{"id":"08c94a67-20ba-4372-9ad2-93630df919c2","arxiv_id":"2510.26648","paper_version":2,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper defines μ_bds, a level-based invariant of the derived category that measures the failure of a scheme to be a birational derived splinter.","lead":"Birational derived splinters are a characteristic-free version of rational singularities, and this paper introduces a number, μ_bds, that measures how far a scheme is from being one. The invariant is computed from the derived category and is shown to behave predictably under field extensions, products, and projective bundles.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 3.4 and the Section 4 inequalities hinge on two lemmas from unpublished preprints; a hidden hypothesis in either would break the central claims.","rationale":"The reader's weakest assumption identifies the same load-bearing point: Theorem 3.4's finiteness and resolution-independence depend on unpublished lemmas. My review of the proof confirms that the internal steps are plausible if those lemmas are true: the diagram from Lütkebohmert is standard, the level multiplication inequality is correct, and the application of [LV25, Lemma 3.16] to a regular target is reasonable. No direct mathematical error appears in the paper itself. The concern is therefore one of verification and generality, not of known falsehood. Since the reader already flagged this and accepted with moderate confidence, my stress-test does not change the verdict; it only underscores that the result should be regarded as conditional on the cited preprints. A concrete check of those preprints' hypotheses is the most efficient way to settle whether the concern lands.","tokens_in":13313,"tokens_out":32853,"duration_ms":300475,"concrete_test":"Download the latest arXiv versions of [DL24] (arXiv:2401.13661) and [DLMR25] (arXiv:2502.08629). Verify the exact hypotheses of Lemma 3.9 in [DL24] and Lemma 5.5 in [DLMR25]. Then re-run the proof of Theorem 3.4 with only those stated hypotheses: choose a Noetherian X that is not quasi-excellent or not finite-dimensional and a resolution f that is not projective, and check whether Lemma 3.9 still gives finiteness; if it does not, compute level_{Rf_*D^b_coh(X̃)}(O_X) for a concrete singular X (e.g., a non-projective proper surface) and see whether the value is infinite. Separately, test Lemma 5.5 by taking S = Rf_*D^b_coh(X̃) and asking whether the level of O_X is attained by a single object of S when S is not closed under coproducts; if not, the equality in Theorem 3.4 fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is Theorem 3.4: for any Noetherian X admitting a resolution f: X̃ → X, μ_bds(X) = level_{Rf_*D^b_coh(X̃)}(O_X) < ∞, independent of the resolution. The finiteness is imported from [DL24, Lemma 3.9], and the existence of an object E attaining the level is imported from [DLMR25, Lemma 5.5]. Both are same-author preprints which are not yet peer-reviewed. If [DL24, Lemma 3.9] actually requires, say, X to be quasi-excellent or finite-dimensional, or f to be projective, then the asserted finiteness may fail for the stated generality of 'Noetherian scheme.' Likewise, if [DLMR25, Lemma 5.5] needs the subcategory to be closed under arbitrary direct sums or the ambient category to be compactly generated, the step in Theorem 3.4 where a single E′∈D^b_coh(Z) is chosen with E∈<Rh′_*E′>_1 ceases to be justified, and the chain of inequalities giving μ_bds(X) = N collapses. Because Theorems 4.4 and 4.6 (field extension and product inequalities) invoke Theorem 3.4 directly, a correction in either preprint would invalidate the paper's headline results. This is not an internal inconsistency, but an unresolved external dependency that the paper itself flags by citing these preprints.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a numerical invariant μ_bds(X) for Noetherian schemes X, defined as the supremum of level_{R f_* D^b_co h(Y)}(O_X) over all proper birational morphisms f:Y→X. It observes that X is a birational derived splinter precisely when μ_bds(X)=1 (Lemma 1.2 and Definition 1.3). The central result, Theorem 3.4, states that if X admits a resolution of singularities f:X~→X, then μ_bds(X) equals level_{R f_* D^b_co h(X~)}(O_X) and is finite, independent of the chosen resolution. The authors then study behavior under completion (Corollary 3.5), smooth pullbacks and projective bundles (Proposition 3.6, Corollary 3.7), localization to stalks (Proposition 3.11), field extensions (Propositions 4.1 and 4.3, Theorem 4.4), and products (Propositions 4.5 and Theorem 4.6). The main new results are in positive characteristic, where derived splinter techniques are