{"id":"39148e32-a023-4a11-989a-a20906a18cf8","arxiv_id":"2607.08612","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"A new pointwise invariant, the Point-Cross Dimension, quantifies the cumulative weighted complexity of independent directional channels through a single germ, separating directionality from isotropic dispersion.","lead":"The paper introduces a new pointwise geometric invariant called the Point-Cross Dimension, which measures how many independent directions coexist at a single point in a set. It provides a local, directional notion of dimension that can exceed classical fractal dimensions at crossing points, potentially useful for analyzing singular and fractal geometries.","discovery_kind":"unclear","skeptic_critique":{"model":"glm-5.2","headline":"Shadowing Lemma's global Lipschitz estimate across countably many disjoint balls needs verification that the constant-2 bound survives the infinite gluing.","rationale":"The reader correctly identifies the Shadowing Lemma as the load-bearing technical step. Having examined the argument in detail, the global Lipschitz estimate across countably many disjoint balls appears sound: the key observation is that disjoint subsegments of [z,w] have total length at most ||z-w||, so even with factor-2 inflation per subsegment, the global bound Lip(H) ≤ 2 follows. The o(||z-x||) condition is also correctly verified. The concern is legitimate in that this is the most intricate step in the paper and is presented at a level of detail that a careful reader would want to expand, but I do not find an actual error. The main limitation of the review remains the truncated manuscript: Chapter 5 (comparison principles) and several model computations (Sierpiński carpets, infinite-rank outlook) cannot be verified. The CONDITIONAL verdict is appropriate given this incompleteness, but the specific concern about the Shadowing Lemma does not, upon close inspection, appear to be an actual defect. The theory is internally consistent where verifiable, the three-layer hierarchy is well-motivated, and the calibration on geometric graphs and smooth manifolds is correct. The C^1-invariance (rather than bi-Lipschitz) is an honest and well-explained limitation stemming from the sensitivity to ambient linear rank, not a flaw.","tokens_in":60768,"tokens_out":923,"duration_ms":368459,"concrete_test":"Independently verify the global Lipschitz estimate for H in Lemma 3.34/3.35 by checking: (1) that the decomposition of segment [z,w] into countably many subsegments intersecting the disjoint balls {B(p_i, r_i)} is well-defined, (2) that monotone convergence of subsegment lengths justifies the countable sum, and (3) that the same argument applies to H^{-1}. If any step fails for a configuration where the balls B(p_i, r_i) accumulate densely along the segment, the closure invariance Proposition 3.37 would need modification.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The reader correctly identifies the Shadowing Lemma (Lemma 3.35) as load-bearing for closure invariance (Proposition 3.37). The construction patches countably many bi-Lipschitz maps H_i, each supported on pairwise disjoint balls B(p_i, r_i) with Lip(H_i) ≤ 2 and Lip(H_i^{-1}) ≤ 2. The global Lipschitz estimate Lip(H) ≤ 2 is argued by decomposing any segment [z,w] into subsegments lying inside or outside the balls, summing lengths with a factor-2 bound per interior subsegment, and passing to a limit via monotone convergence of subsegment lengths. The concern is whether this summation is fully rigorous: the segment [z,w] can intersect countably many disjoint open balls, producing countably many interior subsegments. The total length of interior subsegments is at most ||z-w|| (since they are disjoint subintervals of [z,w]), and each is inflated by at most factor 2, so the total inflation is at most 2·||z-w||, giving Lip(H) ≤ 2. This argument appears sound. However, the inverse requires the same estimate for H^{-1}, which patches H_i^{-1} with the same Lipschitz constant 2 on the same disjoint balls. The same summation applies, so Lip(H^{-1}) ≤ 2. The more subtle point is the condition r_i/||p_i - x|| → 0 ensuring H(z) - z = o(||z - x||), which is needed to preserve the limiting direction of the probe. This is verified in the proof: for z near x inside B(p_i, r_i), ||H(z)-z|| ≤ r_i/4 ≤ η||p_i-x||/4 while ||z-x|| ≥ (1-η)||p_i-x||, giving ratio ≤ η/(4(1-η)) → 0. This is correct. The construction appears to hold as stated, though the infinite gluing argument is presented at a sketch level that warrants careful line-by-line verification.","agreement_with_reader":"agree"},"referee_report":{"model":"glm-5.2","summary":"The paper introduces the Point-Cross Dimension, a new pointwise invariant measuring the directional organization of a set at a single point. The construction proceeds in three layers: (1) the point-vector dimension dim_Pvec, counting exact local segment directions; (2) the point-tangential dimension dim_Ptan, counting asymptotically accessible Bouligand tangent directions; (3) the Point-Cross dimension dim_×, which weights effective projective