 is a statement about the class F of closed finite covers, and at that point in the proof the restrictions are only known to be finite covers, not closed. The note after Lemma 4.1 explicitly says closedness does not follow from [Sim18b], and Theorem 4.4 is supposed to establish it — so the written argument appears to use closedness to prove closedness. I think the gap is repairable: for a closed oligomorphic group, the restriction to an invariant subset is closed by a standard back-and-forth argument, and your setting is oligomorphic throughout. But the paper needs to say that, or give the argument. Without that sentence, a careful reader hits a missing premise at the entrance to the classification.\n\nThe other soft spot is the reliance on Lemma 4.1 from Simon's unpublished [Sim18b]. The author says the details of F and E are not important, but this lemma is the gateway to everything; a referee will need the preprint or a self-contained proof. I also noticed Lemma 6.1 leaves condition (5) of the ω-partition definition unverified — likely routine, but it should be filled in. There are also some notation garbles around G(Y) versus G((Y)) that should be cleaned.\n\nWho is this for? Anyone working on ω-categorical structures, permutation group orbit growth, or Thomas' conjecture. It deserves a serious external referee; I would not desk-reject. The recommendation is to send it out with specific requests to verify Theorem 4.4's stabilization step and the import from [Sim18b].","headline":"A substantial and likely correct classification of the class S, with real consequences, but the written proof has a repairable gap in Theorem 4.4 and leans heavily on an unpublished source.","tokens_in":47065,"tokens_out":30039,"would_cite":true,"duration_ms":322212,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["03C45","03C35","20B27"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper classifies every countable structure whose unlabelled orbit growth stays below $2^n/p(n)$ for all polynomials $p$, showing that such groups are built from finite symmetries layered over the rational order.","keywords":["unlabelled growth","oligomorphic permutation groups","omega-categorical structures","Thomas' conjecture","finite covers of permutation groups","highly set-transitive groups","hereditarily cellular structures","interpretability in (Q;<)"],"falsifier":"Take any closed oligomorphic permutation group $G$ whose unlabelled growth satisfies $u_n(G) \\sim c^n$ for a real $c$ not lying in $\\{\\gamma_d : d \\in \\mathbb{N}\\}$, while still $u_n(G) < 2^n/p(n)$ for every polynomial $p$; if such a group exists, the gap theorem (Theorem 1.5) is false. A natural place to look is among wreath products of finite highly set-transitive groups over structures interpretable in $(\\mathbb{Q};<)$, since the classification predicts that every growth constant appearing in $\\mathscr{S}$ must be one of the $\\gamma_d$.","tokens_in":46059,"feed_emoji":"📈","tokens_out":6997,"duration_ms":71357,"temperature":0.7,"pith_summary":"Let $\\mathscr{S}$ be the class of countable structures whose number $u_n$ of orbits on $n$-element subsets never reaches $2^n/p(n)$ for any polynomial $p$; the paper gives a complete classification of the automorphism groups of these structures. The classification says that every such group is assembled, by finite direct products, wreath products with $\\mathrm{Sym}(\\omega)$, finite-index supergroups, and isomorphisms, from copies of $\\mathrm{Aut}(\\mathbb{Q};<)$ acting on finite highly set-transitive fibres — so the rational order is the only infinite primitive ingredient. From this description the paper derives that $u_n$ grows like $\\gamma_d^n$ for exactly one of the numbers $\\gamma_d$ defined as the largest real root of $x^d-x^{d-1}-\\cdots-1$, with the sequence increasing to $2$. It follows that there are only countably many such structures up to bidefinability, that each is first-order interpretable in $(\\mathbb{Q};<)$ and is interdefinable with a finitely bounded homogeneous structure, and