{"id":"0c10139d-8c65-4517-8f12-05e070735db0","arxiv_id":"2411.16706","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Every non-trivially T-convexly valued o-minimal field is definably spherically complete.","lead":"This paper proves that every non-trivially T-convexly valued o-minimal field is definably spherically complete, meaning every definable nested family of valuation balls has a common point. This answers an open question from 2023 in the negative for a broad class of valued fields, showing that a definable form of completeness can hold even when ordinary spherical completeness fails.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 1's value-group cofinality transfer is asserted rather than proved, and Fact 2 has a sign typo in the same step; Theorem 3 rests on these.","rationale":"The reader's conditional verdict is appropriate. Lemma 1 is doing all the work in the paper: without its cofinality classification, Theorem 3's final appeal to κ-spherical completeness has no ground. The two flagged steps are genuinely compressed and not merely cosmetic. Cofinality is not automatically preserved by the order-reversing valuation map, so the sentence 'cof(v(X)) ∈ {1, cof({1/x : x ∈ X})}' needs a real proof. The sign typo in Fact 2's proof sits in the same cofinality-preservation statement used by Theorem 3; while it is almost certainly a typo, the printed proof as it stands does not fully support the citation. None of this shows the theorem is false; the strategy is plausible and the cited results in [2] and [4] likely supply the missing arguments. The right disposition is to require a rewritten Lemma 1 and a corrected Fact 2 before acceptance, exactly the reader's conditional verdict. I also checked the possible objection that Theorem 3 only proves the property for E* rather than for E: this is not a separate problem, because an E-definable nest's intersection formula is first-order, so a witness in the elementary extension E* transfers downward to E by elementarity.","tokens_in":2517,"tokens_out":34015,"duration_ms":247924,"concrete_test":"Re-derive the final step of Lemma 1 from first principles in a non-trivially T-convex model, e.g. the real closed Puiseux-series field with value group Q and O its convex valuation ring, or a model with an uncountable-cofinality value group. For definable X such as (0,a), (a,∞), {x : v(x)>γ}, and finite boolean combinations, compute cof(X), cof(v(X)), and cof({1/x : x∈X}), and verify membership in {1, cof(r(E,O)), cof(v(E,O)<0), cof(E)}. In particular verify the asserted identities cof(E<b)=cof(r(E,O)) and cof(E>b)=cof(v(E,O)<0), and confirm that the corrected Fact 2 (with '<0') is what [2, Thm. A] actually gives. If any computed cofinality falls outside the listed set, Lemma 1 and Theorem 3 are false.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central theorem rests entirely on Lemma 1, the only place where the paper bounds the cofinality of every (E,O)-definable subset of v(E,O) by {1, cof(r(E,O)), cof(v(E,O)<0), cof(E)}. The proof of Lemma 1 has two compressed steps. First, after reducing to an (E,O)-definable X⊆E>0 and writing it as a finite union of traces of intervals with endpoints in E⟨b⟩, it asserts without proof that cof(E<b)=cof(r(E,O)) and cof(E>b)=cof(v(E,O)<0). These identifications are exactly what lets the cofinality of a definable set be read off from the sort of its endpoints. Second, the final sentence 'if X⊆E>0, then cof(v(X)) ∈ {1, cof({1/x : x ∈ X})}' is not justified: v is order-reversing, so cofinality of v(X) is not automatically the cofinality of X or of its reciprocal set; one needs a separate argument for the definable pieces at hand. Adding to the concern, the proof of Fact 2 prints 'v(E,O)<0 is cofinal in v(E⟨x⟩, O∗∩E⟨x⟩)>0' where the statement of Fact 2 and its use in Theorem 3 require '<0'. If any of these identifications fails in a model of Tconvex, the maximal extension E* need not have every definable nested family of valuation balls of cofinality <κ, and Theorem 3 collapses.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims that every model of the theory Tconvex (non-trivially T-convexly valued o-minimal fields) is definably spherically complete: every definable nested family of valuation balls has non-empty intersection. The proof passes to a maximal κ-bounded wim-constructible elementary extension (E*,O*), uses Fact 2 to preserve the cofinalities of the value group, residue field, and field sort, and applies Lemma 1 to conclude that no definable nested family of balls of small cofinality can exist. The author notes this gives a negative answer to Question 1.1 of [1] for expansions defining exponentiation, since by [3] such fields are not spherically complete.","tokens_in":2849,"tokens_out":1969,"duration_ms":20781,"significance":"If the proof is correct, this is a substantial result: it sharply distinguishes definable