{"id":"2b940428-c99e-4bed-8c4a-89cab408cb9f","arxiv_id":"2512.21278","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Core companions preserve stability, NIP, simplicity, and NSOP_k, but the classes of structures interpretable over (N;=) and (Q;<) are not closed under taking core companions.","lead":"This paper studies 'core companions': simplified versions of mathematical structures that keep their existential-positive content and are unique when they exist. It proves that stability, NIP, simplicity, and similar tameness properties survive taking core companions, and builds counterexamples showing the classes of structures interpretable in the pure set or in the rational order do not survive.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Non-closure over (Q;<) hinges on unproved same-author classification ([BB25, Cor. 25]); preservation theorem itself appears sound.","rationale":"I read Theorem 3.1 as the paper's main positive assertion, and its proof appears internally sound: the patterned-property transfer (Lemma 3.19, Theorem 3.22) correctly uses the equivalence between a property and its positive version in model-complete core theories, and the type-counting argument (Lemma 3.25) is valid because positive types determine full types in a model-complete core and homomorphisms preserve the relevant obstructions. The Morley-rank and trace-definability arguments also check out. The reader's weakest assumption, however, is exactly where the paper is most exposed: the non-closure over (Q;<) depends on Lemma 4.25, which in turn depends on an unproved classification from [BB25]. This is a same-author preprint and the cited Corollary 25 is not reproduced. Since the preservation theorem does not use [BB25], the concern is localized to the negative results, but those are advertised as central contributions. The abstract/full-text mismatch about trace definability in Lachlan's class is real and should be fixed, but it is a presentation issue rather than a flaw in the main proof. Thus the reader's CONDITIONAL verdict is appropriate and unchanged.","tokens_in":44802,"tokens_out":19236,"duration_ms":181455,"concrete_test":"Verify [BB25, Corollary 25] directly for small dimensions: enumerate all Aut(Q^d)-invariant equivalence relations on Q^d_↗ for d=2 and d=3 and check whether each is exactly E_S={(u,v): S⊆(u=v)} for some S⊆[d]. If a d=3 counterexample exists, Lemma 4.25 fails. If the classification holds, check that the proof in [BB25] is complete and that the preprint's hypotheses (e.g. increasing tuples, no parameters) match the use in Lemma 4.25. The same test could be run by asking the authors to include a self-contained proof of Corollary 25 in a revision.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The preservation theorem (Theorem 3.1) is well supported by the pattern-transfer and type-counting arguments, and I found no internal inconsistency there. The load-bearing concern is in the negative results. Lemma 4.25 reduces any A with a d-dimensional interpretation in (Q;<) to a first-order reduct of J_<(d̄) by invoking [BB25, Corollary 25]: every Aut(Q^d)-invariant equivalence relation on increasing tuples in Q^d is of the form {(u,v): S⊆(u=v)} for some S⊆[d]. This classification is quoted as a black box from a same-author preprint and is not proved in the present text. Lemma 4.25 is the foundation for Theorem 4.37, and hence for Corollary 4.38 (non-closure of I((Q;<)) under model companions) and for the non-interpretability of the generic permutation and S(2) via Corollaries 4.40–4.44. If Corollary 25 is false or has a hidden hypothesis, those results collapse. This is a genuine missing support rather than a disagreement with consensus. The abstract additionally overclaims that all structures in Lachlan's class are trace definable in (N;=), whereas Section 5 only states this as a consequence of Conjecture 5.4, so the written version should be corrected.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the model-complete core companion operation for first-order theories. Its main positive result, Theorem 3.1, asserts that if a complete first-order theory T has a core companion S, then S inherits a long list of model-theoretic tameness properties: stability, NIP, simplicity, NSOP, NTP2, NSOP_n, λ-stability, superstability, monadic stability/NIP, total transcendence, strong minimality, k-NIP, strong dependence, and finite U-rank. The proof machinery consists of three transfer methods: a positivization of patterned properties (Subsection 3.2), type-counting over parameter sets (Subsection 3.3), and trace definability of the core companion (Subsection 3.6). The paper then gives negative results: the class of structures interpretable over equality is not closed under taking model-complete cores (Corollary 4.21), and the class of structures interpretable over (Q;<) is not closed under taking model companions or model-complete cores (Corollary 4.38 and surrounding results); in particular (Q;<) does not interpret the generic permutation, (Q;<,S,T), or S(2). The final section introduces Lachlan's class D, conjectures that D equals the class of model-complete cores of structures interpretable over equality, and relates this conjecture to trace definability.","tokens_in":45038,"tokens_out":7985,"duration_ms":81124,"significance":"If the results are fully correct, this is a substantial contribution. The preservation