{"id":"fb2b4448-0f18-49a8-9934-f1646dca0d52","arxiv_id":"2411.10763","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Kausz compactifications and spaces of complete collineations are unified as total space and base of a single torus-equivariant blow-up family over Grassmannians.","lead":"This paper shows that two classical compactifications in algebraic geometry, the Kausz compactification of general linear groups and the spaces of complete collineations, come from one construction: blowing up Grassmannians along a torus action. The Kausz-type space is the total family, and the spaces of complete collineations appear as its fixed-point components.","discovery_kind":"unification","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The unproved general covering statement for the Mille Crêpes charts in §3.2 is load-bearing: smoothness of T_{s,p,n}, the SNC divisor claim, and Theorem 1.2 all depend on it.","rationale":"The reader's weakest_assumption identifies exactly the point on which the proof of the main theorem rests. The paper's central construction is a torus-equivariant iterated blow-up of a Grassmannian, and its smoothness is established only through the asserted Mille Crêpes atlas. The general coverage statement in §3.2 is explicitly deferred with 'Similarly to Lemma 3.3', and Lemma 3.3 itself is proved by a schematic induction whose key step is summarized rather than demonstrated. This is a genuine load-bearing gap rather than a stylistic omission: every global conclusion about divisors, the flat retraction, and the Hilbert quotient uses the local normal forms supplied by those charts. The paper does contain real independent content: worked coordinate computations, explicit examples, a comparison with Kausz's construction, and a plausible route to the main theorems. I do not see an internal contradiction or a clear false statement; the appropriate status is conditional on filling the covering proof and the related omitted reductions. The reader's CONDITIONAL verdict therefore stands unchanged.","tokens_in":23401,"tokens_out":5705,"duration_ms":64737,"concrete_test":"Test the first general case not covered by Lemma 3.3: take (s,p,n)=(3,2,5), so r=2, and take l=1. Over ℚ (then over Spec Z if desired), form the affine open U_1⊂G(2,5) from (27), restrict the ideals S_0,S_1,S_2 from (13) to U_1, and compute the affine blowup algebra defining R^{-1}_{s,p,n}(U_1). For every τ∈J_1, compute the image A_τ of the embedding J^τ_1:A^6→PN_{2,5}×PN^0×PN^1×PN^2 and saturate the union ⋃A_τ. Check that the ideal of the complement of ⋃A_τ inside the blowup is the unit ideal; if the complement is nonempty, the §3.2 covering assertion is false and Proposition 3.6 collapses. If the check passes, repeat on G(3,6) with l=1 and l=2 to gauge whether the omitted induction is routine or hides a real obstruction.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central geometric claims pass through the assertion in §3.2 that the Mille Crêpes charts cover R^{-1}_{s,p,n}(U_l) for every 0≤l≤r. The text says only: 'Similarly to Lemma 3.3, we can further prove that ∪_{τ∈J_l} A_τ = R^{-1}_{s,p,n}(U_l).' The analogous proof of Lemma 3.3 is itself a compressed induction using auxiliary charts A^τ_k for intermediate blowups Y^p_k; the general l case has two Gaussian-elimination blocks and is not written out. This missing coverage is not cosmetic. Proposition 3.6 deduces smoothness of T_{s,p,n} over Spec Z solely from the existence of this atlas. Lemma 4.3 computes D^±_k∩A_τ locally and needs the charts to be an open cover to conclude that the 2r divisors are globally smooth with simple normal crossings. Theorem 1.2(A)–(D), the flat retraction, and the subsequent identification with the Hilbert quotient all inherit this dependence. If some point of R^{-1}_{s,p,n}(U_l) lies outside every A_τ, the atlas is not an atlas and the proof of the main theorem collapses. No independent global smoothness argument is supplied, and the unproved 'WLOG' reduction 2p≤n≤2s means missing cases cannot be dismissed by symmetry. A referee should require a complete proof or a rigorous computational verification of the covering statement before the main theorem is accepted.