{"id":"24a65d4d-e966-4988-9094-44f5be309310","arxiv_id":"2505.04074","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"The diameter of modular flip-graphs for non-orientable filling surfaces lies between 5n/2 and 4n asymptotically, and equals 5n/2 for the Möbius strip.","lead":"New upper and lower bounds are proven for the diameter of modular flip-graphs of non-orientable filling surfaces, growing linearly between 5n/2 and 4n. For the unpunctured Möbius strip, the diameter is shown to grow exactly like 5n/2, pinning down a constant that was previously open in the non-orientable analogue of the orientable theory.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The unproved structural classification in Proposition 4.1 is load-bearing for the Möbius upper bound, but a mod-2 homology parity argument validates it.","rationale":"The reader's weakest-assumption analysis points to Proposition 4.1 as the main load-bearing unproved claim, and I agree. The central upper bound for the Möbius strip depends on it through Theorem 4.3 and Theorem 4.8. However, the claim is not merely plausible but provable by a standard topological argument: the boundary of every triangle is a contractible cycle, so counting non-separating arcs modulo 2 must give zero. This validates Proposition 4.1 and shows that the unproved status is an exposition gap rather than a mathematical flaw. The secondary concern about Theorem 2.4, the transposition of [8, Lemma 3.2] to non-orientable surfaces, is also worth noting, but the local star of a boundary vertex is a disk, so the orientable proof is expected to transfer without difficulty. Since I find no actual counterexample or internal inconsistency, and the missing justifications are elementary, the reader's ACCEPT verdict should stand unchanged.","tokens_in":23048,"tokens_out":30772,"duration_ms":306976,"concrete_test":"Supply a direct proof of Proposition 4.1 via mod-2 homology: for each triangle t, take its boundary walk W_t; because t is a disk, [W_t] = 0 in H_1(M_n; Z/2). Each non-separating arc occurrence contributes the nonzero class, so the number of such occurrences, counted with multiplicity (a non-separating loop serving as two edges counts twice), must be even. Verify that this parity condition reduces the possible triangle types to exactly (i)-(iii). As an independent check, exhaustively generate all triangulations of M_n for small n (say n <= 6) and confirm that every triangle satisfies (i)-(iii); an odd non-separating occurrence count in any legitimate triangulation would falsify Proposition 4.1 and break the Möbius upper bound.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The sharp upper bound diam(F(M_n)) <= floor(5n/2) rests on the central-triangle theorem (Theorem 4.3) and Theorem 4.8, both of which rely on Proposition 4.1, a structural classification of the arcs bounding a triangle of a triangulation of the Möbius strip M_n. Proposition 4.1 is stated without proof, and the reader's weakest-assumption analysis correctly identifies it as load-bearing. In particular, Lemma 4.5 uses part (iii) to conclude that the third side of a triangle with two non-separating sides is separating, and Theorem 4.8 uses part (ii) to handle the two-arc case. The classification is nevertheless correct: each triangle is a disk, so its boundary walk is null-homotopic and hence null-homologous in H_1(M_n; Z/2). Every non-separating arc represents the unique nonzero class in H_1(M_n; Z/2), so the number of non-separating arc-occurrences in the boundary walk, counted with multiplicity, must be even. This forces zero or two non-separating occurrences in the three-arc case, and in the two-arc case forces the non-separating loop to occur twice (as two edges) while the separating loop occurs once. The other unproved transposition, Theorem 2.4 ([8, Lemma 3.2]), is also load-bearing for the general 4n upper bound, but the star of a boundary vertex is a disk even in a non-orientable surface, so the orientable proof should carry over. Neither issue appears to be a real mathematical gap; both are missing justifications that can be supplied.