{"id":"6a10b204-0738-46ff-b68b-9fab5eb13f85","arxiv_id":"2512.06755","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A combinatorial version of the intersection Hodge conjecture for projective toric varieties is verified, conditionally on an assumed geometric-combinatorial compatibility, in dimensions at most 3 and for simplicial fans.","lead":"This paper restates the intersection Hodge conjecture for projective toric varieties in the combinatorial language of fans, defining 'combinatorial cycle classes' inside BBFK intersection cohomology. It reports that these classes span in dimensions up to three and for simplicial fans, but only under a compatibility assumption that remains unproven.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Combinatorial cycle classes are not formally defined: the Gysin map (i_tau)_* in Def. 3.2 is never constructed, and the BBFK–BL compatibility is assumed but unproven; §5's spanning proofs inherit this gap.","rationale":"The paper's central claim reduces the toric intersection Hodge conjecture to a linear-algebra statement about combinatorial classes. The single most load-bearing assumption is that [V(tau)]_comb is a well-defined class that corresponds, under the canonical isomorphism, to the geometric cycle class of V(tau). The reader correctly identifies this as the weakest assumption. I sharpen it: the paper does not even define the combinatorial Gysin map (i_tau)_*; Remark 3.2 is a dangling reference to functoriality, not a construction. This makes the central object of the conjecture formally undefined. The abstract's caveat 'Assuming the stated BBFK--BL compatibility' is not carried into the theorem statements or proofs, so the verification in §5 is conditional in a way the paper's narrative does not fully disclose. The smooth-maximal-cone issue in Theorem 5.1(3) is a separate, more local gap that would affect only that step, not the whole framework; the Gysin-map issue is more fundamental. I do not see a reason to move the verdict from CONDITIONAL: the framework is coherent, the conjecture is honestly labeled, and the gap could be repaired by providing the missing construction and compatibility proof. The reader's verdict of CONDITIONAL is appropriate; my concern does not change it, so I recommend UNCHANGED.","tokens_in":5887,"tokens_out":3008,"duration_ms":31690,"concrete_test":"Take a small non-simplicial projective 3-fan, such as the one used in the paper's promised example (or the cone over a square with a diagonal added). For each cone tau, construct (i_tau)_* explicitly using the BBFK minimal extension sheaf and the open embedding Star(tau) -> Sigma: determine the stalks, the shift by 2 dim(tau), and the induced map on hypercohomology. Then compare phi([V(tau)]_comb) with the cycle class of the invariant subvariety V(tau) in IH^*(X_Sigma) via the Goresky–MacPherson cycle class map. If the comparison fails, or if the map cannot be defined for singular maximal cones, then the degree-4 step of Theorem 5.2 and the equivalence in Conjecture 4.1 are unsupported. A positive result would validate the BBFK–BL compatibility for this example.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central conjecture (Conjecture 4.1) equates combinatorial Hodge classes with geometric Hodge classes, but the definition of the combinatorial cycle class [V(tau)]_comb in Definition 3.2 is not rigorous: the 'fundamental class in the local intersection cohomology of Star(tau)' is not defined, and the map (i_tau)_* : IH^m_comb(Star(tau)) -> IH^{m+2k}_comb(Sigma) is asserted in Remark 3.2 to arise from functoriality, with no construction, degree shift justification, or verification of the required properties. Without this map, the subspace Hdg^k_comb(Sigma) is not well-defined, and the statements 'spanned by combinatorial cycle classes' have no precise content. Moreover, even if such a map could be defined, nothing in the paper proves the 'BBFK–BL compatibility' under the canonical isomorphism phi: the abstract explicitly conditions the n<=3 verification on this compatibility, but Theorem 5.2's degree-4 paragraph silently replaces the needed combinatorial statement with the geometric assertion that intersecting divisors with an ample divisor yields invariant curves. The proof of Theorem 5.1(3) also assumes a smooth maximal cone, which can fail (e.g., the 2D complete fan with rays (±1,±1)); this is secondary but further shows the cycle-class map is not robustly constructed. These gaps are load-bearing because the equivalence in Conjecture 4.1 and the spanning claims in §5 all depend on a well-defined, geometrically compatible Gysin map.