less developed.","tokens_in":13673,"tokens_out":20639,"duration_ms":191274,"significance":"If correct, this paper provides a finite, computable invariant measuring the failure of birational derived splinters, and gives strong new evidence that categorical generation can encode singularities in positive and mixed characteristic. The reduction of μ_bds to a single resolution in Theorem 3.4 is conceptually clean, and the product/base-change inequalities are new. The paper is well written and the proofs are detailed; definitions are motivated by the relation between splitting of the unit and generation (Lemma 1.2). The main results are conditional on the existence of resolutions of singularities, a standard hypothesis in positive-characteristic questions but not always known. The paper's utility would be enhanced if the cited unpublished lemmas were made available or proved in the text.","major_comments":[{"comment":"The proof of Theorem 3.4 depends on two load-bearing external results: [DL24, Lemma 3.9] for finiteness of level_{R f_* D^b_co h(X~)}(O_X) and [DLMR25, Lemma 5.5] for the existence of a single object E attaining that level. Both are unpublished preprints, the latter by three of the authors. The manuscript does not state the hypotheses of these lemmas. If [DL24, Lemma 3.9] requires, for example, quasi-excellence or finite dimensionality, or if [DLMR25, Lemma 5.5] requires closure under arbitrary direct sums or compact generation, then the stated generality of Theorem 3.4 ('Noetherian scheme admitting a resolution') may be too broad. The equality μ_bds(X)=level_{R f_* D^b_co h(X~)}(O_X) is the cornerstone of the paper, and Corollary 3.5, Corollary 3.7, Proposition 3.11, Proposition 4.3, Theorem 4.4, and Theorem 4.6 all invoke it. The authors should either state the needed lemmas in the man","section":"Theorem 3.4 (Section 3.1)"}],"minor_comments":[{"comment":"The proof contains a typo: 'Choose n≤μ_bds(X×_kY) such that O_{X'×Y}∈⟨Rf'_*D^b_coh(X×Y)⟩_n' should read 'O_{X×Y}∈⟨Rf'_*D^b_coh(X'×Y)⟩_n'. As written, the object and subcategory are on the wrong sides of the morphism f'.","section":"Proposition 4.5"},{"comment":"In the proof, the identity ℓ'_*O_XL ≅ ⊕_{n∈Z} O_X^{⊕r_n}[n] is not accurate: for a field extension ℓ':X_L→X, the pushforward ℓ'_*O_XL is a quasi-coherent sheaf concentrated in degree 0 and is isomorphic to a direct sum of copies of O_X, with no shifts. The splitting argument that follows is unaffected.","section":"Proposition 4.1"},{"comment":"The statement uses 'an Noetherian scheme' (typo for 'a Noetherian'). More substantially, the proof reduces to blowups via [Lüt93, Lemma 2.2], which may require integrality or other hypotheses. Since the paper allows non-integral schemes in the definition of birationality, please clarify that the cited lemma applies in this generality or state the proposition under the necessary hypotheses.","section":"Proposition 3.3"},{"comment":"The first inequality in the displayed chain, level_{Rg_*D^b_co h(Y)}(O_X) ≤ level_{Rg_*D^b_co h(Y)}(R(g∘h)_*E') · level_{R(g∘h)_*E'}(O_X), uses the submultiplicativity of level in triangulated categories. This property is standard but not stated in Section 2; adding a reference or a short justification would improve readability.","section":"Theorem 3.4 proof"},{"comment":"The inequality direction μ_bds(R)≥μ_bds(R^) is correct, but it may look counterintuitive at first. A sentence explaining that the level on the completed side is bounded above by the level on the original ring would help.","section":"Corollary 3.5"}],"recommendation":"major_revision","confidential_remarks":"The paper is well structured and the results are likely correct, but the central theorem leans on two lemmas from unpublished preprints by the same authors. I do not suspect an error; however, the referee cannot verify the load-bearing inputs. The authors should be encouraged to include the statements of the needed lemmas or provide self-contained proofs, especially because Theorem 3.4 is used throughout the paper. The typo in Proposition 4.5 is local and easily fixed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my take on De Deyn et al. The paper does something real: it defines μ_bds(X), the birational derived splinter measurement, and proves it behaves like a proper invariant. Theorem 3.4 is the load-bearing result — if X admits a resolution, μ_bds(X) equals the level of O_X with respect to the resolution's pushforward, and that value doesn't depend on the resolution. The proof is coherent, and the supporting results — projective bundle invariance, stalk-local formulas, product and field-extension inequalities — are genuinely new. The paper is also clear about where it needs resolutions, which matters in positive characteristic where they aren't known to exist.