{"id":"73bad90a-b536-448a-a1c9-5df558867040","arxiv_id":"2501.10728","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"ParkView encodes both shift maps of a monotone interleaving as colored active paths and hedges, with an optimal path-branch decomposition computable in linear time and a 3-color guarantee.","lead":"A new schematic drawing, ParkView, shows how two merge trees are matched by a monotone interleaving using colored path segments and enclosing shapes. It comes with proofs that an optimal decomposition is computable in linear time and that at most three colors per tree are needed.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The active-path-top property asserted in Section 2.2 is never proved, and the hedge-to-active-path matching (R0) collapses if it fails; Lemma 1 comes close but the paper omits the corollary and the infinite-path case.","rationale":"The reader's weakest assumption identifies exactly the gap I consider most load-bearing: the unproved active-path suffix property in Section 2.2. The paper's own R0 requirement asserts that a user can determine the image of any part of the tree, but the visual matching between a hedge and an active path relies on the branch image being a contiguous top suffix. Lemma 1 establishes per-component contact with the top edge, which strongly suggests the property is true for finite paths, but the corollary is never articulated, and the infinite path is not handled. The order-matching property is likewise asserted without proof, although monotonicity likely implies it. The rest of the theoretical core, including Theorem 1 and Theorem 2, appears sound on close reading; the flaws I noticed are presentational (e.g., the 'distinct path' wording in Lemma 1, the unmentioned constant contribution of the infinite path in Theorem 1's counting argument) and do not change the conclusions. The lack of a user evaluation and the authors' own scalability caveat for 900+ leaves further support a conditional verdict. Since my concern matches the reader's and does not shift the verdict, I recommend UNCHANGED.","tokens_in":18952,"tokens_out":41940,"duration_ms":445933,"concrete_test":"Build an exhaustive brute-force checker over all ordered merge trees with at most 5 leaves and all monotone δ-shift maps at small integer heights: for each heavy path decomposition, compute each Bπ and check that α(Bπ) is a suffix of π, and that ordering the hedges by their leftmost/lowest leaf reproduces the x-order of active paths. Also ask the authors to add a short lemma deriving the suffix property from Lemma 1 and monotonicity, covering the infinite path; if the derivation cannot be completed, the ParkView encoding does not satisfy R0 as claimed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 2.2 states that for every path π in the path decomposition, if Bπ is nonempty then α maps Bπ to a contiguous suffix of π, placing the square glyph at the top of π and making the hedge's height equal to the active path's height. This structural assertion is never proven; it is load-bearing because R0 (reconstructing the shift maps) and the color/order/height matching all depend on it. Lemma 1 shows each connected component of Bπ contains a point mapping to the top edge e of π, which implies, by continuity and height monotonicity, that each component's image is a segment ending at e, so the union is a suffix for finite paths. But the paper does not state or prove this corollary, and the unique path to infinity has no finite top edge, so the property is not obviously covered. The left-to-right order claim for matching hedges to active paths is also asserted without proof. If any of these fail, a user cannot uniquely determine which hedge maps to which active path, so the central completeness claim is unsupported.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces ParkView, a schematic visual encoding for monotone interleavings between ordered merge trees. The proposed drawing decomposes each tree into paths and represents the preimage of each path under a shift map as a histogram-shaped hedge; the image of a branch is drawn as a colored active path on the other tree. The authors prove that a heavy path decomposition minimizes both the total number and the maximum number of branch components (Theorem 1), that the resulting hedges are 3-colorable (Theorem 2), and that an optimal decomposition can be computed in linear time (Section 4). The encoding is designed to satisfy four visual requirements R0-R3, and the paper demonstrates it on real-world ionization-front and volcanic datasets. The supplementary material contains the omitted proofs of the geometric properties used in the coloring argument and a large-scale example.","tokens_in":19150,"tokens_out":15077,"duration_ms":155374,"significance":"If the construction is correct, ParkView is a genuinely useful contribution: it gives a compact, order-respecting drawing from which both interleaving maps can in principle be reconstructed, and it scales to trees with many leaves while using only six colors. The paper's strength is its formal component: Theorem 1 is proved from the definitions, Theorem 2 reduces 3-colorability to three local geometric properties that are proved in the appendix, and the authors make their implementation publicly available and test it on real data. The paper is also honest about limitations, including readability at 900+ leaves and the open problem of constructing stable leaf orders. The main weakness is that two structural properties that are load-bearing for the completeness