{"id":"b5283400-ac1a-4e1d-a4cc-00b5904a9e22","arxiv_id":"1908.02975","paper_version":7,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper constructs a new category of Nisnevich sheaves with transfers on non-proper modulus pairs, proves it is a Grothendieck abelian category, and computes its Ext groups as filtered colimits of Nisnevich cohomology.","lead":"The paper develops a theory of sheaves with transfers for 'modulus pairs', which are schemes together with a distinguished divisor, generalizing Voevodsky's foundational constructions for motives. It proves an explicit sheafification theorem and an Ext formula, giving the categorical foundation for 'motives with modulus' that aim to capture cohomology theories such as additive Chow groups.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified; the Raynaud-Gruson dependence is external but standard and internally consistent.","rationale":"The reader's weakest assumption already names Lemma 1.6.1 and the Raynaud–Gruson citation, and my independent pass agrees that this is the most load-bearing input. I did not find an internal inconsistency: every application of Theorem 1.6.2 has the required normal open U=M^o and proper closures; the localization squares are completed without leaving Sigma_fin; and exactness of a_Nis follows from exactness of b_Nis, a^fin_Nis, and b^*. The central claim would fail only if the cited platification corollary is misquoted or inapplicable, and that can be checked directly against [RG71] and by working through a blow-up example. Absent such evidence, the reader's ACCEPT verdict stands unchanged.","tokens_in":46821,"tokens_out":23700,"duration_ms":269710,"concrete_test":"Obtain the original statement of [RG71, Cor. 5.7.10] and check the translation used in Lemma 1.6.1: for Noetherian X, normal dense open U, and proper integral Z->X with Z_U->U finite, the U-admissible blow-up X'->X must make the strict transform Z' subset Z x_X X' finite over X'. Then trace this through Theorem 1.6.2 and Proposition 1.9.2(a)(2) on the test case X=A^2, U=A^2\\{p}, Z=Bl_p(A^2), X'=Z; if finiteness is not recovered, the localization (Sigma_fin)^{-1}MCor_fin ~ MCor—and hence a_Nis and the Ext formula—would need repair.","verdict_should_be":"UNCHANGED","load_bearing_attack":"No significant objection identified. The central construction (Theorem 2 via Theorems 4.5.5 and 4.6.3) rests on one external, standard result—Lemma 1.6.1 (Raynaud–Gruson platification [RG71, Cor. 5.7.10])—feeding Theorem 1.6.2 and Proposition 1.9.2, which give the calculus of right fractions for Sigma_fin. I traced the use: left properness of alpha survives pullback along a proper f in Sigma_fin, minimality of the blow-up f' preserves admissibility, and the finite-closure hypothesis supplies the finiteness required by MCor_fin. The proof is internally consistent and the cited theorem is the standard statement needed; no hidden assumption or circularity surfaced. The erroneous [KSY15] claim is explicitly corrected, and Question 2 is openly left open, which is consistent with the paper's claims. Therefore the reader's identified risk is a genuine external dependency but not a demonstrated gap, and it does not move the verdict.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper develops a theory of modulus sheaves with transfers for non-proper modulus pairs, generalizing Voevodsky's sheaves with transfers. The authors introduce the categories MCor and MSm along with their finite variants, define admissible correspondences and the class Sigma_fin, and prove a calculus of right fractions for this class (Proposition 1.9.2). On this foundation they define the category MNST of Nisnevich sheaves with transfers on MCor and establish the main theorem (Theorem 2, detailed in Theorems 4.5.5 and 4.6.3): the inclusion MNST into MPST has an exact left adjoint aNis given by an explicit filtered colimit formula, and Ext groups in MNST are computed as filtered colimits of Nisnevich cohomology groups. The paper also corrects a false statement from the withdrawn preprint [KSY15] and explicitly leaves open two questions concerning possible simplifications of the Ext formula.","tokens_in":46958,"tokens_out":5174,"duration_ms":54392,"significance":"If the main theorem is correct, this is a foundational contribution: it gives a genuine Grothendieck abelian category of modulus sheaves with transfers, with a computable Ext formula, which is essential for the sequel [KMSY20] on motives with modulus. The paper is careful and detailed: it provides long proofs, collects the required categorical machinery in an appendix, and is transparent about