{"id":"344cb9ac-275a-4921-a94c-61b0694c2222","arxiv_id":"2411.10689","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Merging two smooth structure classes can preserve the original generic limits, and with a parallel strongness condition every infinite existentially definable subset of the merge's first reduct is again the first generic; 1-local classes yield new EPPA and Ramsey examples.","lead":"This paper studies 'merges' of two classes of finite structures, each with a generic limit, and asks when the merged class still has a generic limit and what properties it inherits. It shows that under extra regularity, every infinite definable subset of the merged limit's first side is again a copy of the first generic, and it gives new EPPA and Ramsey results for a family of smooth classes.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 2.12's proof needs B1 strong in the dPS amalgam, but Definition 2.3(3) only yields Wn≤D1 and B'⊆D1; the line 'Notice that B∗≤1 α(B′)' is therefore unjustified.","rationale":"The reader's weakest assumption targets the same-universe construction in the proof of Theorem 2.12. I do not think that is the principal problem: because L1 and L2 are disjoint by Definition 2.6, one may freely put an L1-structure and an L2-structure on the same carrier provided both restrict to Wn; the maps p and h do exactly this, so no compatibility constraint exists. The genuinely load-bearing step is later: “Notice that B∗≤1 α(B′)” in the proof of Theorem 2.12. The structure D1 is obtained by dPS (Def. 2.3(3)) from A=Wn∩C*, C=Wn, B=B′. The definition yields Wn≤1D1 and B′⊆D1, but not B′≤1D1, since it explicitly says g(B)⊆D. From B1≤1B′ one only gets truth of universal formulas for B1 in B′; these formulas can fail in the larger structure D1. Thus B1≤1D1 does not follow. The rest of the proof—α(B*)⊆C* and D*∩C*=α(B′)—does not repair this: to embed B1 into C* as a strong substructure one needs B1≤1D before applying the generic embedding α. Hence condition (2) of Definition 2.4 for C*|L1 is not established. This is a proof gap rather than a disproof; it might be closed by adding a two-sided strongness variant of dPS or by proving a missing lemma. It is load-bearing because all of Corollaries 2.14–2.18 rely on Theorem 2.12. The secondary issues noted by the reader (missing smooth intersections in Theorem 2.26, unsupported Ramsey transfer in Corollary 3.22) are real but do not affect the main theorem if the central gap is fixed. I therefore keep the CONDITIONAL verdict: the paper should not be relied on until Theorem 2.12's strongness step is justified or the dPS hypothesis is strengthened.","tokens_in":22718,"tokens_out":20545,"duration_ms":200019,"concrete_test":"Re-derive the disputed step from Definition 2.3(3): instantiate dPS with A=Wn∩C*, C=Wn, B=B′, and check whether the conclusion B1≤1D1 follows for the substructure B1⊆B′. Concretely, exhibit a smooth class satisfying the hypotheses of Theorem 2.12 (closed under substructure, dAP, dPS, smooth intersections) and a finite configuration Wn, B′, B1 for which some dPS amalgam D1 over S has B1 not strong in D1; if such a configuration exists, the proof fails as written. If none exists, identify the missing lemma that lets the step go through, or strengthen dPS to require both embeddings strong and verify that all cited examples still satisfy the stronger property.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central proof of Theorem 2.12 has a gap at the line “Notice that B∗≤1 α(B′)” (proof of Thm 2.12, after the definition of B*). Let S = Wn∩C*. The preceding construction of D1 used dPS (Def. 2.3(3)) with A = S, C = Wn, B = B′. That definition guarantees only Wn≤1D1 and B′⊆D1; it does not guarantee B′≤1D1, nor that the substructure B1 of B′ is strong in D1. From B1≤1B′ and B′⊆D1 one cannot infer B1≤1D1: the universal formulas that hold in B′ need not hold in the larger structure D1. Without B1≤1D1, the image α(B*) need not be ≤1C*, so condition (2) of Definition 2.4 is not established for C*|L1. This is not the same-universe compatibility issue: since L1 and L2 are disjoint (Def. 2.6), defining D by independently pulling back D1 and D2 onto one carrier is legitimate. The unsupported step is the transfer of strongness from B′ to B1 in D1. The proof would go through if dPS were strengthened to require both embeddings to be strong, but that is not the definition used.