theorem for patterned properties gives a uniform explanation of why so many dividing lines survive passage to the core companion, and the trace-definability transfer (Lemma 3.46) is a useful new tool. The non-closure results answer natural questions and the construction involving finite covers of the Johnson graph is elegant. The paper also makes progress on a conjecture of Walsberg, provided the conditional statement in Section 5 can be made unconditional. However, a significant part of the negative results over (Q;<) currently rests on an unproved classification result from a same-author preprint, and the abstract overstates the status of the trace-definability result for Lachlan's class. These issues need to be resolved before the paper can be accepted.","major_comments":[{"comment":"The proof of Lemma 4.25 is entirely dependent on [BB25, Corollary 25], quoted as a black box: every Aut(Q^d)-invariant equivalence relation on increasing tuples is of the form {(u,v) : S ⊆ (u=v)}. This classification is not proved in the present paper and is taken from a same-author arXiv preprint. Lemma 4.25 is then the foundation for Theorem 4.37 and hence for Corollary 4.38 (non-closure of I((Q;<)) under model companions) and for Corollaries 4.40, 4.41, and 4.44 (non-interpretability of the generic permutation, S(2), and its betweenness reduct), as well as for the inclusion MI((Q;<)) ⊆ E in Theorem 6.1. If Corollary 25 of [BB25] is false, or if it carries an additional hidden hypothesis, these results collapse. Please include a proof of this classification in an appendix or replace the reference by a published, peer-reviewed source; otherwise the non-closure claims over (Q;<) should b","section":"Abstract and §5"},{"comment":"The abstract states: 'To support our conjecture we prove that all structures in Lachlan's class are trace definable in (N;=), confirming a conjecture of Walsberg.' This is not what the full text proves. In Section 5, after Conjecture 5.4, the text only says that a positive answer to Conjecture 5.4 would imply a positive answer to Walsberg's question; no unconditional proof of trace definability for all of Lachlan's class is given. The abstract and the introduction should be corrected to say that trace definability is a consequence of Conjecture 5.4, not an independently proved theorem.","section":"Conjecture 5.4 equivalence"},{"comment":"The reduction in Lemma 5.6 and Proposition 5.7 shows the equivalence of Conjectures 5.4 and 5.5. This is fine. However, the sentence 'Note that a positive answer to Conjecture 5.4 would imply a positive answer to Walsberg's question' appears in the text only after stating Conjecture 5.4. If the authors intend trace definability of Lachlan's class as a contribution, it must be either proved or clearly marked as conditional. As written, this is a missing support passage that should be corrected in revision.","section":"§4.2, Remark 4.22"}],"minor_comments":[{"comment":"In the induction step of Lemma 3.41, the sentence 'if MR(ϕ(x,a))≥α in S, then MR(ϕ(x,f(a)))≥α in S' should read '...in T' for the second occurrence; otherwise the proof is circular.","section":"§3.5, Lemma 3.41 proof"},{"comment":"The proof refers to 'Theorem 4.35' when it should refer to 'Lemma 4.35'. Please correct the cross-reference.","section":"§4.3, Theorem 4.37 proof"},{"comment":"Remark 4.22 says 'We mention without proof that both X and Y are finitely homogenizable'. Since this fact is not used in the main argument, this is acceptable, but if it is mentioned it should be marked as 'not used in this paper' or given a reference with a proof.","section":"§4.2, Remark 4.22"},{"comment":"The letter Q is used both for the rationals and for the set of cardinality pairs in Definition 3.16. This is harmless, but the notation clash could be avoided to improve readability.","section":"§3.2, Definition 3.16"},{"comment":"The proof of Lemma 4.19 uses the classical Schreier–Ulam theorem. It would be helpful to state the precise form being used, since the statement 'every proper closed normal subgroup of Sym(Q) is trivial' is the key ingredient.","section":"§4.2.4, Lemma 4.19"},{"comment":"The implication (3)⇒(1) relies on Macpherson's theorem that finitely homogeneous relational structures cannot interpret an infinite group. This is a substantial external result and should be cited with the exact theorem number in [Mac91], so the reader can verify the translation to the present setting.","section":"§5, Theorem 5.2"}],"recommendation":"major_revision","confidential_remarks":"The preservation theorem for patterned properties appears sound and is the strongest part of the paper. The main obstacle is the reliance on [BB25, Corollary 25] for the non-interpretability results over (Q;<); since this is a same-author preprint and no proof is supplied, I would require either a proof in an appendix or a published reference before publication. I would also insist that the abstract's trace-definability claim be corrected to match Section 5's conditional statement. If the authors cannot supply the [BB25] classification, the paper should be reframed as conditional on that result."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The preservation theorem is the real contribution, and I think it is sound. The non-closure results are important, but the (Q;<) half rests on an unproved classification from the authors' own preprint, and the abstract overstates what Section 5 proves.