claim (R0) are stated in Section 2.2 without proof, and one algorithmic claim in Section 4 is only sketched.","major_comments":[{"comment":"The statement that a nonempty branch Bπ maps to a contiguous suffix π* of π, so that the active path always forms the top part of π, is load-bearing for requirement R0 and for the height- and color-based matching, but it is never proved. Lemma 1 in Section 3.1 shows that each connected component of Bπ contains a point mapping to the top edge of π, which comes close, but the paper does not state the corollary that the image of each component is a connected interval ending at the top of π, nor does it treat the unique path that extends to infinity, for which there is no finite top edge. Please add a proof of the suffix property for all paths, including the infinite path, or, if the property can fail, give a concrete counterexample and adjust the encoding.","section":"Section 2.2 (Active paths)"},{"comment":"The claim that the left-to-right order of the lowest leaf in each hedge matches the left-to-right order of the corresponding active paths is asserted without proof, and it is explicitly used in Section 5 to resolve the matching in the path-3 example. This order property is essential for R0 because it allows the user to disambiguate the hedge-to-active-path matching. Please provide a proof or a precise reference, and specify how the property is defined for compound branches, which have several connected components and hence potentially several lowest leaves.","section":"Section 2.2 (Properties) and Section 5 (Showcase)"},{"comment":"The linear-time computation of the heavy path-branch decomposition relies on precomputing the values |Be|, the number of connected components of the preimage of each edge e of T'. The text says this is done by 'simultaneously traversing the trees' and 'iteratively traversing T from each leaf to its root', but it does not explain how connected components are counted rather than merely points. Since linear-time optimal decomposition is one of the central claims, please give a precise description of the component-counting procedure and its linear-time bound, or point to the specific part of the released implementation that realizes it.","section":"Section 4 (Algorithm 1 and running-time paragraph)"}],"minor_comments":[{"comment":"The phrase 'hedges and active paths have the same height' is ambiguous: it is reconciled with the later statement that the top of each hedge is δ lower than the top of the active path only if 'height' means vertical extent, not top position. Please clarify this terminology.","section":"Figure 1 caption"},{"comment":"In the path-3 discussion, the sentence 'as path 3 is left of path 2, it must map to hedge 3 as it the leaf it encloses is left of the lowest leaf hedge 2 encloses' contains a typo ('as it the leaf') and is hard to parse; please rewrite it.","section":"Section 5 (Showcase)"},{"comment":"The abstract's 'maximum of 6 colors' is not directly tied to Theorem 2, which proves 3-colorability per tree; adding a sentence explaining that the total of six comes from using two 3-colorings, one for each shift map, would help readers connect the two statements.","section":"Abstract and Section 3.2"},{"comment":"The construction of leaf orders via a space-filling curve is described in two sentences, and the paper itself notes that stable orders are an open question; a brief comment on the observed stability of the space-filling-curve orders in the examples would strengthen the proof-of-concept.","section":"Section 5 (Pipeline)"},{"comment":"The supplementary 900-leaf example is reported as difficult to read; consider adding a sentence in Section 5 that summarizes this scalability limitation in the main text, since it is relevant to the paper's scalability claims.","section":"Supplementary material"}],"recommendation":"major_revision","confidential_remarks":"The paper is a solid design/theory contribution for visualizing monotone interleavings, and the core theorems are plausible and well-motivated. The main concern is that two load-bearing structural properties in Section 2.2 (the suffix property of active paths and the left-to-right order matching) are stated without proof; both appear repairable with short arguments building on Lemma 1 and the monotonicity definition in the appendix. The linear-time weight computation in Section 4 also needs a more precise description. The lack of a user study is worth noting given the R0-R3 claims, but for a primarily theoretical design paper this is acceptable if the structural claims are made rigorous. I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read this if you care about merge tree interleavings and visualization. The genuinely new part is the optimal path-branch decomposition (Theorem 1) and the 3-colorability result (Theorem 2), both backed by proofs in the appendix and an open-source implementation. The visual design is sensible and the real-world figures are informative; the authors are honest about scalability breaking down around 900 leaves.\n\nThe soft spot is the unproved \"active-path-top\" property in Section 2.2. The paper asserts that the image of a branch is a contiguous suffix of its path and that the square glyph sits at the path's top. This is load-bearing for R0. The stress-test note is right that it is never stated as a formal claim. But it is not a hidden flaw: Lemma 1 shows each connected component of a branch contains a point mapping to the top edge, and continuity plus height monotonicity give the suffix property for finite paths. The paper should add the one-paragraph corollary. The unique path to infinity lacks a top edge and needs a separate sentence; that is a minor gap, not a fatal one.