spherical completeness from ordinary spherical completeness in a broad class of valued o-minimal fields, and it resolves a question from [1] for a natural family of expansions. The paper is short and relies on the author's earlier work [2] for the construction of κ-bounded wim-constructible extensions; that reliance is not circular because [2] is about weakly immediate types and spherical completions, not about definable spherical completeness. The argument's architecture is coherent: maximal extensions with controlled cofinalities are a standard and promising route. However, the correctness of the proof hinges on a few unproved cofinality identifications in Lemma 1, and on a likely sign typo in the printed statement of Fact 2. The paper is not yet self-contained at the load-bearing step.","major_comments":[{"comment":"The proof asserts, without proof, that cof(E<b) = cof(r(E,O)) and cof(E>b) = cof(v(E,O)<0). These equalities are load-bearing: they are exactly what allows the cofinality of an arbitrary (E,O)-definable subset of E>0 to be read off from the endpoints of its o-minimal traces. Since the proof of the lemma gives no argument and cites no reference for these identifications, the reader cannot verify the central transfer from subsets of the field to subsets of the residue field and value group. Please supply a proof or a precise citation to a result that establishes these cofinality formulas in Tconvex.","section":"Lemma 1, proof"},{"comment":"The sentence \"if X ⊆ E>0, then cof(v(X)) ∈ {1, cof({1/x : x ∈ X})}\" is not justified as written. The valuation map v is order-reversing, so cofinality of v(X) is not automatically equal to the cofinality of X or of its reciprocal set; the cofinality of the image under a reversing map depends on initial segments of X, not on the cofinality of X in the usual upward sense. A separate argument is needed for the definable pieces at hand. Without it, the lemma does not establish the claimed bound on cofinalities of definable subsets of v(E,O).","section":"Lemma 1, final sentence"},{"comment":"The proof of Fact 2 states that [2, Thm. A] gives that \"v(E,O)<0 is cofinal in v(E⟨x⟩, O∗∩E⟨x⟩)>0\", but the statement of Fact 2 (and its use in Theorem 3) requires cofinality in v(E∗,O∗)<0, i.e. with the inequality '<0' on both sides. This looks like a sign typo in the printed proof. If the '<0' on the target side is indeed meant, please correct the typo; if the intended statement really is '>0', then Fact 2 as stated does not follow and Theorem 3 lacks the cofinality equality cof(v(E,O)<0)=cof(v(E∗,O∗)<0) that it invokes. This is a load-bearing point and must be clarified.","section":"Fact 2"}],"minor_comments":[{"comment":"There are minor typographical errors in the abstract ('Thi s', 'field') and in the text ('cofinality', 'cofinal') that should be fixed in a revision.","section":"Abstract and Introduction"},{"comment":"The notation E<b and E>b is used without an explicit definition; please define these as the sets of elements less than b and greater than b, respectively, where b is the element with O < b < E>O.","section":"Lemma 1, proof"},{"comment":"The term \"p.c. sequence\" is used without recalling its definition; although the reader is referred to [2, Def. 3.15] for the notion of κ-bounded wim-constructible extension, the abbreviation p.c. (probably 'pseudo-Cauchy') should be spelled out or defined.","section":"Fact 2"},{"comment":"The conclusion of Theorem 3 is stated as \"every model of Tconvex is definably spherically complete\", but the proof only treats the expansion (E,O) after passing to E∗. It would help to explicitly state that definable spherical completeness of (E∗,O∗) implies the same for (E,O), since any (E,O)-definable family is also (E∗,O∗)-definable and the intersection property is elementary.","section":"Theorem 3"}],"recommendation":"major_revision","confidential_remarks":"The paper's central theorem is plausible and would be a nice contribution, but the current manuscript does not supply enough detail at the two steps that the argument depends on: the cofinality identifications in Lemma 1 and the sign in Fact 2. These are not merely cosmetic. If the author can provide a complete proof of Lemma 1 (or replace it with a reference that covers the cofinality transfer), and can fix the Fact 2 typo, the paper would likely be acceptable. I encourage the editor to request a revision that addresses these points before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Pietro proves a clean theorem: every T-convex o-minimal field is definably spherically complete. This answers Question 1.1 of Bradley-Williams–Halupczok in the negative for expansions defining exponentiation, since those are T-convex and cannot be spherically complete by Kuhlmann–Kuhlmann–Shelah. The result is new, and the overall strategy—pass to a maximal κ-bounded wim-constructible extension, preserve cofinalities, apply Lemma 1—is sensible and, as far as I can see, works.