directions by the point-extended box complexity detected along admissible Lipschitz probes and aggregates over projectively independent channels. The main structural results establish the hierarchy dim_Pvec ≤ dim_× ≤ dim_Ptan, C^1-diffeomorphism invariance, closure invariance, a Grassmann-type union formula, and calibration on smooth submanifolds and geometric graphs. The paper also introduces a combined Point-Box-Cross dimension dim_⊠ = max(dim_Pbox, dim_×). Model examples include oscillatory germs, fractal coordinate frames, and Sierpiński-type carpets. The final part develops comparison principles between directional and dispersive layers.","tokens_in":61874,"tokens_out":1402,"duration_ms":302460,"significance":"The paper addresses a genuine gap in local dimension theory: classical isotropic dimensions (Hausdorff, box, packing, Assouad) obey a max-rule on finite unions and thus collapse the directional richness present at crossing points. The n-axis frame in R^n carrying pointwise dimension 1 everywhere under classical dimensions is a compelling motivating example. The graded per-direction weight θ_A_x(ξ) ∈ [0,1], combined with aggregation over projectively independent directions, is a novel construction that appears to differ from existing directional notions (tangential dimensions of Guido–Isola, direction sets of Koike–Paunescu, Assouad spectra, decomposability bundles of Alberti–Marchese) in the precise sense summarized in Table 3.1. The C^1-invariance (Theorem 4.7) and the smooth calibration (Proposition 3.51) are clean results. The collapse on geometric graphs (Proposition 3.56) serves as a useful calibration check. The comparison with existing notions in Section 3.7 is thorough and helps position the contribution.","major_comments":[{"comment":"The Shadowing Lemma (Lemma 3.35) is the load-bearing technical step for closure invariance (Proposition 3.37), which in turn underpins the well-definedness of the theory on closed germs. The construction patches countably many bi-Lipschitz maps H_i on pairwise disjoint balls B(p_i, r_i). The global Lipschitz estimate Lip(H) ≤ 2 is argued via decomposition of segments into subsegments and monotone convergence. While the argument appears sound on inspection, the infinite gluing across countably many balls is subtle enough that a brief explicit verification of the summation argument—perhaps as a dedicated paragraph or a reference to a standard gluing lemma for bi-Lipschitz maps on disjoint supports—would strengthen the paper. Specifically, the claim that the segment [z,w] intersects at most countably many disjoint open balls, producing subsegments whose total inflation is bounded by 2||z-w−","section":null}],"minor_comments":[{"comment":"Sections 1.1–1.3 (philosophical, mathematical, and historical motivations) span approximately 5 pages. While the author explicitly states that the phenomenological analysis is not foundational for the mathematics, the length is disproportionate relative to the technical content. Consider condensing to 2–3 pages, retaining the key motivating examples (Figures 1.1, 1.2).","section":null},{"comment":"The Lebesgue quotation in French is untranslated. A brief English paraphrase or translation would aid readers unfamiliar with French.","section":null},{"comment":"In Definition 3.2, the condition A ∩ γ((0,η]) ≠ ∅ for every η ∈ (0,δ] is the recurrence condition. This is clear, but the term 'recurrence' is not introduced before its use in the proof of Proposition 3.4. A brief gloss when first used would help.","section":null},{"comment":"Remark 3.8 discusses the orientation convention (v vs −v) and the passage from spherical to projective directions. The distinction between oriented and projective formulations recurs throughout Section 3. A summary diagram or a consolidated remark collecting all conventions would improve readability.","section":null},{"comment":"In Proposition 2.41, the Grassmann formula for dim_Ptan on unions is exact, but Remark 2.43 notes that the overlap term is not dim_Ptan{x}(A∩B). This is an important subtlety. Consider flagging it more prominently, perhaps with a labeled cautionary example.","section":null},{"comment":"The notation θ^A_x(v) uses a superscript for the set A, which can be confused with exponentiation. Consider θ_A(x,v) or θ(x,v;A).","section":null},{"comment":"Table 3.1 is helpful but could benefit from a column indicating whether each notion is defined pointwise (at every point) or almost everywhere, since this is a key distinction for the decomposability bundle row.","section":null},{"comment":"The paper references 'Proposition 5 in [25]' and 'Corollary 2 in [25]' for upper semicontinuity of dim_Pbox (Remark 3.23) and 'Remark 7 in [25]' for locality. Since these are load-bearing properties, brief restatements of the relevant statements would make the paper more self-contained.","section":null},{"comment":"In Example 3.10(2), the computation of Eff_{(0,0)}(Γ_{α,β}) for the three regimes (α>1, α=1, 0<α<1) is detailed but the role of β (oscillatory frequency) is mentioned only at the end. Stating upfront that β does not affect the