\n\nTheorem 3.1 is genuinely general. Using Bailetti's pattern framework, the authors show that moving to a core companion preserves stability, NIP, simplicity, NSOP_n, NTP2, and more. The key trick, turning a patterned property into its positive version and proving that positive patterns transfer along homomorphic equivalence, is clean and seems right. The type-counting arguments for λ-stability and superstability are standard and correct. This unifies and extends the older ω-categorical orbit-growth results, and it will be the part people cite. I did not find a gap here.\n\nThe problems are in Section 4. The I((Q;=)) half is on firmer ground: the finite-cover argument for Y is detailed, and the involutions argument for non-interpretability is self-contained. The total categoricity example is a nice bonus.\n\nThe (Q;<) half is fragile. Lemma 4.25 reduces arbitrary d-dimensional interpretations in (Q;<) to first-order reducts of J_<(d̄). That reduction needs a classification of Aut(Q^d)-invariant equivalence relations on increasing tuples, imported as Corollary 25 of [BB25], a same-author preprint that is not proved here. Everything downstream — Theorem 4.37 and the non-interpretability of (Q;<,S,T), the generic permutation, and S(2) — depends on it. If that classification is unavailable or has a hidden hypothesis, the non-closure claim for I((Q;<)) is unsupported. This is a missing proof, not a matter of taste.\n\nTwo smaller issues. The abstract claims the paper proves trace-definability of all of Lachlan's class in (N;=); Section 5 only derives that from Conjecture 5.4. That mismatch should be fixed. Also Remark 4.22 contains an explicit 'without proof' assertion about finite homogenizability, which is not load-bearing but is a loose end.\n\nThis paper deserves a serious referee. The preservation theorem alone is publishable. The non-closure results are valuable if correct, but a referee needs to check the [BB25] dependency carefully. I would ask the authors to either prove the classification or cite a publicly available version, and to correct the abstract. It is a strong paper that needs revision, not a desk reject.","headline":"Preservation of dividing lines under core companions is the solid core; the (Q;<) non-closure chain hinges on an unproved same-author classification and the abstract overclaims the trace-definability result.","tokens_in":45589,"tokens_out":3887,"would_cite":true,"duration_ms":38711,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["03C45"],"pacs":[],"model":"deepseek-v4-flash","headline":"Taking a model-complete core preserves the model-theoretic tameness of a theory: stability, NIP, simplicity, NSOP, and strong minimality all pass from a theory to its core companion.","keywords":["model-complete cores","positive model theory","patterned properties","stability","NIP","simplicity","Lachlan's class","interpretability"],"falsifier":"Check the cited classification directly: search for an automorphism-invariant equivalence relation on increasing d-tuples in Q^d that is not determined by fixing a subset of coordinates. If such a relation exists for some d, then the reduction lemma for interpretations in (Q;<) fails, and the non-interpretability results for (Q;<,S,T), the generic permutation, and the dense local order all break. Alternatively, find an omega-categorical structure in Lachlan's class that is not homomorphically equivalent to any structure interpretable over equality; that would refute Conjecture 5.5 and hence Co","tokens_in":44624,"feed_emoji":"🧮","tokens_out":10033,"duration_ms":94036,"temperature":0.7,"pith_summary":"The paper establishes that the model-complete core companion—the positive-logic analogue of the model companion, obtained by passing to the same universal-negative consequences while making every formula existential positive—behaves tamely for classification theory. If a complete first-order theory has a core companion, then stability, NIP, simplicity, NSOP, NSOP_n, superstability, monadic stability, monadic NIP, total transcendence, and strong minimality all transfer to the core companion. This matters because cores are the standard simplification device for homomorphism and constraint-satisfaction problems: one can replace a structure by its core without losing tameness. The paper also shows the limits: total categoricity is not preserved, and the classes of structures interpretable in (Q;=) and in (Q;<) are both not closed under taking cores, with the generic permutation as a concrete witness for the ordered case. A final conjecture identifies the cores of equality-interpretable structures with Lachlan's class of omega-stable finitely homogeneous reducts.","feed_headline":"Stability, NIP, and simplicity survive core companions","feed_subtitle":"Tameness passes from a first-order theory to its model-complete core; two natural interpretability classes fail to close.","key_machinery":"The central working notion is the core companion S of a first-order theory T: the unique, when it exists, model-complete core theory with the same h-universal (universal-negative) consequences as T, meaning models of the two theories map homomorphically to each other. Because S is a model-complete core, every formula modulo S is existential positive, and existential positive formulas are preserved by homomorphisms; this is what allows formulas and their