that each has finitely many first-order reducts — Thomas' conjecture for $\\mathscr{S}$.","feed_headline":"Unlabelled growth below 2^n forms a discrete ladder","feed_subtitle":"The classification shows every such structure is interpretable in the rational order and confirms Thomas' conjecture.","key_machinery":"The central object is the cover construction $L(\\tilde{G}, G^*, D)$: starting from a linked finite cover $\\tilde{G}$ of a hereditarily cellular group $G^*$, one attaches to each orbit of $G^*$ a datum $D(a) = (F_a, B_a, \\phi_a)$ consisting of a fibre group $F_a$, a normal pointwise binding group $B_a$, and a surjection onto the fibre of $\\tilde{G}$, and the resulting group consists of all permutations whose coordinate actions lie in the prescribed fibre data and are witnessed by an element of $\\tilde{G}$. This construction captures exactly the covers with finite fibre factors. The second load-bearing mechanism is the ladder of classes $\\mathscr{S}_d$ with thresholds $\\gamma_d$, the largest real roots of $x^d - x^{d-1} - \\cdots - 1$; the recurrence $u_n = \\sum_{i=1}^{|F|} u_i(H) u_{n-i}(G_0)$ for wreath products $H \\wr \\mathrm{Aut}(\\mathbb{Q};<)$ forces the growth constant to be one of the $\\gamma_d$. Finite highly set-transitive groups $H$ are the only finite ingredients needed, and the rational order supplies the unique infinite primitive behaviour.","core_discovery":"The central claim is Theorem 7.10: for $d \\in \\mathbb{N} \\cup \\{\\infty\\}$, a permutation group $G$ lies in the class $\\mathscr{S}_d$ of groups with unlabelled growth below $(\\gamma_d+\\varepsilon)^n$ if and only if $G$ is isomorphic to a group $L(\\tilde{G}, G^*, D)$ constructed from finite covers of highly set-transitive groups over a hereditarily cellular base, and equivalently if and only if $G$ is built from $\\mathrm{id}(\\{\\emptyset\\})$ and groups $H \\wr \\mathrm{Aut}(\\mathbb{Q};<)$, where $H$ is a finite highly set-transitive group of degree at most $d$, by isomorphisms, closed supergroups, finite direct products, and wreath products with $\\mathrm{Sym}(\\omega)$. Corollary 7.11 then gives $\\mathscr{S} = \\bigcup_d \\mathscr{S}_d$, so every structure in $\\mathscr{S}$ has exponential growth with base strictly below $2$. From the classification the paper derives the gap theorem (Theorem 1.5), interpretability in $(\\mathbb{Q};<)$, finite homogenizability and finite boundedness, and Thomas' conjecture for $\\mathscr{S}$.","pith_inferences":["Editorial inference: the classification suggests that the exponential growth bases appearing among oligomorphic structures form a discrete ladder $1, \\gamma_2, \\gamma_3, \\ldots \\to 2$, so a structure whose growth constant is not one of these values cannot lie in $\\mathscr{S}$; the paper does not claim such a statement outside $\\mathscr{S}$.","Editorial inference: for constraint satisfaction, Theorem 1.6 reduces structures with polynomial unlabelled growth to combinations of $(\\mathbb{Q};<)$ and cellular structures, but the paper notes complexity questions require primitive-positive definability, so the classification alone does not settle the infinite-domain CSP dichotomy for this class.","Editorial inference: a natural testable extension is to ask whether the same $L(\\tilde{G},G^*,D)$ construction still classifies groups with unlabelled growth bounded by $c^n$ for a fixed $c<2$, rather than by $2^n/p(n)$ for every polynomial $p$."],"forward_implications":["Every structure in $\\mathscr{S}$ has unlabelled growth whose $n$-th root tends to one of the numbers $\\gamma_d$; no intermediate growth constants occur in this class.","The class $\\mathscr{S}$ contains only countably many structures up to bidefinability.","Every structure in $\\mathscr{S}$ is first-order interpretable in $(\\mathbb{Q};<)$ and is interdefinable with a finitely bounded homogeneous structure.","Thomas' conjecture holds for $\\mathscr{S}$: every structure in $\\mathscr{S}$ has finitely many first-order reducts up to interdefinability.","A structure has at most