directional support would orient the reader.","section":null},{"comment":"The bibliography appears to use a non-standard format. Ensure consistency with the journal's citation style.","section":null}],"recommendation":"minor_revision","confidential_remarks":"The paper is a continuation of the author's prior work [24, 25] and is largely self-contained mathematically. The philosophical framing is unusual for a math.MG paper but the author is transparent about its non-foundational role. The main technical risk is concentrated in the Shadowing Lemma, which I have verified to my satisfaction but which deserves more explicit treatment. The novelty claim relative to existing directional notions (Section 3.7) appears well-supported. The paper is appropriate for the journal's scope in geometric measure theory and fractal geometry."},"author_rebuttal":null,"desk_editor":{"model":"glm-5.2","letter":"The main thing to know: this paper introduces a genuinely new pointwise invariant, the Point-Cross Dimension dim×, which measures how many independent effective directions coexist at a single point. The construction is a three-layer hierarchy — point-vector dimension (exact segment directions), point-tangential dimension (Bouligand tangent directions), then the Point-Cross dimension which weights each effective direction by the point-extended box complexity detected along admissible Lipschitz probes. The key structural result is the sandwich dimPvec ≤ dim× ≤ dimPtan, with equality on smooth manifolds and geometric graphs. This is not in the existing literature; the graded weight θ_A_x(ξ) ∈ [0,1] combined with projective independence aggregation is new, and the comparison table in Section 3.7 honestly distinguishes it from Guido-Isola tangential dimensions, Koike-Paunescu direction sets, and Alberti-Marchese decomposability bundles. The calibration on standard geometries (crosses, coordinate frames, oscillatory germs) is correct where I checked. The Grassmann-type union formula for dimPtan (Proposition 2.41) is clean and the failure of upper semicontinuity (Proposition 4.3) is a nice honest observation. The C^1-invariance theorem (4.7) is correct, and the author is upfront that this is not a bi-Lipschitz invariant — the three-coplanar-lines vs three-independent-lines example in Remark 4.9 makes the obstruction concrete. The stress-test concern about the Shadowing Lemma (3.35) is the right thing to flag, but I think it actually holds. The global Lipschitz estimate across countably many disjoint balls works because the interior subsegments of any segment [z,w] are disjoint subintervals with total length at most ||z-w||, each inflated by factor 2, giving Lip(H) ≤ 2. The same argument applies to the inverse. The condition r_i/||p_i - x|| → 0 ensuring H(z)-z = o(||z-x||) is verified correctly in the proof. The argument is presented at sketch level but the details check out. The real soft spot is that the manuscript is truncated — Chapter 5 (comparison principles) and several promised computations (Sierpiński carpets, infinite-rank outlook) are not in the text I can see. The philosophical motivation runs about three pages longer than it needs to. The invented entities are all properly defined and not circular; the Point-Extended Box Dimension from the author's prior work is used as a standard input tool. This is for researchers in fractal geometry and geometric measure theory who care about local directional structure. It deserves a serious referee who can verify the missing Chapter 5 computations and push for tighter exposition.","headline":"New pointwise directional dimension theory with a load-bearing shadowing lemma that holds up under scrutiny","tokens_in":61634,"tokens_out":644,"would_cite":true,"duration_ms":111582,"reading_group":"no","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"glm-5.2","headline":"Crossing lines carry hidden dimension that classical fractals miss","keywords":[],"falsifier":"Find a closed set A ⊂ R^n and a point x ∈ A where a direction v is an effective Bouligand direction and the closure trace Ā ∩ Γ_γ along some probe has point-extended box dimension s > 0, but where no admissible probe meeting A itself can recover dimension s—i.e., where the Shadowing Lemma's bi-Lipschitz construction degrades the separated-set cardinality so badly that the box exponent drops. This would break θ^A_x(v) = θ^{Ā}_x(v) and hence closure invariance of dim_×.","tokens_in":60937,"feed_emoji":"✚","tokens_out":994,"duration_ms":204772,"temperature":0.7,"pith_summary":"The paper introduces the Point-Cross Dimension, a new point-by-point invariant for subsets of Euclidean space that measures how many independent directional channels coexist at a single point. Classical fractal dimensions—Hausdorff, box, packing, Assouad—obey a max-rule on unions: the dimension of two intersecting sets is just the larger of the two. This erases the local interaction at a crossing. The Point-Cross Dimension is built to recover that interaction. At a point where two transverse line segments meet, it returns 2 rather than 1, because two genuinely independent directions are