\n\nThe main soft spot is exactly what the stress-test note flags: the proof of Theorem 3.4 imports finiteness from [DL24, Lemma 3.9] and the existence of an object attaining the level from [DLMR25, Lemma 5.5]. Both are by the same group. [DL24] is accepted at PAMS, so that's probably fine, but [DLMR25] is still a preprint. If either lemma has a hidden hypothesis — say the subcategory needs to be closed under certain sums, or the scheme needs to be quasi-excellent — the equality in Theorem 3.4 could unravel. That's not an internal error; it's an external dependency that the paper itself doesn't hide. A referee should verify those importations carefully.\n\nThe resolution assumption is a second, more familiar caveat. The paper is honest that the main inequalities in Section 4 are conditional on resolutions existing. That's a real limitation but not a flaw, since it's stated plainly.\n\nI didn't find a mathematical error in the arguments I checked. The definitions are sound, the proofs are detailed, and the writing is accessible. The reader's verdict — accept, moderate confidence — strikes me as right. I'd send it to peer review. A good referee can check the preprint dependencies; if those hold up, this is a solid contribution to singularity theory via derived categories.","headline":"A genuinely useful invariant with clean proofs, but the main theorem leans on two lemmas from the same group's preprints; worth sending to a careful referee.","tokens_in":14133,"tokens_out":3022,"would_cite":true,"duration_ms":29906,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14A30","14F08","14B05","18G80"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper defines µ_bds, a numerical invariant measuring failure to be a birational derived splinter, and proves it is finite and resolution-independent when a resolution of singularities exists.","keywords":["birational derived splinters","derived categories","level","generation in triangulated categories","measurements","singularities","resolution of singularities","positive characteristic"],"falsifier":"Compute the level of the structure sheaf with respect to the derived pushforward of a resolution for a known singular scheme, such as a cone over a smooth projective curve of genus g≥1 over a perfect field, using two different resolutions (e.g., different blowups). If the two levels differ, or if the level is infinite, Theorem 3.4 fails.","tokens_in":13251,"feed_emoji":"📏","tokens_out":11036,"duration_ms":92884,"temperature":0.7,"pith_summary":"This paper introduces a numerical invariant, µ_bds(X), that quantifies how far a Noetherian scheme X is from being a birational derived splinter — a scheme for which the natural morphism O_X → Rf_*O_Y splits for every proper birational morphism Y→X. The central theorem shows that when X admits a resolution of singularities, µ_bds(X) is finite and equals the level of O_X with respect to the derived pushforward of the bounded derived category of coherent sheaves on the resolution, independent of the chosen resolution. Consequently, X is a birational derived splinter if and only if µ_bds(X)=1, so the invariant gives a quantitative handle on a condition that is otherwise defined by a supremum over all proper birational maps. The paper then proves robustness properties: µ_bds behaves predictably under smooth pullbacks, projective bundles, localizations, completion, field extensions, and fiber products, the last of these yielding new positive-characteristic results about when the splinter property is preserved or reflected.","feed_headline":"A single number measures birational derived splinter failure","feed_subtitle":"It equals 1 exactly when the scheme is a splinter, and it is always finite for resolved schemes.","key_machinery":"The central mechanism is the level of an object in a triangulated category: the minimum number of cones, shifts, and direct summands needed to build the object from a given subcategory. The paper's invariant µ_bds(X) is the supremum of level_{Rf_*D^b_coh(Y)}(O_X) over all proper birational morphisms f:Y→X. Theorem 3.4 computes this supremum as a single level when a resolution f:X~→X exists, using a dévissage result that reduces any proper birational morphism to a blowup, the splitting of the natural morphism O_{X~}→Rh_*O_Z over a regular scheme (which lets Lemma 3.2 promote the structure sheaf to all perfect complexes), and companion lemmas ensuring finiteness and the existence of an object","core_discovery":"The paper's central discovery is that for a Noetherian scheme X admitting a resolution of singularities, the invariant