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a combinatorial analogue of the intersection Hodge conjecture for projective toric varieties. It defines, within the BBFK combinatorial intersection cohomology IH^*_comb(Σ), a combinatorial cycle class [V(τ)]_comb for each cone τ (Definition 3.2) and lets Hdg^k_comb(Σ) be their span (Definition 3.3). Conjecture 4.1 asserts that the canonical isomorphism φ: IH^*_comb(Σ,ℚ) ≅ IH^*(X_Σ,ℚ) carries Hdg^k_comb onto the geometric Hodge classes, which for projective toric varieties amounts to asking that the classes [V(τ)]_comb span IH^{2k}_comb. The paper claims to verify the conjecture for n≤2 (Theorem 5.1), n=3 (Theorem 5.2), and for simplicial fans (Theorem 5.3), and it outlines an algorithm for checking the spanning property. The abstract conditions the n≤3 verification on a 'BBFK–BL compatibility' statement, and the proofs are labeled sketches.","tokens_in":6272,"tokens_out":8777,"duration_ms":81077,"significance":"If the combinatorial Gysin maps and the BBFK–BL compatibility were actually established, the paper would provide a clean reduction of the toric intersection Hodge conjecture to a concrete linear-algebraic rank computation, and the low-dimensional checks would be valuable. The author correctly identifies Karu's Hard Lefschetz as the key input and separates the simplicial case. However, the manuscript does not deliver these prerequisites: the central objects are not well-defined, and the main theorem statements overclaim relative to the abstract. The strengths are the formulation of the question and the honest disclosure of the compatibility assumption in the abstract.","major_comments":[{"comment":"The combinatorial Gysin map (i_τ)_* is never constructed. The stated origin (functoriality of minimal extension sheaves for open embeddings of fans) does not yield a degree-shifting pushforward on intersection cohomology; open embeddings induce restriction maps in the opposite direction, and no properness or duality argument is supplied. Likewise 'the fundamental class of Star(τ)' is not defined. Consequently [V(τ)]_comb and Hdg^k_comb are not well-defined, and Conjecture 4.1 and Theorems 5.1–5.3 have no precise content as stated. This is load-bearing because every 'spanned by combinatorial cycle classes' statement depends on this definition.","section":"Def. 3.2, Rem. 3.2"},{"comment":"Theorem 5.2 asserts Conjecture 4.1 holds unqualifiedly for n=3, but the proof silently uses the BBFK–BL compatibility that the abstract discloses as an assumption. The sentence 'Since the intersection of algebraic cycles (divisors) with an ample divisor yields algebraic cycles (invariant curves)' is a geometric assertion about ordinary cycle classes; it does not prove that φ maps [V(τ)]_comb onto those classes. The theorem statement must include the compatibility hypothesis, and the proof must spell out the transfer through φ.","section":"Thm 5.2, degree-4 paragraph"},{"comment":"The degree-4 step for surfaces assumes the existence of a smooth maximal cone ('Since X is projective, it contains smooth points'). This is not guaranteed for an arbitrary complete fan: the 2D fan with rays (±1,±1) has no smooth maximal cone. The argument needs a replacement using a singular cone or a different proof that a nonzero class exists. As written, the proof of Theorem 5.1(3) is incomplete.","section":"Thm 5.1(3)"},{"comment":"For simplicial fans the proof asserts that '[V(τ)]_comb correspond to monomials in these divisor classes' without proof. The isomorphism IH^*_comb ≅ H^*(X) identifies the vector spaces, but identifying the specifically defined combinatorial cycle classes with torus-invariant subvariety classes is exactly the compatibility that has not been established. Thus Theorem 5.3 is also conditional on the unproved identification.","section":"Thm 5.3"},{"comment":"The claim that Conjecture 4.1 is equivalent to the combinatorial cycle classes spanning IH^{2k}_comb is only valid if the BBFK–BL compatibility holds. Since that compatibility is unproved, the equivalence is not established. Moreover, Hdg^k_comb was defined as the span of those classes, so 'spanning' is definitional; the substantive assertion is the compatibility φ([V(τ)]_comb)=[V(τ)]. The paper's low-dimensional arguments verify only the geometric generation statement and leave the combinatorial statement untouched.","section":"Conj. 4.1, equivalence paragraph"}],"minor_comments":[{"comment":"The term 'BBFK–BL compatibility' is used without a formal definition; it should be stated precisely in the introduction and clearly referenced in the statements of the main theorems.","section":"Abstract"},{"comment":"The Danilov–Jurkiewicz theorem is attributed to [3] (Cox), but the standard citation is Danilov's paper [5] or Fulton's book [7]. The current citation is misleading.","section":"§5.3"},{"comment":"The assertion that the combinatorial classes [V(ρ_i)]_comb are non-zero is not justified by any computation; an explicit description of the maps involved would be needed for the example to be illustrative.","section":"Example 3.1"},{"comment":"The algorithm presumes the existence of the very sheaf map and Gysin pushforward that are not defined in Section 3. The algorithm cannot be executed without resolving Definition 3.2.","section":"§6.1, Step 4"},{"comment":"Several minor typographical issues: spacing in 'IH ∗', the '2010 Mathematics Subject Classification' should be '2020', and the phrase 'Key words and phrases' is not standard for this journal format.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The manuscript has a promising idea but the central definitions are incomplete and the main theorems overclaim relative to the abstract. The referee report requests a major revision that would require constructing the combinatorial Gysin maps, proving or explicitly axiomatizing the BBFK–BL compatibility, and correcting the smooth-cone assumption in Theorem 5.1(3). If the author cannot supply the missing constructions, the paper should be reformulated as a conditional contribution with all hypotheses made explicit."