obstructions to be pushed from S into T. The proof engine is the positivisation of patterned properties: a scheme converting any property defined by omitting a combinatorial pattern in formulas into a positive-logic counterpart, so that the p","core_discovery":"In the paper's own terms, the central discovery is that core companionship preserves essentially all mainstream dividing lines of model theory. Given a complete theory T with core companion S, T and S have the same universal-negative consequences, and S is a model-complete core theory, so every S-formula is equivalent to an existential positive formula. Any model-theoretic property that can be expressed as omitting a combinatorial pattern—stability, NIP, tree properties, SOP_n—can be 'positivised', and a formula exhibits the positivised pattern in S exactly when it exhibits the original pattern in T; hence NP passes from T to S. Independently, a homomorphism from a monster model of S to a mo","pith_inferences":["If Conjecture 5.4 is true, the model-complete core operation inverts equality-interpretation on Lachlan's class: every omega-stable finitely homogeneous reduct would be homomorphically equivalent to a structure interpretable over equality, and trace minimality for that class would follow as a strengthening the paper does not fully spell out.","The trace-definability result suggests a practical strategy beyond the paper's explicit statements: any model-theoretic property preserved under trace definability can be verified on the core companion, which is often structurally simpler, even when no pattern-omission proof is available.","The group-theoretic non-interpretability method—comparing involutions and kernels of finite covering maps—could plausibly be adapted to other homogeneous graphs whose cores arise as finite covers, since the four-fold fibres in the equality example are likely not the only possible obstruction.","A direct test of Conjecture 4.45 would be to search for an omega-categorical structure not interpretable in (Q;<) with unlabelled growth slower than the betweenness reduct of the dense local order; the paper explicitly leaves that gap open."],"forward_implications":["If T is stable, NIP, simple, NSOP, NSOP_n, superstable, monadically stable, monadically NIP, or strongly minimal, then its core companion has the same property.","The core companion of T is trace definable in T, so any property preserved under trace definability—including finite U-rank, finite Morley rank, finite dp-rank, and strong dependence—transfers to the core.","For omega-categorical structures, where core companions always exist, every tame omega-categorical structure has a tame model-complete core, which is the case relevant to constraint satisfaction and orbit-finite computation.","The hierarchy I((Q;=)) is strictly contained in MI((Q;=)), which is contained in Lachlan's class D, which is strictly contained in I((Q;<)); the final inclusion into E also holds, and D is closed under taking cores.","The classes of structures interpretable over equality and over (Q;<) are not closed under taking cores; for (Q;<) this already fails for model companions, witnessed by the generic permutation."],"fun_headline_variants":["Core companions keep stability, NIP, and simplicity","Tameness survives model-complete cores","Core companions preserve all major dividing lines","Stability and NIP survive core companions","Core companions: tameness preserved, interpretability fails"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The non-interpretability proofs over (Q;<) rest on a classification, cited without proof, saying that every automorphism-invariant equivalence relation on increasing d-tuples in Q^d is of the form S ⊆ (u=v) for a fixed set S of coordinates; if that classification fails, the proofs that structures such as the generic permutation are not interpretable in (Q;<) collapse, as does the separate abstract claim about trace definability of Lachlan's class, whose proof is missing from","fun_headline_variants_meta":{"raw":{"variants":["Core companions keep stability, NIP, and simplicity","Tameness survives model-complete cores","Core companions preserve all major dividing lines","Stability and NIP survive core companions","Core companions: tameness preserved, interpretability fails"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000597,"raw_usage":{"total_tokens":2650,"prompt_tokens":785,"completion_tokens":1865,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":529,"completion_tokens_details":{"reasoning_tokens":1797}},"tokens_in":529,"tokens_out":1865,"duration_ms":14946,"temperature":1.0,"reasoning_tokens":1797,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T14:09:22.822013+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Check the cited classification directly: search for an automorphism-invariant equivalence relation on increasing d-tuples in Q^d that is not determined by fixing a subset of coordinates. If such a relation exists for some d, then the reduction lemma for interpretations in (Q;<) fails, and the non-interpretability results for (Q;<,S,T), the generic permutation, and the dense local order all break. Alternatively, find an omega-categorical structure in Lachlan's class that is not homomorphically equivalent to any structure interpretable over equality; that would refute Conjecture 5.5 and hence Co","supporting_citations":[],"review_version":1}