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper constructs schemes T_{s,p,n} and M_{s,p,n} as closures of a rational map from the Grassmannian G(p,n) to a product of projective spaces, and realizes T_{s,p,n} as an iterated blow-up of G(p,n) according to the torus action (7). The main theorems assert that T_{s,p,n} is smooth and projective over Spec Z, that its boundary consists of 2r smooth divisors with simple normal crossings, that there is a flat retraction onto M_{s,p,n}, that M_{s,p,n} is wonderful, and that M_{s,p,n} is isomorphic to the Hilbert quotient G(p,n)//G_m. The paper also identifies T_{p,p,2p} with the Kausz compactification of GL_p, derives toroidal embeddings and resolutions of the Landsberg-Manivel birational maps, and gives fibrations of these spaces over smaller Grassmannians.","tokens_in":23664,"tokens_out":6006,"duration_ms":71232,"significance":"If the missing proof of the atlas covering is supplied, the paper would be a substantial and useful unification: it would give a single blow-up picture for complete collineations and Kausz compactifications, identify the total family over the Hilbert quotient, and provide concrete coordinate charts that should be checkable in examples. The explicit coordinate constructions and worked examples (Examples 3.1 and 3.5) are valuable, and the claimed divisor equations in Lemma 4.3 are concrete and testable. The main geometric claims are not circular and do not rely on fitted parameters, but several load-bearing assertions are currently stated without proof.","major_comments":[{"comment":"The assertion after the definition of J^τ_l that ∪_{τ∈J_l} A_τ = R^{-1}_{s,p,n}(U_l) is the only proof that the Mille Crêpes charts form an atlas, but it is not proved: the text says 'Similarly to Lemma 3.3, we can further prove' and then proceeds. Lemma 3.3 itself is proved by an induction over auxiliary charts A^τ_k on intermediate blow-ups Y^p_k, and the general l-case contains two Gaussian-elimination blocks, so the reduction is not automatic. This coverage statement is used in Proposition 3.6 for smoothness of T_{s,p,n}, in Lemma 4.3 for the simple normal crossing divisor structure, and in Theorem 1.2(C) for the orbit stratification; if some point of R^{-1}_{s,p,n}(U_l) lies outside every A_τ, the main theorem is not supported. Please provide the complete proof, or a rigorous computational verification of the covering for all 0≤l≤r and all parameter ranges. The companion statement that J^τ_l extends to an embedding is also left as 'Similarly to Lemma 3.2' and should be written out.","section":"§3.2"},{"comment":"The statement 'Without loss of generality, we assume that 2p ≤ n ≤ 2s' is not justified. The quantities r, the charts U_l, and the blow-up centers depend on s,p,n, and no symmetry is stated that reduces the general case to this range. Since Theorem 1.2 is asserted for all 0<p<n, either the reduction should be proved explicitly (for instance by a duality or transpose argument) or the statement of the theorem should be restricted to the range 2p≤n≤2s.","section":"§2"},{"comment":"In Proposition 5.1, the claim that K_{n-p,p,n}∘LM∘KA extends to an embedding is central to the identification KA_{p,n} ≅ T_{n-p,p,n} and hence to Theorem 1.3 and Corollary 1.8, but the proof is only 'Similar to the proof of Lemma 3.2'. The coordinate systems used for KA_{p,n} in (35) and the Mille Crêpes charts of §3 are not identical, and the required cancellation and injectivity computations are not shown. This step should be written out in detail or replaced by a precise reference to a proved statement that applies directly.","section":"§5.1"},{"comment":"Property (D) of Theorem 1.2 asserts that D^-_1 is wonderful. The proof says 'as in the case of the spaces of complete collineations ... we can show' without supplying the orbit/stratum correspondence or verifying the defining properties of a wonderful variety on D^-_1. In particular, one needs to check that the boundary divisors ˇD_i are smooth with simple normal crossings and that the closures of G-orbits in D^-_1 match the intersections of the ˇD_i. This is a main assertion of the paper and should be proved at the same level of detail as the other parts of Theorem 1.2.","section":"§4"}],"minor_comments":[{"comment":"In the definition of J_l, 'paritial permutation' is a typo for 'partial permutation'.","section":"§3.2"},{"comment":"In the proof of Lemma 4.12, the phrase 'the restriction of ϕ to Z_q' uses an undefined morphism ϕ; it should presumably be R_{s,p,n}.","section":"§4"},{"comment":"In the proof of Theorem 1.7, 'Property (c) in Lemma 1.7' should read 'Lemma 4.12(c)'; the reference to Lemma 1.7 does not match the numbering.","section":"§4"},{"comment":"In Example 3.5, the notation C^{16} is used for the affine scheme of coordinates over Spec Z; using A^{16} would be consistent with the rest of the paper.","section":"§3.2"}],"recommendation":"major_revision","confidential_remarks":"The decisive gap is the missing proof of the Mille Crêpes covering statement in §3.2, on which smoothness, the SNC divisor structure, and Theorem 1.2 all depend. The paper is coherent and the framework is promising, and the worked examples suggest that the missing verification is likely to be routine rather than false, so I would not recommend rejection. However, the current version does not yet contain enough detail for the main theorem to be accepted as proved. The authors should also clarify the relation to the first author's earlier preprint arXiv:2007.06200 and state precisely which results are new in this expanded version."