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper extends the theory of modular flip-graphs of filling surfaces to non-orientable surfaces. A filling surface Σ_n is obtained from a non-orientable surface Σ by fixing all marked points except n marked points on a privileged boundary component; the modular flip-graph MF(Σ_n) is the quotient of the flip-graph of triangulations by the mapping class group fixing that boundary component pointwise. The main theorems assert: (1) for every non-orientable filling surface there is a constant c_Σ with diam(MF(Σ_n))/n converging to c_Σ, and 5/2 ≤ c_Σ ≤ 4, with improved upper bounds for one-holed surfaces of demigenus g; (2) for the unpunctured Möbius strip M, diam(F(M_n)) lies between floor(5n/2)-2 and floor(5n/2), so the lower bound of the general interval is sharp; and (3) the subgraph of simplicial triangulations satisfies analogous linear bounds. The proofs adapt the orientable constructions of Parlier–Pournin, introduce a new general lower bound via boundary-arc contractions, and prove an upper bound for the Möbius strip through a structural theorem on central triangles.","tokens_in":23368,"tokens_out":11397,"duration_ms":112559,"significance":"If the results hold, this is a substantial extension of the known asymptotics of flip-graph diameters from orientable to non-orientable filling surfaces. The exact asymptotic diam(F(M_n)) ~ 5n/2 for the Möbius strip is particularly valuable, as it identifies the first non-orientable surface for which the constant is determined and shows that the general lower bound is sharp. The paper is also methodologically strong: the lower-bound proof is a genuine induction using boundary-arc contractions and explicit pairs of triangulations, the central-triangle argument is a new structural tool, and the statements are precise and falsifiable. The proofs are mostly self-contained except for two explicitly flagged transpositions from earlier work. There are no fitted parameters or numerical computations; the arguments are combinatorial and checkable by hand.","major_comments":[{"comment":"Proposition 4.1 is stated without proof, yet it is load-bearing for the Möbius-strip upper bound. Lemma 4.5 uses part (iii) to identify the third side of a triangle with two non-separating sides as separating, and Theorem 4.8 uses part (ii) in the two-arc case. A proof should be supplied. The statement is correct: each triangle is a disk, hence its boundary walk is null-homologous in H_1(M_n; Z/2), and since every non-separating arc represents the unique nonzero class, the number of non-separating arc-occurrences in the boundary walk must be even. This short parity argument would close the gap, but as written the manuscript asks the reader to accept an unproved structural classification on which the main upper bound rests.","section":"Section 4, Proposition 4.1"},{"comment":"Theorem 2.4 is quoted from [8, Lemma 3.2] with the remark that the proof can be immediately transposed to non-orientable surfaces, but no proof or verification is included. This lemma is used to prove Theorem 2.5, which in turn gives the existence of the constant c_Σ and the general 4n upper bound in Theorem 1.1. Since the lemma is load-bearing, the manuscript should either prove it or explain in detail why the orientable proof carries over. The natural justification is that the star of a boundary vertex is a disk even in a non-orientable surface, so the local flip-counting argument from [8] should apply unchanged; nevertheless, this needs to be stated rather than left to the reader.","section":"Section 2, Theorem 2.4 and Theorem 2.5"},{"comment":"In the case where T+ contains a separating loop twice incident to u, the proof asserts that \"the non-separating loop contained in T+\" exists, but no justification is given. The presence of a separating loop based at u does not by itself rule out the coexistence of non-separating non-loop arcs incident to u elsewhere in the same triangulation. This case split is used to define the arc γ along which T+ is cut, so the proof of the upper bound is incomplete at this point. The authors should justify the asserted existence of a non-separating loop, or restructure the argument to avoid this assumption.","section":"Section 4, proof of Theorem 4.8"}],"minor_comments":[{"comment":"The sentence \"Combining Proposition 2.3 with Theorem 2.7 shows that for every non-orientable filling surface...