external dependencies, about the correction of [KSY15], and about the open questions. The central construction is purely deductive, with no fitted parameters or empirical inputs. The main risk is the reliance on the Raynaud–Gruson platification theorem (Lemma 1.6.1) for the calculus of right fractions, but this is a standard external result and I did not find any gap in its application.","major_comments":[],"minor_comments":[{"comment":"The decomposition of a modulus pair into the sum of its irreducible components when M^o is disconnected is stated with the proof left to the reader. Since this remark is used in later reductions (for example, to reduce to irreducible interiors), a brief proof or a precise reference would make the paper more self-contained.","section":"1.3, Remark 1.3.8"},{"comment":"The proof of Lemma 1.6.1 is a very short reduction to [RG71, Corollary 5.7.10], and this lemma is load-bearing for Proposition 1.9.2 and hence for the main theorems. A few more sentences explaining how the cited result applies—in particular why the admissible blow-up can be chosen to be a scheme rather than an algebraic space—would improve verifiability, even though the cited theorem is standard.","section":"1.6, Lemma 1.6.1"},{"comment":"In the definition of the blow-up center, the text writes q_1^*(U_1^∞) ×_{W_1} q_2^*(U_1^∞); this appears to be a typo, as the second factor should presumably be q_2^*(U_2^∞). As written, the center is the self-product of the first divisor, which would not yield the intended exceptional divisor.","section":"1.10, Proposition 1.10.4(3)"},{"comment":"The equivalence (i)⇔(iii) relies on [Voe10a, Corollary 2.17] for the cd-structure PMV. The paper has already shown PMV is strongly complete and regular, so the citation is appropriate, but a short reminder of how the cited corollary applies to the exact-sequence formulation would help the reader.","section":"4.2, Lemma 4.2.3"},{"comment":"The formula in Theorem 2(1) for aNis uses the notation (F_N)_Nis(N), where F_N is not explicitly defined until later in the paper. A forward reference to Definition 4.5.2 and Notation 4.6.2 would improve readability.","section":"Introduction"}],"recommendation":"accept","confidential_remarks":"The paper is built on a standard external theorem (Raynaud–Gruson platification) and the authors have been transparent about the flaws in the earlier preprint [KSY15]. I see no need for further changes beyond the minor presentational points listed. The paper fits the scope of the journal and the main theorem is a solid foundation for the sequel."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is the real thing — a working sheaf theory with transfers for non-proper modulus pairs, with a computable sheafification and an Ext formula, and it honestly fixes a false proposition from the authors' earlier preprint. Send it to a good referee; cite it if you work anywhere near motives with modulus.\n\nWhat is new: Voevodsky's NST/PST story is rebuilt on MCor, the category of admissible correspondences for non-proper modulus pairs. The paper's central pieces are the MV_fin cd-structure on MSm_fin (Definition 3.2.1) and the exact left adjoint a_Nis of Theorem 2 (detailed in 4.5.5) with the explicit formula (a_Nis F)(M) = colim over N in Sigma_fin down M of (F_N)_Nis(N). That formula is genuinely useful, and Theorem 4.6.3 — Ext^i_{MNST}(Z_tr(M),F) isomorphic to the filtered colimit of H^i_Nis(N,F_N) — reduces extension groups to ordinary Nisnevich cohomology. The paper also states precisely what was wrong in [KSY15, Prop. 3.5.3]: b_Nis is left exact, not exact. Questions 1 and 2 are openly left open. That is the right way to run a foundation paper.\n\nThe load-bearing external input is Raynaud–Gruson platification (Lemma 1.6.1, citing [RG71, Cor. 5.7.10]), which feeds Theorem 1.6.2 and hence the calculus of right fractions for Sigma_fin in Proposition 1.9.2. It is a heavy theorem, and it is genuinely load-bearing: without it the sheafification formula collapses. But it is standard, correctly cited, and the authors thank Gabber and Raynaud for help with it. I traced the use and found it consistent rather than a hidden gap. Minor nits: Remark 1.3.8 leaves an 'easy' proof to the reader, and several steps lean on standard cd-structure formalism from Voevodsky. Both are minor.\n\nThe citation pattern is healthy. The self-references are programmatic, and one of them is a paper the present text corrects rather than leans on.