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies merges of smooth classes (K1,≤1) and (K2,≤2) into a new smooth class K1⊛K2 and the model-theoretic properties of the resulting generic structure M*. The central result, Theorem 2.12, claims that if K1 has smooth intersections and dPS, then for any infinite set C* definable in M* by an existential L2-formula, the L1-reduct C*|L1 is isomorphic to the generic of K1. The paper derives several self-similarity corollaries from this theorem, proves transfer and non-transfer results for atomicity and saturation of generics, and in Section 3 proves EPPA results for a class of 1-local smooth classes and their merges, including a generalization of Ivanov's equivalence-relation construction. The paper is written in a clear expository style and builds on standard external results (EHN19, HKN22, Iva15, KL92, Bod12).","tokens_in":22875,"tokens_out":26835,"duration_ms":246219,"significance":"If Theorem 2.12 is correct, it is a substantial strengthening of the known fact that the generic of a merge of substructure-closed smooth classes expands the generic of each component: it shows that certain definable subsets of the merged generic already encode the original generic, yielding strong self-similarity phenomena for Shelah-Spencer classes and other smooth classes. The EPPA results for 1-local classes and their merges are also a useful contribution, as smooth classes with EPPA are rare and the paper gives a flexible construction via extended structures. The paper ships no code or machine-checked proofs, but the arguments are mostly self-contained and contain no fitted parameters or circular steps. The main theorem's proof, however, requires a careful repair before the stated results can be regarded as established.","major_comments":[{"comment":"The proof of Theorem 2.12 is not rigorous as written because it conflates D1, D, and D*=α(D). Specifically, ρ is defined by ρ(b_i)=d_i, but B* is α({d_i})∪A, so the image of ρ is not contained in B*; the intended map is ρ(b_i)=α(d_i). The sentence 'α(ρ(B1))≤1α(B′)' then cannot be parsed, and the equality 'D*∩C*=α(W_n∪B′)∩C*' is written as though α were defined on W_n∪B′, which is a subset of D1, not of D. These are notation errors that can be repaired by working with the preimage p^{-1}(B′)⊆D and proving {d_i}∪A≤1p^{-1}(B′) from B1≤1B′, which does hold by the dAP construction; the dPS step only supplies W_n≤1D1 and B′⊆D1, and does not need to supply B′≤1D1. The proof also omits the needed condition that E be disjoint from W_n: if e_i∈W_n for some i, the map f:D2→E∪W_n is not injective. Because C* is infinite, one can choose E outside W_n (or replace W_n by an isomorphic copy avoiding E), but this must be stated explicitly.","section":"§2.3.1, Theorem 2.12 proof"},{"comment":"Theorem 2.26 invokes Theorem 2.23, but Theorem 2.23 requires (K,≤) to have smooth intersections, and Theorem 2.26 neither assumes smooth intersections for K1 and K2 nor proves that the merge (K*,≤*) has smooth intersections. The proof of Theorem 2.23 uses smooth intersections in the step where B≤M and A_i≤M imply B∩A_i≤A_i. Consequently the proof of Theorem 2.26 is incomplete; either add smooth intersections to the hypotheses or prove that they hold for the merge.","section":"§2.3.3, Theorem 2.26"},{"comment":"Definition 2.22 is internally inconsistent with its use: it defines (A,B) to be a ≤-minimal pair by 'for every A⊆B′⊊B, A≤B′ but A̸≤B', yet Theorem 2.23 and Theorem 2.26 construct chains of minimal pairs in which each consecutive pair satisfies A_i≤A_{i+1}. As printed, the final clause A̸≤B makes it impossible for consecutive members of such a chain to be strong extensions. The definition should presumably read 'A≤B and for every proper intermediate B′, A̸≤B′' (or the intended dual version); please correct it and re-check the argument in Theorem 2.23.","section":"§2.3.3, Definition 2.22"},{"comment":"In the proof of Proposition 2.21, the sentence 'By each of Th(M1) and Th(M2) being ω-categorical' assumes ω-categoricity of M1, but the proposition only assumes that K1 is a Fraïssé class. A Fraïssé class in a countable relational language need not have an ω-categorical limit (e.g., a language with countably many binary relations can