\n\nThe left-to-right order matching between hedges and active paths is also asserted more than proved. It is plausible from the monotonicity definition and Observation 1, but a formal statement would strengthen the paper. R0-R3 are not evaluated with users or baselines; the authors treat them as design goals and show illustrative examples. For a VIS-style paper that is acceptable but leaves the effectiveness claims softer than the theory claims.\n\nOverall: the mathematics is coherent, the algorithms are linear time, and the code is available. The paper is not overclaiming. It deserves peer review with requests to prove the missing corollary and add a short evaluation or clearer scope language. I would cite it if I worked on merge-tree comparison.","headline":"Solid theory for a new merge-tree interleaving visualization; the main missing proof is a small corollary, not a flaw.","tokens_in":19703,"tokens_out":2485,"would_cite":true,"duration_ms":25253,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["55N31","68U05"],"pacs":[],"model":"deepseek-v4-flash","headline":"ParkView is a compact schematic encoding of monotone interleavings that lets a viewer reconstruct both shift maps, read off the interleaving height $\\delta$, and see mapped and unmapped regions, using three colors per tree and a…","keywords":["merge trees","monotone interleavings","interleaving distance","path-branch decomposition","heavy path decomposition","3-colorable drawings","topological data analysis","scalar field visualization"],"falsifier":"Enumerate all ordered merge trees with a small number of leaves, say four or five, and all monotone $\\delta$-shift maps between them; for each path in the heavy decomposition, compute the preimage branch and check whether its image is a contiguous suffix ending at the path's top. Any example where the image has a gap, or ends below the top, would refute the completeness of ParkView's matching rules.","tokens_in":1699,"feed_emoji":"🌳","tokens_out":1699,"duration_ms":93592,"temperature":0.7,"pith_summary":"The paper introduces ParkView, a schematic visual encoding for monotone interleavings between two ordered merge trees. A monotone interleaving consists of two shift maps that respect the trees' leaf orders; ParkView compresses both maps into one drawing by decomposing each tree into paths and drawing the preimage of each path as a colored rectilinear hedge that links to a colored active path in the other tree. The authors prove that a heavy path-branch decomposition is optimal, meaning it minimizes the number of branch components both per path and in total, and that the resulting hedges can always be colored with three colors per tree, hence at most six colors overall, regardless of tree size. They also give a linear-time algorithm for constructing the encoding. If these claims hold, ParkView lets a user reconstruct both shift maps, read the interleaving height $\\delta$ relative to the trees, and see which regions are mapped, ignored, or share an image, at the scale of roughly one hundred leaves.","feed_headline":"Six colors suffice to draw any monotone interleaving","feed_subtitle":"Both shift maps become readable in one drawing, with the offset delta and mapped or unmapped regions visible at a glance.","key_machinery":"The load-bearing object is the heavy path-branch decomposition. For a shift map $\\alpha$ from $T$ to $T'$, a path decomposition of $T'$ is obtained by choosing, at each internal vertex, one down edge as the through edge; each resulting leaf-to-root path $\\pi$ has a preimage branch $B_\\pi$ in $T$. Choosing the down edge of maximum weight, where weight is the number of connected components of the preimage of that edge, gives a heavy decomposition, and Theorem 1 shows this choice is optimal for both component-count objectives. The visual machinery is the hedge: a histogram-shaped union of tree bars, fillers, and bridges whose top height equals the top of the branch and whose left-to-right position follows the leaf order. Three structural properties of these hedges, namely interior-disjointness, having at most one parent, and having no hedge adjacent to the bottom of a longest bar, together yield the 3-colorability proof.","core_discovery":"ParkView claims that a monotone interleaving is completely and compactly captured by an optimal decomposition of both ordered merge trees into paths, together with the branches these paths pull back in the opposite tree. By choosing at every internal vertex the heaviest down edge, meaning the one whose preimage has the most connected components, the heavy path-branch decomposition simultaneously minimizes the maximum number of branch components per path and the total number of branch components (Theorem 1), and a greedy traversal computes it in linear time. Each non-empty branch is drawn as a histogram-shaped hedge and maps to one contiguous active path at the top of its path, so the two shift maps can be superimposed: hedges and active paths are matched by left-to-right leaf order, by equal height, and by color, and the paper proves the hedge set is 3-colorable (Theorem 2). The accompanying design elements, including square glyphs, grid lines at multiples of $\\delta$, and narrowed columns for unmapped regions, are chosen to satisfy the stated requirements R0--R3: reconstruct the image of any point, read $\\delta$, identify mapped versus