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First thing to know: this paper has a genuinely new organizing idea, but the proof of its main smoothness theorem is not complete as written. The unproved covering statement for the Mille Crêpes charts in §3.2 is load-bearing, and the paper should not be accepted until it is written out.\n\nWhat is actually new: the construction by blowing up Grassmannians according to the Gm action gives a uniform family T_{s,p,n} that contains both the Kausz compactification of GL_p (n=2s=2p) and the classical complete collineations (as the divisors D±_1). The identification of complete collineations as fixed-point components, the total-family statement over the Hilbert quotient, and the toroidality of the Kausz compactifications are not in the cited literature. The special cases p=n-s are worked out explicitly (Examples 3.1, 3.5), and the coordinate computations are convincing enough that the construction is plausible. The paper is honest about the provenance in the first author's earlier arXiv post.\n\nThe soft spots. The stress-test note is right: §3.2 asserts, without proof, that the charts A_τ cover R^{-1}_{s,p,n}(U_l) for every l. Lemma 3.3 is proven only for l=p (i.e., p=n-s), and the general case has two Gaussian-elimination blocks. The smoothness of T_{s,p,n} (Prop 3.6), the simple normal crossings claim (Lemma 4.3), and hence Theorem 1.2 all rest on this covering. This is not a cosmetic gap. The 'WLOG' reduction 2p ≤ n ≤ 2s is also asserted without proof; the symmetry is not evident from the construction, and a referee should ask for the reduction or a statement of which cases remain open. Several computations are summarized with 'Computation yields,' which is acceptable for a first version but should be expanded. The missing reference to arXiv:2007.06200 in the bibliography, despite the abstract citing it, is minor and fixable.\n\nProportion: the central argument is plausible and the gap is likely fillable. This is a good paper that needs a serious referee.\n\nRecommendation: send to peer review. The referee should require a complete proof of the covering statement and the WLOG reduction before acceptance.","headline":"Genuine new idea, but the main smoothness proof has a load-bearing gap that needs a complete covering argument before acceptance.","tokens_in":24252,"tokens_out":5469,"would_cite":false,"duration_ms":64289,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14M15","14L30","14E05","14M27"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that one torus-equivariant blow-up construction on a Grassmannian produces both the Kausz compactification and the spaces of complete collineations, with the Kausz-type compactification as the total family over the…","keywords":["Kausz compactification","complete collineations","Grassmannians","torus action","Hilbert quotient","Bialynicki-Birula decomposition","Landsberg-Manivel birational maps","simple normal crossing divisors"],"falsifier":"Carry out the omitted covering argument for a concrete small case, such as $G(3,6)$ with $s=3$ and $l=1$: write the equations of the union of the Mille Crêpes charts inside $R^{-1}_{s,p,n}(U_1)$ and check whether some closed point lies outside every chart. An uncovered point would falsify the smoothness and simple-normal-crossing conclusions; finding none would supply the missing step.","tokens_in":23144,"feed_emoji":"🧩","tokens_out":15585,"duration_ms":144508,"temperature":0.7,"pith_summary":"The paper aims to establish that two classical compactification theories—the Kausz compactification of the general linear group and the spaces of complete collineations of linear maps—are two halves of one canonical operation: blow up a Grassmannian along the stable and unstable strata of a $\\mathbb G_m$-action induced by a splitting $E=E_1\\oplus E_2$. The resulting schemes $T_{s,p,n}$ are smooth and projective, with a boundary of $2r$ simple-normal-crossing divisors, and the two end divisors are isomorphic to $M_{s,p,n}$, the generalized space of complete collineations. Over an algebraically closed field, $M_{s,p,n}$ is the Hilbert quotient $G(p,n)/\\!