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper's actual new content is the formulation of Conjecture 4.1: a combinatorial analogue of the intersection Hodge conjecture for projective toric varieties, with combinatorial Hodge classes defined as the span of proposed combinatorial cycle classes. That framing is coherent and worth thinking about. The authors also sketch an algorithmic rank-check procedure that could be useful if the underlying definitions were made precise. What they verify for n≤2 and for simplicial fans is, by their own admission, known material; the interesting part is the attempt to extend to n=3. The abstract is admirably explicit that this extension assumes the 'BBFK–BL compatibility.'\n\nThe soft spots are real and load-bearing. Definition 3.2 never constructs the combinatorial Gysin map (i_τ)_*; Remark 3.2 points to functoriality in the literature, but the degree shift, the fundamental class of Star(τ), and the required properties are all unspecified. Without that map, Hdg^k_comb is not well-defined, and the spanning claim has no precise content. The compatibility that would identify these combinatorial classes with geometric cycle classes under φ is exactly what the paper needs, and it is never proved or even stated as a separate explicit conjecture. Theorem 5.2's degree-4 step silently uses it, swapping in the geometric statement that intersecting divisors with an ample class gives curves. The stress-test note is right that this is the core gap. There is also a genuine but secondary flaw in Theorem 5.1(3), which assumes a smooth maximal cone that need not exist (the 2D fan with rays (±1,±1) is a counterexample), and the promised non-simplicial 3D example is missing. Metadata inconsistencies (title, affiliations) are minor but sloppy.\n\nThat said, the paper is not a sham. It is honestly framed as a conjecture, and the authors identify exactly what would need to be proved. The equivalence in Conjecture 4.1 is partly tautological because Hdg^k_comb is defined as a span, but that is a presentation issue; the real mathematical question is the compatibility, and the paper says so in the abstract. I would not cite the main theorem as a result, but I would bring the paper to a reading group to discuss what a rigorous cycle-class map in BBFK theory should look like.\n\nMy recommendation: send it to peer review, but with the expectation of major revision. The authors should construct the Gysin map or make the compatibility a separate, precisely stated conjecture, fix the smooth-cone step, and include the promised example. If they cannot fill those gaps, the published version should present the n=3 statement as conditional, not as a theorem.","headline":"A well-framed conjecture paper that is honest about its main assumption, but the central combinatorial cycle-class construction is not rigorous and the n=3 verification is conditional on an unproven compatibility.","tokens_in":6805,"tokens_out":1479,"would_cite":false,"duration_ms":17053,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14M25","14C30","52B20","14F43"],"pacs":[],"model":"deepseek-v4-flash","headline":"Combinatorial Hodge Conjecture for projective toric varieties: the classes of torus-invariant subvarieties span every even-degree rational intersection cohomology group, verified up to dimension three and for simplicial fans.","keywords":["toric varieties","intersection cohomology","Hodge conjecture","combinatorial cycle classes","fans","Hodge classes","simplicial fans","Hard Lefschetz theorem"],"falsifier":"Run the paper's rank check (Algorithm, Step 5) on a non-simplicial projective 4-dimensional fan with no smooth maximal cone—for instance, a fan whose maximal cones are all singular—and compare the rank of the candidate cycle-class matrix with the computed dimension of IH^{2k}_comb for each even degree; a shortfall in any degree would refute the spanning conjecture. For the compatibility premise, compute phi([V(tau)]_comb) on a singular weighted projective 3-fold and compare it with the geometric cycle class of V(tau); any nonzero difference would invalidate the stated compatibility.","tokens_in":1331,"feed_emoji":"🧮","tokens_out":8331,"duration_ms":115669,"temperature":0.7,"pith_summary":"This paper formulates a purely combinatorial version of the Intersection Hodge Conjecture for projective toric varieties. It attaches to each cone of a fan a combinatorial cycle class in the combinatorial intersection cohomology of the fan and conjectures that these classes span every even-degree group. Because projective toric intersection cohomology is Hodge-Tate, this spanning statement is exactly the assertion that all rational Hodge classes are algebraic. The authors verify the conjecture for projective toric varieties of dimension at most three under a stated compatibility between combinatorial and geometric cycle classes, and unconditionally for simplicial fans. If correct, the conjecture turns a deep geometric problem into a finite linear-algebra rank check on fans.","feed_headline":"Toric Hodge classes all come from invariant subvarieties in dim ≤ 3","feed_subtitle":"Even-degree intersection cohomology of any