unmapped regions, and detect equal-image points. The paper demonstrates the encoding on ionization-front and volcanic-eruption scalar fields.","pith_inferences":["Beyond the paper: the 3-coloring argument applies to any histogram drawing satisfying the three stated properties, so the same color bound should transfer to other rectilinear encodings of tree matchings, not only ParkView.","Beyond the paper: the usefulness of ParkView for time-series analysis hinges on choosing stable, meaningful leaf orders; the space-filling-curve order used in the demos is a heuristic, and a stability guarantee for order construction would be a natural next step.","Beyond the paper: the contiguity-of-active-paths assumption could be stress-tested by exhaustive search over small ordered merge trees; if a counterexample exists, ParkView would need a modified glyph marking the actual image endpoints to remain complete.","Beyond the paper: an interactive folding of inactive columns, not just static compression, would directly address the cluttered large-tree case that the supplementary material shows for trees with over nine hundred leaves."],"forward_implications":["A user can recover either shift map point by point: every point inside a hedge maps to the corresponding active path, so requirement R0 is met by the drawing alone.","The interleaving height $\\delta$ is read as the vertical gap between the topmost square glyph and the hedge below it, giving requirement R1 without extra computation.","Regions with no active path are drawn in narrow columns, so mapped versus unmapped parts of each tree are visually separated (R2), and shared images can be spotted through matching hedges (R3).","Because heavy decompositions are optimal and computable in linear time, the encoding can be produced for ordered merge trees with roughly one hundred leaves, the scale the design targets.","Only three colors per tree are ever needed, and with a separate hue for each of the two shift maps the full drawing uses at most six colors, keeping the color channel discriminable for arbitrary tree size."],"supporting_citations":[{"why":"Supplies the monotone interleaving model and the relation to Fréchet distance that the pipeline uses to obtain an interleaving from ordered merge tree curves, and provides the ordered-merge-tree formalism used in the appendix.","marker":"[2]"},{"why":"Defines interleavings and the labeled interleaving distance, giving the shift-map formalism and prior visualization context that ParkView extends.","marker":"[9]"},{"why":"One of the sources for the interleaving distance between merge trees and the definition of $\\delta$-shift maps that underlies the visualization requirements.","marker":"[14]"},{"why":"Provides the rectilinear merge tree drawing style that ParkView adapts, along with the ionization-front and volcanic datasets used in the showcase.","marker":"[19]"},{"why":"Supplies the branch-decomposition terminology and the edit-distance context that ParkView's path and branch decomposition follows and reinterprets.","marker":"[31]"}],"fun_headline_variants":["Six colors, linear time, no clutter: monotone interleavings","Optimal path decomposition draws any monotone interleaving","ParkView: monotone interleavings drawn with 6 colors","Heavy paths cut monotone interleavings to 3-colorable hedges","Linear-time drawing of interleavings via heavy paths"],"cache_read_input_tokens":21888,"weakest_assumption_plain":"The encoding assumes that every non-empty branch maps to one contiguous stretch of its path ending exactly at the path's top, a property stated in Section 2.2 but not proved; if a branch's image could have gaps or stop short, the hedge-to-active-path matching and the square glyph would not line up, and the drawing would no longer be complete.","fun_headline_variants_meta":{"raw":{"variants":["Six colors, linear time, no clutter: monotone interleavings","Optimal path decomposition draws any monotone interleaving","ParkView: monotone interleavings drawn with 6 colors","Heavy paths cut monotone interleavings to 3-colorable hedges","Linear-time drawing of interleavings via heavy paths"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000228,"raw_usage":{"total_tokens":1515,"prompt_tokens":1024,"completion_tokens":491,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":640,"completion_tokens_details":{"reasoning_tokens":397}},"tokens_in":640,"tokens_out":491,"duration_ms":4301,"temperature":1.0,"reasoning_tokens":397,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T19:04:25.015322+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Enumerate all ordered merge trees with a small number of leaves, say four or five, and all monotone $\\delta$-shift maps between them; for each path in the heavy decomposition, compute the preimage branch and check whether its image is a contiguous suffix ending at the path's top. Any example where the image has a gap, or ends below the top, would refute the completeness of ParkView's matching rules.","supporting_citations":[{"cited_title":"Agarwal, K","cited_arxiv_id":null,"evidence_quote":"Supplies the monotone interleaving model and the relation to Fréchet distance that the pipeline uses to obtain an interleaving from ordered merge tree curves, and provides the ordered-merge-tree formalism used in the appendix."},{"cited_title":"Pont and J","cited_arxiv_id":null,"evidence_quote":"Provides the rectilinear merge tree drawing style that ParkView adapts, along with the ionization-front and volcanic datasets used in the showcase."}],"review_version":1}