\n\nWho this is for: anyone doing motives with modulus, reciprocity sheaves, or additive Chow groups. It is foundational, not flashy, and the payoff lives in the sequels. Read Theorem 2 and the introduction for the shape of the results; the rest is machinery written carefully. I would bring it to a motivic reading group, and if I were an editor I would send it to referees without hesitation.","headline":"A solid, honest foundation paper: a genuine sheaf theory with transfers for non-proper modulus pairs, with a computable Ext formula, resting on one standard but heavy external input.","tokens_in":47570,"tokens_out":4244,"would_cite":true,"duration_ms":37050,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["19E15","14F42","19D45","19F15"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that modulus sheaves with transfers admit an exact sheafification functor, with Ext groups expressed as filtered colimits of Nisnevich cohomology.","keywords":["modulus pair","presheaf with transfers","Nisnevich sheaves","cd-structure","motives with modulus","sheafification","extension groups","non-proper modulus pairs"],"falsifier":"Take $M=(\\mathbb{P}^1,\\infty)$ and a non-$\\square$-invariant $F\\in\\mathbf{MNST}$, and compute both sides of the formula $\\mathrm{Ext}^i_{\\mathbf{MNST}}(Z_{\\mathrm{tr}}(M),F)\\simeq\\varinjlim_{N\\in\\Sigma_{\\mathrm{fin}}\\downarrow M}H^i_{\\mathrm{Nis}}(N,F_N)$; any degree in which the two sides differ would falsify Theorem 2.","tokens_in":46586,"feed_emoji":"📐","tokens_out":7942,"duration_ms":77200,"temperature":0.7,"pith_summary":"This paper develops a sheaf theory for modulus pairs—schemes equipped with an effective Cartier divisor as a modulus and a smooth open interior—by replacing the classical category of finite correspondences with the larger category of admissible correspondences. The main theorem constructs an exact left adjoint $a_{\\mathrm{Nis}}$ from presheaves with transfers to Nisnevich sheaves with transfers, so the sheaf category is a Grothendieck abelian category. It also identifies extension groups from representable sheaves as filtered colimits of ordinary Nisnevich cohomology, which makes the theory computable. If the construction works, it gives a foundation for invariants that are not $\\mathbb{A}^1$-invariant, such as additive Chow groups and higher Chow groups with modulus.","feed_headline":"Exact sheafification exists for modulus sheaves with transfers","feed_subtitle":"Ext groups become filtered colimits of Nisnevich cohomology, making the sheaf theory computable.","key_machinery":"The engine is the class $\\Sigma_{\\mathrm{fin}}$ of minimal morphisms in $\\mathbf{MSm}_{\\mathrm{fin}}$ and $\\mathbf{MCor}_{\\mathrm{fin}}$: proper morphisms that extend an isomorphism between interiors and pull back the target divisor exactly to the source divisor. Proposition 1.9.2 gives $\\Sigma_{\\mathrm{fin}}$ a calculus of right fractions, so localization at $\\Sigma_{\\mathrm{fin}}$ identifies $\\mathbf{MCor}_{\\mathrm{fin}}$ with $\\mathbf{MCor}$, and all filtered colimits in the sheafification and Ext formulas are indexed by the comma categories $\\Sigma_{\\mathrm{fin}}\\downarrow M$. The calculus is proved via a platification-type lemma that produces a proper birational modification making a given finite correspondence finite over its source; this is the step that makes the whole machinery work.","core_discovery":"Theorem 2 states that the inclusion $\\mathbf{MNST}\\to\\mathbf{MPST}$ has an exact left adjoint $a_{\\mathrm{Nis}}$ given by $(a_{\\mathrm{Nis}}F)(M)=\\varinjlim_{N\\in\\Sigma_{\\mathrm{fin}}\\downarrow M}(F_N)_{\\mathrm{Nis}}(N)$, making $\\mathbf{MNST}$ a Grothendieck abelian category. For every modulus pair $M$, the representable presheaf $Z_{\\mathrm{tr}}(M)$ is a sheaf, and $\\mathrm{Ext}^i_{\\mathbf{MNST}}(Z_{\\mathrm{tr}}(M),F)\\,\\simeq\\,\\varinjlim_{N\\in\\Sigma_{\\mathrm{fin}}\\downarrow M} H^i_{\\mathrm{Nis}}(N,F_N)$. The central point is that the sheaf condition is governed by a genuine Grothendieck topology, not by an artificial construction: it arises from a cd-structure, and the earlier mistake in the preprint is corrected by weakening an exactness statement to left exactness for one auxiliary functor. This yields a computable description of extension groups and a workable foundation for motives with modulus.","pith_inferences":["Inference: If Theorem 2 is correct, the same localization formula should give a working definition of motivic cohomology with modulus as Ext groups in $\\mathbf{MNST}$, with the filtered colimit replacing the classical Nisnevich cohomology of smooth schemes.","Inference: The paper's Question 1 suggests a concrete testable strengthening: under $\\square$-invariance and the proper-image condition, the filtered