\n\nWhere the paper is soft is in Lemma 1, which is the load-bearing step. The proof reduces a definable subset of E>0 to traces of intervals with endpoints in E⟨b⟩, and then says 'one easily reckons' two cofinality identifications: cof(E<b)=cof(r(E,O)) and cof(E>b)=cof(v(E,O)<0). Those are plausible but not trivial; for a paper that advertises a structural property, this deserves a written argument. The final transfer to the value group is even more compressed: cof(v(X)) ∈ {1, cof({1/x : x∈X})} is asserted without proof, and since v is order-reversing the cofinality of v(X) is not obviously determined by the cofinality of X. A few extra sentences would settle it. The sign typo in Fact 2's proof ('>0' where '<0' is meant) is minor but should be fixed.\n\nI don't see a circularity problem in citing the author's own [2] for Fact 2 and for the existence of maximal κ-bounded wim-constructible extensions. Those are independent technical results, not the property being proved. The citation pattern is fine.\n\nIf I were refereeing this, I'd ask for a revision that expands Lemma 1 and fixes the typo. The theorem is significant enough within model theory of valued fields to deserve that referee time.","headline":"A short, promising note: the theorem is likely correct, but Lemma 1's proof needs to be written out before this is referee-ready.","tokens_in":3372,"tokens_out":2800,"would_cite":false,"duration_ms":729670,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["03C64","12J10"],"pacs":[],"model":"deepseek-v4-flash","headline":"Every model of $T_{\\mathrm{convex}}$ is definably spherically complete: definable nested families of valuation balls always have non-empty intersection.","keywords":["T-convexity","o-minimal field","definable spherical completeness","valuation balls","value group cofinality","weakly immediate extensions","T-convex valuation rings","spherical completeness"],"falsifier":"Produce a model $(E,O)$ of $T_{\\mathrm{convex}}$ and an $(E,O)$-definable nested family of valuation balls whose intersection is empty; equivalently, find an $(E,O)$-definable subset of $v(E,O)$ whose cofinality is none of $\\mathrm{cof}(r(E,O))$, $\\mathrm{cof}(v(E,O)^{<0})$, $\\mathrm{cof}(E)$, or $1$. Either would directly contradict Theorem 3 or Lemma 1 and settle the matter.","tokens_in":2315,"feed_emoji":"","tokens_out":8391,"duration_ms":74402,"temperature":0.7,"pith_summary":"This paper proves that in every o-minimal field carrying a non-trivial $T$-convex valuation ring, every definable nested family of valuation balls has a common point. That property, definable spherical completeness, can hold even when the field is not spherically complete in the classical sense, for instance in expansions defining exponentiation. The proof obtains a maximal weakly-immediate elementary extension, then uses a cofinality classification of definable subsets of the value group to show that no definable ball nest can escape to an empty intersection. A direct corollary is that a definably spherically complete valued field need not admit a spherically complete elementary extension, answering the motivating question negatively in a natural class of examples.","feed_headline":"Definable ball nests always intersect in T-convex fields","feed_subtitle":"Every definable nest of valuation balls has a common point; classical completeness may still fail.","key_machinery":"The workhorse is the cofinality classification in Lemma 1: for $(E,O)\\models T_{\\mathrm{convex}}$, every $(E,O)$-definable subset $X\\subseteq v(E,O)$ has cofinality in $\\{\\mathrm{cof}(r(E,O)), \\mathrm{cof}(v(E,O)^{<0}), \\mathrm{cof}(E), 1\\}$. The proof of the theorem chooses a cardinal $\\kappa$ larger than all three infinite cofinalities and passes to a maximal $\\kappa$-bounded wim-constructible extension $(E^*,O^*)$, where 'wim' abbreviates weakly immediate: the extension is generated by pseudolimits of weakly immediate sequences that have no pseudolimit in the ground field. Fact 2 ensures this extension preserves the residue field, leaves $E$ cofinal in $E^*$, and leaves $v(E,O)^{<0}$ cofinal in $v(E^*,O^*)^{<0}$; with those cofinalities fixed, Lemma 1 forces every definable nested ball family to intersect, making the extension (and by elementarity the original structure) definably spherically complete.","core_discovery":"The central claim is Theorem 3: every model of $T_{\\mathrm{convex}}$ is definably spherically complete. In detail, if $(E,O)$ is an o-minimal field equipped with a non-trivial $T$-convex valuation ring, then every $(E,O)$-definable nested family of valuation balls has non-empty intersection. The proof builds a maximal $\\kappa$-bounded wim-constructible elementary extension $(E^*, O^*)$ of $(E,O)$ for a cardinal $\\kappa$ exceeding the relevant cofinalities; such an extension is $\\kappa$-spherically complete, and Fact 2 shows it preserves the residue field, the cofinality of the field, and the cofinality of the negative part of the value group. Lemma 1 then classifies the cofinality of every $(E,O)$-definable subset of the value group as one of four values, forcing every definable ball nest in $(E^*, O^*)$—and hence in $(E,O)$—to meet. Since known results show that non-trivially $T$-convexly valued o-minimal fields defining exponentiation are not spherically complete, the paper concludes that Question 1.1 of [1] has a negative answer for such expansions.","pith_inferences":["One could test whether the same conclusion holds for other classes of valued fields with restricted cofinality spectra for definable subsets of the value group, such as Hensel-minimal or dp-minimal fields.","Because the proof is non-constructive through a maximal chain, it only asserts existence of intersection points; explicit descriptions in concrete exponential-logarithmic fields remain an open direction.","The apparent sign typo in Fact 2 should be checked: if the intended statement is that $v(E,O)^{<0}$ is cofinal in $v(E^*,O^*)^{<0}$, the induction works; otherwise the proof of cofinality preservation needs repair."],"forward_implications":["Every model of $T_{\\mathrm{convex}}$ is definably spherically complete, so no definable nested family of valuation balls can have empty intersection.","In expansions defining exponentiation, definable spherical completeness coexists with the absence of spherical completeness, giving a negative answer to the motivating question about whether definably spherically complete expansions always have spherically complete elementary extensions.","The cofinality classification of Lemma 1 is a structural fact about definable subsets of the value group in all $T_{\\mathrm{convex}}$ models and may hold independently of the completion argument.","Maximal $\\kappa$-bounded wim-constructible extensions provide $\\kappa$-spherically complete elementary extensions that preserve the residue field and the cofinalities of the field and of the negative value group."],"supporting_citations":[{"why":"Poses the motivating question (Question 1.1) about whether definably spherically complete expansions must have spherically complete elementary extensions; this paper answers it negatively for a natural class.","marker":"[1]"},{"why":"Supplies the construction of $\\kappa$-bounded wim-constructible extensions and the cofinality preservation facts used in Fact 2.","marker":"[2]"},{"why":"Shows that fields defining exponentiation cannot be spherically complete, which turns the new definable spherical completeness result into a negative answer to the motivating question.","marker":"[3]"},{"why":"Provides the $T$-convexity results used in Lemma 1 to reduce definable subsets of $E^{>0}$ to traces of intervals and preimages of the valuation ring $O$.","marker":"[4]"}],"fun_headline_variants":["T-convex fields: every definable ball nest intersects","Definable ball nests always meet in T-convex fields","T-convexity ensures definable spherical completeness","Every definable nest of valuation balls has a point"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is the lemma that every definable subset of the value group has one of four possible 'sizes at infinity'; if that classification fails, the proof collapses, and the printed Fact 2 also appears to contain a sign typo ('>0' where '<0' seems intended) in the cofinality-preservation step used next.","fun_headline_variants_meta":{"raw":{"variants":["T-convex fields: every definable ball nest intersects","Definable ball nests always meet in T-convex fields","T-convexity ensures definable spherical completeness","Every definable nest of valuation balls has a point"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000232,"raw_usage":{"total_tokens":1430,"prompt_tokens":825,"completion_tokens":605,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":441,"completion_tokens_details":{"reasoning_tokens":540}},"tokens_in":441,"tokens_out":605,"duration_ms":6414,"temperature":1.0,"reasoning_tokens":540,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T15:59:56.474098+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Produce a model $(E,O)$ of $T_{\\mathrm{convex}}$ and an $(E,O)$-definable nested family of valuation balls whose intersection is empty; equivalently, find an $(E,O)$-definable subset of $v(E,O)$ whose cofinality is none of $\\mathrm{cof}(r(E,O))$, $\\mathrm{cof}(v(E,O)^{<0})$, $\\mathrm{cof}(E)$, or $1$. Either would directly contradict Theorem 3 or Lemma 1 and settle the matter.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Poses the motivating question (Question 1.1) about whether definably spherically complete expansions must have spherically complete elementary extensions; this paper answers it negatively for a natural class."}],"review_version":1}