polynomial unlabelled growth exactly when an expansion by finitely many constants is bidefinable with a finite disjoint union of copies of $(\\mathbb{Q};<)$ together with a cellular structure."],"supporting_citations":[{"why":"Supplies Lemma 4.1, the decomposition of every structure in S as a cover of a hereditarily cellular base whose non-trivial fibres are finite covers of highly set-transitive but not highly transitive groups; the whole classification feeds through this lemma.","marker":"[Sim18b]"},{"why":"Provides the classification and closure properties of hereditarily cellular groups, including the rank machinery and Theorem 2.21, used for the base group G* and for the counting finiteness argument in Theorem 9.3.","marker":"[Bod24]"},{"why":"Gives the stable dichotomy and the identification of stable structures with subexponential unlabelled growth as hereditarily cellular, and formulates the conjectures on gaps and on structures with subexponential orbit growth that this paper confirms.","marker":"[Bra22]"},{"why":"Classifies groups with polynomial unlabelled growth and underlies Theorem 8.2 and Theorem 1.6 describing P-oligomorphic groups as unions of linear orders and cellular structures after adding constants.","marker":"[FT20]"},{"why":"Classifies finite covers of reducts of (Q;<), which yields Theorem 5.2 and the explicit description of the class F of closed finite covers of highly set-transitive groups.","marker":"[Iva99]"},{"why":"Cameron's classification of closed highly set-transitive groups as the five reducts of (Q;<) fixes the possible base actions and fibre actions used throughout the classification.","marker":"[Cam76]"},{"why":"States Thomas' conjecture, the finitely-many-reducts property for finite-signature homogeneous structures, which the paper verifies for the whole class S.","marker":"[Tho91]"},{"why":"Classifies finite highly set-transitive groups, giving the finite list of possible degree-d fibre groups in Theorem 5.23.","marker":"[L W65]"}],"fun_headline_variants":["Sub-2^n growth: full classification","All sub-2^n structures interpretable in (Q;<)","Thomas' conjecture confirmed for sub-2^n growth","Discrete ladder of structures below 2^n","Complete classification of sub-2^n unlabelled growth"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is the imported structural lemma saying that every structure in $\\mathscr{S}$ has a cover whose base is hereditarily cellular and whose non-trivial fibres are finite covers of highly set-transitive, but not highly transitive, groups; the paper relies on this lemma to enter the classification, and if it fails for some $G \\in \\mathscr{S}$, Theorem 7.10 and its consequences collapse.","fun_headline_variants_meta":{"raw":{"variants":["Sub-2^n growth: full classification","All sub-2^n structures interpretable in (Q;<)","Thomas' conjecture confirmed for sub-2^n growth","Discrete ladder of structures below 2^n","Complete classification of sub-2^n unlabelled growth"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000402,"raw_usage":{"total_tokens":2091,"prompt_tokens":936,"completion_tokens":1155,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":552,"completion_tokens_details":{"reasoning_tokens":1079}},"tokens_in":552,"tokens_out":1155,"duration_ms":10129,"temperature":1.0,"reasoning_tokens":1079,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T15:01:00.471630+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take any closed oligomorphic permutation group $G$ whose unlabelled growth satisfies $u_n(G) \\sim c^n$ for a real $c$ not lying in $\\{\\gamma_d : d \\in \\mathbb{N}\\}$, while still $u_n(G) < 2^n/p(n)$ for every polynomial $p$; if such a group exists, the gap theorem (Theorem 1.5) is false. A natural place to look is among wreath products of finite highly set-transitive groups over structures interpretable in $(\\mathbb{Q};<)$, since the classification predicts that every growth constant appearing in $\\mathscr{S}$ must be one of the $\\gamma_d$.","supporting_citations":[],"review_version":1}