colimit should collapse to $H^q(M_{\\mathrm{Nis}},F_M)$; the blow-up case in Question 2 is a natural place to test this.","Inference: Because the cd-structure plays an essential role, extending the theory to the étale topology would require a different completeness argument; the paper's methods therefore leave open whether a similar colimit formula holds étale-locally.","Inference: The relationship between $\\mathbf{MCor}_{\\mathrm{fin}}$ and $\\mathbf{MCor}$ via right fractions suggests that many computations in the non-proper setting can be reduced to proper models, which may simplify future calculations of additive Chow groups with modulus."],"forward_implications":["The category $\\mathbf{MNST}$ is a Grothendieck abelian category, so it has enough injectives and all small colimits, making homological algebra available in the modulus setting.","Extension groups from representable modulus sheaves are filtered colimits of ordinary Nisnevich cohomology, giving an explicit way to compute them.","The  Cech complexes attached to strict Nisnevich covers are exact in $\\mathbf{MNST}$, so covers behave as they do in the classical theory.","The theory provides the sheaf-theoretic foundation on which the sequel can build categories of motives with modulus.","The corrected left exactness of the auxiliary functor $b_{\\mathrm{Nis}}$ explains the original preprint's error and still supports the main Ext formula."],"supporting_citations":[{"why":"Supplies the classical sheafification-with-transfers theorem being generalized, including the Ext formula for representable presheaves.","marker":"[Voe00]"},{"why":"Provides the platification corollary used to construct proper birational modifications that make correspondences finite, the key input for the calculus of right fractions.","marker":"[RG71]"},{"why":"Defines admissible correspondences and the containment lemma used to compose them, a prerequisite for the category MCor.","marker":"[KSY16]"},{"why":"Supplies the divisor-comparison lemma for surjective morphisms of normal integral schemes, used in admissibility proofs.","marker":"[KP12]"},{"why":"Gives the exactness of Cech complexes for finite correspondences, which the paper adapts to the modulus setting in Theorem 3.4.1.","marker":"[MVW06]"},{"why":"Supplies the cd-structure completeness and regularity criteria used to identify the Nisnevich topology on MSm_fin.","marker":"[Voe10a]"},{"why":"Provides the cd-structure characterization of Nisnevich sheaves used in the comparison with the small Nisnevich site.","marker":"[Voe10b]"},{"why":"Supplies the blow-up argument used in Lemma 1.6.3 to factor étale morphisms through proper birational modifications, needed for refining covers.","marker":"[SV00]"}],"fun_headline_variants":["Exact sheafification makes modulus sheaf Ext computable","Ext for modulus sheaves: filtered colimits of Nisnevich cohomology","Exact sheafification for non-proper modulus pairs","Sheafification exact: preprint corrected, Ext now computable"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the platification statement used in the proof of the calculus of right fractions holds for separated finite-type schemes over the base field: every finite correspondence on a normal open dense subscheme, whose closure is proper over the source, can be made finite over a proper birational modification; if this fails in the asserted generality, the localization equivalence collapses and with it the exact sheafification and the Ext formula.","fun_headline_variants_meta":{"raw":{"variants":["Exact sheafification makes modulus sheaf Ext computable","Ext for modulus sheaves: filtered colimits of Nisnevich cohomology","Exact sheafification for non-proper modulus pairs","Sheafification exact: preprint corrected, Ext now computable"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00117,"raw_usage":{"total_tokens":4772,"prompt_tokens":814,"completion_tokens":3958,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":430,"completion_tokens_details":{"reasoning_tokens":3882}},"tokens_in":430,"tokens_out":3958,"duration_ms":28954,"temperature":1.0,"reasoning_tokens":3882,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:28:35.841896+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take $M=(\\mathbb{P}^1,\\infty)$ and a non-$\\square$-invariant $F\\in\\mathbf{MNST}$, and compute both sides of the formula $\\mathrm{Ext}^i_{\\mathbf{MNST}}(Z_{\\mathrm{tr}}(M),F)\\simeq\\varinjlim_{N\\in\\Sigma_{\\mathrm{fin}}\\downarrow M}H^i_{\\mathrm{Nis}}(N,F_N)$; any degree in which the two sides differ would falsify Theorem 2.","supporting_citations":[],"review_version":1}