produce a limit with infinitely many 2-types). The proposition should either add the hypothesis that M1 is ω-categorical, restrict to finite relational languages, or justify why the Fraïssé limit is ω-categorical in the present setting.","section":"§2.3.2, Proposition 2.21"}],"minor_comments":[{"comment":"After defining D2, '∪{wi,...,wℓ}' should read '∪{w_1,...,w_ℓ}'.","section":"§2.3.1, Theorem 2.12 proof"},{"comment":"The statement '{P_i}_{i≤n} a set of n equivalence classes' should say n+1 classes if the index set is {0,...,n}.","section":"§2.3.1, Corollary 2.16"},{"comment":"The equality 'D*∩C*=α(Wn∪B′)∩C*' should be rewritten using the preimage of B′ in D, since α is not defined on Wn∪B′.","section":"§2.3.1, Theorem 2.12 proof"},{"comment":"The cross-reference 'By Theorem 2.5' should be 'By Proposition 2.5'.","section":"§2.2.1, proof of Theorem 2.9"},{"comment":"The introduction advertises structural Ramsey theory, but Section 3 is mostly about EPPA; only Corollary 3.22 directly concerns the Ramsey property. The scope statement could be adjusted for accuracy.","section":"§1, Introduction"}],"recommendation":"major_revision","confidential_remarks":"The central theorem is plausible and the paper is in scope, but the proof of Theorem 2.12 needs a careful rewrite, and the hypotheses of Theorem 2.26 and Proposition 2.21 need adjustment. I would be willing to review a revised version. No concerns about novelty or unattributed prior work."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read Bryant's merges paper. The short version: the main theorem (2.12) is very likely correct, and the flagged gap in the stress test is not actually a gap. The line B*≤1α(B') follows from B1≤1B', which the dAP step explicitly gives; α is an L1-isomorphism onto its image, so strongness transfers. The same-universe issue is a non-issue because L1 and L2 are disjoint. The proof does have a typo: ρ maps B1 to {d_i}∪A, not to B*, and the notation around ρ/α needs cleaning. But the argument holds together.\n\nWhat's new: Theorem 2.12 strengthens the EHN19 strong-extension result to get self-similarity of existentially L2-definable infinite subsets of the merge's generic; the corollaries (point deletion, automorphism group intersection, equivalence relation expansions, Shelah-Spencer applications) are real. The 1-local EPPA results and the KEω extension of Iva15/HKN22 to proper smooth classes are a genuine advance for the EPPA/Ramsey program.\n\nSoft spots: Theorem 2.26 invokes Theorem 2.23, which needs smooth intersections, but neither class is assumed to have them. Easy fix: add the assumption, or prove the chain suffices without it. Corollary 3.22 asserts the Ramsey property for K⊛KLO without proving that the extended-structure Ramsey witness descends to the original class; that transfer is not automatic and needs an argument or the corollary should be softened. Proposition 2.11's construction of B2 assumes there is an L2-structure on the same universe as B1 extending A; that is not automatic for arbitrary smooth classes, though dPS may rescue it with a bit of work. These are fixable, not fatal.\n\nThe citation pattern is clean, no self-citation, and the paper relies on standard external tools (EHN19, HKN22, Iva15, KL92). If the author patches the flagged assumptions and supplies the missing Ramsey transfer, this becomes a solid paper for the smooth-class community.\n\nRecommendation: send it to a serious referee. The main theorem is worth refereeing now; the EPPA section is worth refereeing. I would not desk-reject.","headline":"Main theorem survives the stress test; the real problems are a missing assumption in Theorem 2.26 and an unproved Ramsey transfer in Corollary 3.22.","tokens_in":23545,"tokens_out":17785,"would_cite":true,"duration_ms":160678,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["03C50","03C45","05D10"],"pacs":[],"model":"deepseek-v4-flash","headline":"The generic of a merge of smooth classes contains, on every infinite set definable by an existential formula in the second class's language, a reduct isomorphic to the first class's generic.","keywords":["smooth classes","generic limits","amalgamation property","parallel strongness","structural Ramsey theory","Hrushovski property (EPPA)","Shelah-Spencer graphs","1-local classes"],"falsifier":"Take the