/\\mathbb G_m$ and $T_{s,p,n}$ is its flat total family. In the square case $n=2s=2p$, $T_{p,p,2p}$ is exactly Kausz's compactification, so the paper gives a uniform explanation of why Kausz's iterative blow-ups are the natural resolution for this torus action. The same construction resolves the Landsberg-Manivel birational maps from projective spaces to Grassmannians, turning them into morphisms.","feed_headline":"Blowing up Grassmannians yields Kausz compactifications","feed_subtitle":"One torus-equivariant blow-up of a Grassmannian yields both compactifications in a single family.","key_machinery":"The load-bearing object is the Kausz-type compactification $T_{s,p,n}$, the closure of the graph of the rational map from $G(p,n)$ to a product of Plücker projective spaces. The mechanism carrying the proofs is the Mille Crêpes coordinate atlas: coordinate charts built by iterated rank-one Gaussian elimination adapted to the Grassmannian's Plücker coordinates, on which the $\\mathbb G_m$-action has the normal form $b\\mapsto tb$, $a\\mapsto t^{-1}a$. In these coordinates the Bialynicki-Birula decomposition is explicit: fixed loci are subschemes $D_{(p-k,k)}$, the boundary divisors $D^\\pm_i$ become coordinate hyperplanes, the retraction $P_{s,p,n}$ is projection onto the invariant coordinates, and the fibers are chains of $\\mathbb G_m$-stable rational curves of Plücker degree $r$. The same charts let the paper read off the isomorphism between source and sink and identify the Kausz blow-up with the resolution of the Landsberg-Manivel maps.","core_discovery":"On its own terms, the central discovery is that the Kausz-type compactification $T_{s,p,n}$—the closure of the graph of the rational map $K_{s,p,n}:G(p,n)\\dashrightarrow \\mathbb P(\\wedge^p E)\\times\\prod_{k=0}^r \\mathbb P(\\wedge^k E_1\\otimes\\wedge^{n-k}E_2)$—is the canonical total space attached to the $\\mathbb G_m$-action (7): its source and sink divisors $D^-_1$ and $D^+_1$ are each isomorphic to $M_{s,p,n}$, and the flat projection $P_{s,p,n}$ makes $T_{s,p,n}$ the total family over the Hilbert quotient $G(p,n)/\\!/\\mathbb G_m$, which is $M_{s,p,n}$ after base change to an algebraically closed field. For $n=2s=2p$, $T_{p,p,2p}$ is isomorphic to Kausz's compactification of $\\mathrm{GL}_p$, so the classical Kausz compactification and the classical spaces of complete collineations are two faces of the same blow-up. The same comparison identifies the Kausz blow-up with the resolution of the Landsberg-Manivel birational maps, and the source/sink isomorphism yields explicit birational maps between projective bundles over Grassmannians.","pith_inferences":["Editorial inference: the paper works under the standing assumption $2p\\le n\\le 2s$, but the coordinate construction is symmetric under exchanging the two summands of $E$ and under $p\\leftrightarrow n-p$; if the missing covering proof in §3.2 is supplied, the same smoothness and Hilbert-quotient statements should extend to the full parameter range without new geometry.","Editorial inference: because the fibers of $P_{s,p,n}$ are explicit chains of rational curves in the Mille Crêpes charts, one can use $T_{s,p,n}$ as a computational bridge: intersection numbers on $M_{s,p,n}$ could be pushed forward from a smooth variety whose boundary divisors are coordinate hyperplanes.","Editorial inference: the paper establishes the construction but not a uniqueness theorem; a natural test of 'canonical' would be whether every smooth $\\mathbb G_m$-equivariant compactification of the general linear group with source and sink isomorphic and resolving the Euler foliation must coincide with $T_{s,p,n}$."],"forward_implications":["Kausz's compactification of $\\mathrm{GL}_p$ is a toroidal embedding of a general linear group, inheriting the divisor structure of Theorem 1.2.","The spaces of complete collineations of $p\\times(n-p)$ matrices appear as a divisor in $T_{p,p,n}$, and the general $T_{s,p,n}$ are obtained from the Kausz compactification by iterated locally trivial fibrations over Grassmannians.","Over an algebraically closed field, $M_{s,p,n}$ is the Hilbert quotient $G(p,n)/\\!