µ_bds(X) — defined as the supremum over all proper birational morphisms Y→X of the level of O_X in Rf_*D^b_coh(Y) — coincides with the single level of O_X with respect to Rf_*D^b_coh(X~) for any resolution f, and is finite. This makes X a birational derived splinter precisely when µ_bds(X)=1. The proof reduces an a priori global condition to a categorical computation, using a dévissage result that reduces proper birational maps to blowups and the splitting of the natural morphism on a regular resolution. The paper also establishes that µ_bds is characterized by blowups alone","pith_inferences":["The resolution-independence of µ_bds suggests the invariant could be defined purely categorically, without fixing a resolution, potentially extending to schemes where resolutions are not known; a natural test is whether the level remains finite for quasi-excellent schemes in positive characteristic that are known to have resolutions only in low dimension.","Because µ_bds measures failure via cones in the derived category, it may connect to other generation-theoretic invariants such as Rouquier dimension or strong generation time of the singular locus; one could hypothesize that µ_bds is bounded by the Rouquier dimension of the bounded derived category, with equality for certain minimal singularities.","The product inequality max{µ_bds(X),µ_bds(Y)} ≤ µ_bds(X×_k Y) ≤ µ_bds(X)µ_bds(Y) leaves open whether the upper bound is ever strict; finding examples where the product invariant is strictly less than the product of the factors would show subtler interaction of singularities under products.","The paper's reliance on unpublished companion lemmas for finiteness and attainment suggests that a standalone proof of those facts might unlock the same results under weaker hypotheses, possibly allowing a purely categorical statement that the level of O_X with respect to Rf_*D^b_coh(X~) is independent of f without assuming X has a resolution."],"forward_implications":["µ_bds(X) is finite and computable from any resolution of singularities, so the birational derived splinter property can be checked by a single level computation.","X is a birational derived splinter if and only if µ_bds(X)=1; when µ_bds(X)>1, the value quantifies how far X is from being one.","µ_bds is invariant under projective bundles: for a vector bundle E of rank r+1 on X, µ_bds(X)=µ_bds(P_X(E)).","For an affine normal X admitting a resolution, µ_bds(X) equals the supremum of µ_bds(O_{X,p}) over all points p, so the invariant is stalk-local in this setting.","Over a perfect field, for proper schemes X,Y with resolutions, µ_bds(X) ≤ µ_bds(X×_k L) for every field extension L (with equality for finite L), and max{µ_bds(X),µ_bds(Y)} ≤ µ_bds(X×_k Y) ≤ µ_bds(X)µ_bds(Y). In particular, the birational derived splinter property descends from base change and is closed under products."],"fun_headline_variants":["One number measures birational derived splinter failure","Invariant equals 1 exactly for splinters, finite otherwise","Categorical invariant quantifies singularities beyond char zero","Blowups alone determine birational splinter invariant","Finite invariant measures splinter failure"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The main formula equating µ_bds(X) with a single derived-category level assumes that X admits a resolution of singularities; without that, finiteness and resolution-independence are not established, and in positive characteristic such resolutions are not known to exist for arbitrary finite-type schemes.","fun_headline_variants_meta":{"raw":{"variants":["One number measures birational derived splinter failure","Invariant equals 1 exactly for splinters, finite otherwise","Categorical invariant quantifies singularities beyond char zero","Blowups alone determine birational splinter invariant","Finite invariant measures splinter failure"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000281,"raw_usage":{"total_tokens":1421,"prompt_tokens":583,"completion_tokens":838,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":327,"completion_tokens_details":{"reasoning_tokens":774}},"tokens_in":327,"tokens_out":838,"duration_ms":7529,"temperature":1.0,"reasoning_tokens":774,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T07:07:38.424097+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the level of the structure sheaf with respect to the derived pushforward of a resolution for a known singular scheme, such as a cone over a smooth projective curve of genus g≥1 over a perfect field, using two different resolutions (e.g., different blowups). If the two levels differ, or if the level is infinite, Theorem 3.4 fails.","supporting_citations":[],"review_version":1}