merge of a Shelah-Spencer class $K_\\alpha$ with the Fraïssé class of finite equivalence relations, let $M^*$ be its generic, fix $a\\in M^*$, and let $C^*=\\{b\\in M^*:E(a,b)\\}$; check whether the $L_\\alpha$-reduct of $C^*$ satisfies the extension property of Definition 2.4 for every finite $A\\le_\\alpha C^*$ and every finite $B$ with $A\\le_\\alpha B$—if some such $B$ cannot be embedded into $C^*|L_\\alpha$ fixing $A$, then Theorem 2.12 is false.","tokens_in":22356,"feed_emoji":"🔁","tokens_out":10144,"duration_ms":90216,"temperature":0.7,"pith_summary":"Two smooth classes are families of finite structures with a relation saying when one structure is a strong substructure of another, and their generics are the countable limits built by amalgamation. The paper asks what survives when two such families are merged into one class of structures carrying both languages. It proves that if the first class is closed under substructure, has disjoint amalgamation, smooth intersections, and disjoint parallel strongness, then inside the merged generic every infinite set definable by an existential formula in the second class's language is an isomorphic copy of the first class's generic. This yields self-similarity corollaries: deleting a point, taking finite unions of equivalence classes, or intersecting a family of definable sets all give back the original generic. The paper also shows atomicity transfers to merged generics, gives a counterexample showing saturation does not transfer, and proves that certain '1-local' smooth classes and their merges have the Hrushovski EPPA property, a stepping stone toward Ramsey properties.","feed_headline":"Definable subsets of merged generics copy the first generic","feed_subtitle":"Those subsets are isomorphic to the first generic, yielding self-similarity consequences.","key_machinery":"The central object is the merge $K_1 \\circledast K_2$: the class of finite $L_1\\cup L_2$-structures whose $L_1$-reducts lie in $K_1$ and whose $L_2$-reducts lie in $K_2$, ordered by requiring both strong relations to hold. The argument in Theorem 2.12 runs on disjoint parallel strongness (dPS) of $K_1$: it lets the proof amalgamate in $L_1$ on a universe that simultaneously carries an $L_2$-structure copied from the finite witness set $E\\cup W_n$, so that a single $L^*$-structure $D$ exists with $W_n\\le^* D$. Because the defining formula $\\varphi$ is existential, the copied $L_2$-relations force the embedded $L_1$-amalgam to land inside $C^*$. For the EPPA half, the machinery is the 1-local property—a smooth class whose closure operator is computed pointwise—expanded into an extended structure with unary functions that encode closures, which converts smooth amalgamation into free amalgamation of extended structures.","core_discovery":"The paper's central claim is Theorem 2.12: if $(K_1,\\le_1)$ and $(K_2,\\le_2)$ are smooth classes closed under substructure with disjoint amalgamation, and $K_1$ additionally has smooth intersections and disjoint parallel strongness, then for the generic $M^*$ of the merge, every infinite set $C^*=\\varphi(M^*,m)$ definable by an existential $L_2$-formula has $C^*|L_1 \\cong M_1$, the generic of $K_1$. The proof verifies the two defining properties of a generic for $C^*|L_1$. It builds, for any finite $A\\le_1 C^*$ and any finite $B_1$ with $A\\le_1 B_1$, an $L_1$-amalgam over a finite approximation $W_n\\cap C^*$, then uses disjoint parallel strongness to place the needed $L_1$-structure on the same universe as the $L_2$-witnesses of the existential formula; genericity of $M^*$ embeds this combined structure back into $M^*$, and the existential formula guarantees the embedded copy of $B_1$ lies inside $C^*$. The theorem is presented as the main result of the merging study, and the paper derives its automorphism-group and self-similarity corollaries from it.","pith_inferences":["An implicit consequence is that any smooth class satisfying the hypotheses of Theorem 2.12 has a generic that is self-similar in a strong sense: the corollaries list point removal and equivalence-class unions, but the same mechanism should produce a copy of $M_1$ inside any infinite set definable in the second language, regardless of which Fraïssé class is merged in.","A testable extension is to isolate the same-universe compatibility