/\\mathbb G_m$, and $P_{s,p,n}:T_{s,p,n}\\to M_{s,p,n}$ is its flat total family, with general fiber a chain of $\\mathbb G_m$-stable rational curves of Plücker degree $r$.","The blow-up $KA:T_{p,p,2p}\\to\\mathbb P^{p^2}$ resolves the Landsberg-Manivel birational map $\\mathbb P^{p^2}\\dashrightarrow G(p,2p)$, and the same comparison handles the general rational maps from projective spaces to Grassmannians.","The isomorphism between the source and sink divisors gives explicit birational isomorphisms between the projectivized normal bundles $\\mathbb P(N_1)$ and $\\mathbb P(N_2)$ of two Grassmannian subvarieties."],"supporting_citations":[{"why":"supplies the modular compactification of GL_p and its iterative determinantal blow-up, which Proposition 5.1 identifies with T_{s,p,n} in the square case.","marker":"[16]"},{"why":"establishes over C that the generalized complete collineations equal the Hilbert quotient G(p,n)//G_m, the statement Theorem 1.7 extends to algebraically closed fields.","marker":"[31]"},{"why":"provides the torus-action decomposition into stable and unstable strata used to describe the divisors D^±_i and the fixed loci of T_{s,p,n}.","marker":"[2]"},{"why":"introduces the birational maps from projective space to Grassmannians whose resolution is stated in Proposition 5.2 and Corollary 1.8.","marker":"[21]"},{"why":"describes complete collineations as closures of rational maps via Plücker coordinates, the classical model generalized by M_{s,p,n} and Theorem 1.5.","marker":"[34]"},{"why":"supplies the Chow-quotient smoothness argument used in the proof that M_{s,p,n} is the Hilbert quotient.","marker":"[24]"},{"why":"defines the toroidal-embedding framework used in Corollary 1.6 for the Kausz compactification.","marker":"[5]"}],"fun_headline_variants":["Blow-up unifies Kausz compactifications and collineations","One torus blow-up yields both compactifications","Kausz compactifications as canonical blow-ups of Grassmannians","A single blow-up captures both Kausz and complete collineations","Torus blow-up of Grassmannian yields Kausz family"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is the paper's claim in §3.2 that the Mille Crêpes coordinate charts cover every affine piece $R^{-1}_{s,p,n}(U_l)$, asserted by 'similarly to Lemma 3.3' without a proof; on that coverage rest the smoothness of $T_{s,p,n}$, the simple-normal-crossing divisor structure, and Theorem 1.2.","fun_headline_variants_meta":{"raw":{"variants":["Blow-up unifies Kausz compactifications and collineations","One torus blow-up yields both compactifications","Kausz compactifications as canonical blow-ups of Grassmannians","A single blow-up captures both Kausz and complete collineations","Torus blow-up of Grassmannian yields Kausz family"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000621,"raw_usage":{"total_tokens":2947,"prompt_tokens":1081,"completion_tokens":1866,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":697,"completion_tokens_details":{"reasoning_tokens":1777}},"tokens_in":697,"tokens_out":1866,"duration_ms":11317,"temperature":1.0,"reasoning_tokens":1777,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T19:20:18.778648+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Carry out the omitted covering argument for a concrete small case, such as $G(3,6)$ with $s=3$ and $l=1$: write the equations of the union of the Mille Crêpes charts inside $R^{-1}_{s,p,n}(U_1)$ and check whether some closed point lies outside every chart. An uncovered point would falsify the smoothness and simple-normal-crossing conclusions; finding none would supply the missing step.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"supplies the modular compactification of GL_p and its iterative determinantal blow-up, which Proposition 5.1 identifies with T_{s,p,n} in the square case."},{"cited_title":"Thaddeus","cited_arxiv_id":null,"evidence_quote":"establishes over C that the generalized complete collineations equal the Hilbert quotient G(p,n)//G_m, the statement Theorem 1.7 extends to algebraically closed fields."},{"cited_title":"Bia lynicki-Birula","cited_arxiv_id":null,"evidence_quote":"provides the torus-action decomposition into stable and unstable strata used to describe the divisors D^±_i and the fixed loci of T_{s,p,n}."},{"cited_title":"Landsberg and L","cited_arxiv_id":null,"evidence_quote":"introduces the birational maps from projective space to Grassmannians whose resolution is stated in Proposition 5.2 and Corollary 1.8."},{"cited_title":"Vainsencher","cited_arxiv_id":null,"evidence_quote":"describes complete collineations as closures of rational maps via Plücker coordinates, the classical model generalized by M_{s,p,n} and Theorem 1.5."},{"cited_title":"Brion and S","cited_arxiv_id":null,"evidence_quote":"defines the toroidal-embedding framework used in Corollary 1.6 for the Kausz compactification."}],"review_version":1}