projective toric threefold is spanned by classes of torus-invariant subvarieties.","key_machinery":"The central object is the combinatorial cycle class [V(tau)]_comb := (i_tau)_*([Star(tau)]), the image of the fundamental class of the star of a cone under a combinatorial Gysin pushforward inside the minimal-extension sheaf complex. The argument also relies on the combinatorial Hard Lefschetz operator, which plays the role of intersecting with an ample divisor, and, in the simplicial case, on the description of rational cohomology as a quotient of the ring generated by rays. The paper's algorithm computes local intersection cohomology, assembles the global sheaf, forms the matrix of candidate cycle classes, and checks whether the rank matches the target dimension.","core_discovery":"The paper's central claim is Conjecture 4.1: the canonical isomorphism between combinatorial and geometric intersection cohomology maps the span of combinatorial cycle classes [V(tau)]_comb onto the geometric Hodge classes. Since projective toric intersection cohomology is Hodge-Tate, this is equivalent to asserting that the classes [V(tau)]_comb span IH^{2k}_comb(Sigma, Q) for every k. The paper proves this for all projective toric varieties of dimension n <= 3, assuming the stated BBFK-BL compatibility, and unconditionally for simplicial fans. In the simplicial case the proof uses the presentation of rational cohomology as a quotient of the ring generated by ray variables, so the combinato","pith_inferences":["The paper's low-dimensional proof is only as strong as the compatibility it assumes; a mismatch between [V(tau)]_comb and the geometric class of V(tau) would sever the link between the combinatorial theorem and the geometric Hodge conjecture.","The rank-check formulation means the conjecture for a given fan is decidable in principle; testing non-simplicial 4-dimensional fans, especially those lacking smooth maximal cones, is the natural next experiment.","If the combinatorial Gysin map is constructed explicitly rather than assumed, the same framework would likely prove the conjecture for all rational polytopal fans, not just low dimensions.","The reliance on smooth maximal cones in the top-degree argument suggests that a cleaner proof may come from local-to-global sheaf theory on fans, avoiding the need for smooth points."],"forward_implications":["For every projective toric 3-fold, every rational class in even-degree intersection cohomology is a Q-linear combination of classes of torus-invariant subvarieties.","For simplicial projective toric varieties (toric orbifolds), the same spanning statement holds unconditionally, recovering the known fact that cohomology is generated by invariant divisors.","The conjecture becomes a finite computational problem: a rank check on a matrix of combinatorial cycle classes attached to the fan.","If the conjecture holds in general, the Intersection Hodge Conjecture for projective toric varieties follows, since the combinatorial spanning statement is equivalent to the geometric one under the compatibility assumption.","The framework suggests a path to verifying the conjecture in higher dimensions by algorithmic rank computations rather than by geometric construction of cycles."],"fun_headline_variants":["Toric threefolds: invariant cycles span intersection cohomology","Invariant cycles generate all Hodge classes in toric 3-folds","Torus-invariant cycles generate all Hodge classes in dim ≤3","Combinatorial cycle classes span IH for projective toric 3-folds"],"cache_read_input_tokens":7936,"weakest_assumption_plain":"The conjecture collapses if the combinatorial class assigned to a cone does not match, via the canonical isomorphism, the geometric class of the corresponding invariant subvariety; the low-dimensional proof further assumes this compatibility and, for top-degree classes, the presence of a smooth maximal cone.","fun_headline_variants_meta":{"raw":{"variants":["Toric threefolds: invariant cycles span intersection cohomology","Invariant cycles generate all Hodge classes in toric 3-folds","Torus-invariant cycles generate all Hodge classes in dim ≤3","Combinatorial cycle classes span IH for projective toric 3-folds"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001225,"raw_usage":{"total_tokens":4883,"prompt_tokens":765,"completion_tokens":4118,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":509,"completion_tokens_details":{"reasoning_tokens":4038}},"tokens_in":509,"tokens_out":4118,"duration_ms":26803,"temperature":1.0,"reasoning_tokens":4038,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T18:08:05.714529+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the paper's rank check (Algorithm, Step 5) on a non-simplicial projective 4-dimensional fan with no smooth maximal cone—for instance, a fan whose maximal cones are all singular—and compare the rank of the candidate cycle-class matrix with the computed dimension of IH^{2k}_comb for each even degree; a shortfall in any degree would refute the spanning conjecture. For the compatibility premise, compute phi([V(tau)]_comb) on a singular weighted projective 3-fold and compare it with the geometric cycle class of V(tau); any nonzero difference would invalidate the stated compatibility.","supporting_citations":[],"review_version":1}