condition that the proof of Theorem 2.12 invokes; if formalized, it would likely extend the theorem to merges where $K_1$ lacks disjoint parallel strongness, since that property is used only to place the two amalgams on one carrier.","The EPPA results suggest a route toward Ramsey properties for smooth classes: because merges of 1-local classes with the tuple-equivalence-relation class have $\\le^*$-EPPA, adding linear orders to those merges and checking the Ramsey property would test whether the EPPA-to-Ramsey connection extends beyond Fraïssé classes."],"forward_implications":["The automorphism group of the merged generic satisfies $\\operatorname{Aut}(M^*)\\cong \\operatorname{Aut}(M_1)\\cap \\operatorname{Aut}(M_2)$ as subgroups of the symmetric group on $M^*$.","Point removal is trivial for such generics: for any $a\\in M_1$, the structure $M_1$ is isomorphic to $M_1-\\{a\\}$.","There is an expansion of $M_1$ by a binary equivalence relation $E$ such that every finite union of $E$-classes in $M_1$ is isomorphic to $M_1$.","For Shelah-Spencer generics, merging with the class of finite linear orders yields a family of subsets whose finite intersections are all isomorphic to the original generic $M_\\alpha$.","If the two original generics are atomic, then the merged generic is atomic; in contrast, if both original classes have many minimal pairs, the merged generic is not saturated."],"supporting_citations":[{"why":"Supplies the definition of smooth classes and the theorem that amalgamation plus countably many isomorphism types gives a unique generic, which the paper's constructions build on.","marker":"[KL92]"},{"why":"Establishes the strong-extension result that a merge of classes closed under substructure with dAP has a generic whose reducts are the original generics, the baseline that Theorem 2.12 strengthens.","marker":"[EHN19]"},{"why":"Introduces disjoint parallel strongness and smooth intersections and shows Shelah-Spencer classes satisfy them, the exact hypotheses needed for Theorem 2.12.","marker":"[BS96]"},{"why":"Provides the $\\Gamma_L$-EPPA theorem for free amalgamation classes of extended structures, which the paper uses to prove $\\le$-EPPA for 1-local classes.","marker":"[HKN22]"},{"why":"Proves that the tuple-equivalence-relation class $K_{E_\\omega}$ has EPPA, an argument the paper generalizes to merges with 1-local classes.","marker":"[Iva15]"}],"fun_headline_variants":["Merged generics: definable subsets mirror the first generic","In merged generics, definable sets are copies of the first generic","Merging smooth classes yields definable subsets isomorphic to first generic","Definable sets in merged generics replicate the first generic"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof of Theorem 2.12 assumes that the $L_1$-amalgam and the copied $L_2$-witness structure can be placed on the same finite carrier set to form a single $L^*$-structure with $W_n\\le^* D$, but no argument in the paper shows that the two structures are compatible on that carrier.","fun_headline_variants_meta":{"raw":{"variants":["Merged generics: definable subsets mirror the first generic","In merged generics, definable sets are copies of the first generic","Merging smooth classes yields definable subsets isomorphic to first generic","Definable sets in merged generics replicate the first generic"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000561,"raw_usage":{"total_tokens":2622,"prompt_tokens":862,"completion_tokens":1760,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":478,"completion_tokens_details":{"reasoning_tokens":1685}},"tokens_in":478,"tokens_out":1760,"duration_ms":10491,"temperature":1.0,"reasoning_tokens":1685,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T19:27:58.690689+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the merge of a Shelah-Spencer class $K_\\alpha$ with the Fraïssé class of finite equivalence relations, let $M^*$ be its generic, fix $a\\in M^*$, and let $C^*=\\{b\\in M^*:E(a,b)\\}$; check whether the $L_\\alpha$-reduct of $C^*$ satisfies the extension property of Definition 2.4 for every finite $A\\le_\\alpha C^*$ and every finite $B$ with $A\\le_\\alpha B$—if some such $B$ cannot be embedded into $C^*|L_\\alpha$ fixing $A$, then Theorem 2.12 is false.","supporting_citations":[],"review_version":1}