simultaneously active. The construction proceeds in three layers. First, a point-vector dimension counts exact straight-segment directions issuing from the point. Second, a point-tangential dimension replaces exact segments by Bouligand tangent directions, capturing asymptotic approach directions even when no straight segment is present. Third, the full Point-Cross Dimension weights each effective tangent direction by the point-extended box complexity detectable along admissible Lipschitz probes approaching the point in that direction, then sums these weights over projectively independent directions. The result is an invariant satisfying dim_Pvec ≤ dim_× ≤ dim_Ptan, agreeing with manifold dimension on smooth submanifolds, recovering the incident-edge rank on geometric graphs, and invariant under local C^1-diffeomorphisms and closure. The paper then computes the invariant on oscillatory germs, fractal coordinate frames, Sierpiński-type carpets, and Cantor dusts, and establishes comparison principles showing when directional richness forces dispersive largeness and when it does not.","feed_headline":"Crossing lines carry hidden dimension classical fractals miss","feed_subtitle":"A new pointwise invariant detects directional coexistence at crossings, returning 2 where box dimension sees only 1","key_machinery":"The directional contribution θ^A_x(v) is the central object. It is defined as the supremum, over all admissible Lipschitz probes γ approaching x with limiting direction v, of the point-extended box dimension dim_Pbox{x}(A ∩ Γ_γ). Each probe is an injective Lipschitz curve whose image has box dimension at most 1, so each direction contributes at most 1. The Point-Cross Dimension then sums these per-direction weights over projectively independent effective directions. The Shadowing Lemma (Lemma 3.35) is the load-bearing technical step for closure invariance: it constructs a bi-Lipschitz ambient homeomorphism supported in disjoint balls around points of the closure trace, moving them onto an ad","core_discovery":"The Point-Cross Dimension dim_×{x}A is defined by assigning to each effective Bouligand tangent direction v a directional contribution θ^A_x(v) = sup over admissible Lipschitz probes γ of the point-extended box dimension of A ∩ Γ_γ, then aggregating: dim_×{x}A = sup over projectively independent families (ξ_1,…,ξ_m) of Σ θ^A_x(ξ_i). This invariant is squeezed between the point-vector dimension (exact segment rank) and the point-tangential dimension (Bouligand tangent span rank), equals k on C^1 k-dimensional submanifolds, equals the linear rank of incident edge directions on geometric graphs, is invariant under closure and local C^1-diffeomorphisms, and—crucially—exceeds the classical max of","pith_inferences":[],"forward_implications":[],"fun_headline_variants":["New pointwise dimension counts independent directions at crossings","Point-Cross Dimension separates directional structure from local covering complexity","Invariant detects coexisting directional channels classical dimensions collapse","Point-Cross Dimension weights effective tangent directions by local box complexity","Cross-based local dimension distinguishes dispersion from directional independence"],"cache_read_input_tokens":0,"weakest_assumption_plain":"The closure invariance of the directional contribution relies on a Shadowing Lemma that constructs a bi-Lipschitz homeomorphism across infinitely many disjoint balls to transfer box-dimensional mass from closure traces back to the original set. If the bi-Lipschitz constants accumulate badly across infinitely many disjoint balls, the point-extended box exponent might not survive the transfer, which would compromise the well-definedness of the theory on closed germs.","fun_headline_variants_meta":{"raw":{"variants":["New pointwise dimension counts independent directions at crossings","Point-Cross Dimension separates directional structure from local covering complexity","Invariant detects coexisting directional channels classical dimensions collapse","Point-Cross Dimension weights effective tangent directions by local box complexity","Cross-based local dimension distinguishes dispersion from directional independence"]},"model":"glm-5.2","effort":"low","cost_usd":0.0,"raw_usage":{"total_tokens":727,"prompt_tokens":667,"completion_tokens":60,"prompt_tokens_details":null},"tokens_in":667,"tokens_out":60,"duration_ms":63533,"temperature":1.0,"reasoning_tokens":null,"cache_read_input_tokens":0,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-10T04:24:22.103826+00:00","model_set":{"reader":"glm-5.2"},"falsifier":"Find a closed set A ⊂ R^n and a point x ∈ A where a direction v is an effective Bouligand direction and the closure trace Ā ∩ Γ_γ along some probe has point-extended box dimension s > 0, but where no admissible probe meeting A itself can recover dimension s—i.e., where the Shadowing Lemma's bi-Lipschitz construction degrades the separated-set cardinality so badly that the box exponent drops. This would break θ^A_x(v) = θ^{Ā}_x(v) and hence closure invariance of dim_×.","supporting_citations":[],"review_version":1}