\" appears to refer to Theorem 2.5, which is the result just proved; Theorem 2.7 is defined later. The same paragraph also says the first upper bound on c_Σ follows from Theorem 2.7, where Theorem 2.5 seems intended.","section":"Section 2, paragraph after Theorem 2.5"},{"comment":"There is a typo in the phrase \"it sufficees to lower bound\" in the final case; it should read \"it suffices\".","section":"Section 3, proof of Theorem 3.1"},{"comment":"The phrase \"two (non-necessarily distinct) points\" should be \"two (not necessarily distinct) points\", and in Lemma 3.6 the word \"canot\" should be \"cannot\".","section":"Section 4, paragraph before (21)"},{"comment":"The figure and its caption are dense, and the four cases would be easier to follow if the text explicitly named which of the four panels is being referenced in each sentence of the proof of Theorem 4.8.","section":"Section 4, Figure 13"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is a serious contribution and the main results appear correct, but two unproved structural ingredients are load-bearing: Proposition 4.1 and Theorem 2.4. Both are repairable with short arguments, and the third major comment on Theorem 4.8 may also be fixable by a local modification. I recommend major revision rather than rejection because the gaps are localized and the central claims remain plausible. The extensive reliance on [8] and [9] is appropriate because those are published, checkable results, but the manuscript should make the claimed transpositions explicit before acceptance."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick verdict: this is a genuinely useful paper. It does what the title says—extends the orientable bounds on modular flip-graph diameters to non-orientable filling surfaces, and then pins down the Möbius strip case exactly: diameter asymptotic to 5n/2, which matches the general lower bound. Theorem 1.1 and Theorem 1.2 are new results, not just restatements. The proof of Theorem 3.1 genuinely generalizes the boundary-contraction machinery from [9], and the Möbius upper bound via central triangles is real work. The paper even catches a constant error in [9], which is the kind of care that earns trust.\n\nSoft spots are real but minor. Proposition 4.1, the structural classification of arcs bounding a triangle in the Möbius strip, is stated without proof and is load-bearing for the upper bound: Lemma 4.5 and Theorem 4.8 both rely on it. The reader flagged this, and I agree. It is not an error—the mod-2 homology parity argument works. The boundary of a triangle is null-homologous, and the unique Z/2 homology class of the Möbius strip forces the number of non-separating arc-occurrences in the boundary walk to be even. That gives exactly the two-arc and three-arc cases claimed. The paper should supply that half-page proof. Theorem 2.4 is likewise quoted from [8] with the proof omitted; here the transposition is as straightforward as the authors say, because the star of a boundary vertex is a disk, so the orientable proof carries over. Still, an editor should ask to see it.\n\nCitations are honest: heavy use of [8] and [9] is justified because these are the prior results being extended, not window dressing. The exposition is careful, aside from a few typos (\"triangilations\", \"sufficees\"). Theorem 1.3 and Question 1.4 are reasonable side contributions.\n\nWho is this for? People who work on flip graphs, triangulations, and mapping-class-group geometry. It will not change the world outside that niche, but inside it, the exact asymptotic for the Möbius strip is a clean benchmark. I would send it to a competent referee. With the structural lemma added, it should be accepted.","headline":"Solid extension of the Parlier–Pournin flip-graph results to non-orientable surfaces, capped by an exact 5n/2 diameter for the Möbius strip; the few unproved structural facts are load-bearing but easy to fill.","tokens_in":23914,"tokens_out":3787,"would_cite":true,"duration_ms":35980,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["57K20","05C12"],"pacs":[],"model":"deepseek-v4-flash","headline":"The modular flip-graph diameter of every non-orientable filling surface grows linearly at a rate between 5/2 and 4, and the unpunctured Möbius strip attains the exact rate 5/2.","keywords":["flip graphs","filling surfaces","Möbius strip","triangulations of surfaces","mapping class groups","graph diameter","non-orientable surfaces","simplicial triangulations"],"falsifier":"For $n=5$ or $n=6$, enumerate all triangulations of the unpunctured Möbius strip and check each triangle against Proposition 4.1: any triangle with fewer than two bounding arcs, any two-arc triangle with both or neither of its loops non-separating, or any three-arc triangle with exactly one non-separating arc would refute the structural claim behind the upper bound. Alternatively, compute the exact flip distance between the paper's $A^-_n$ and $A^+_n$ by exhaustive search; a distance below $\\lfloor 5n/2 \\rfloor - 2$ would contradict Theorem 1.2.","tokens_in":22847,"feed_emoji":"🔀","tokens_out":12553,"duration_ms":110023,"temperature":0.7,"pith_summary":"This paper studies how many elementary moves, called flips, are needed in the worst case to turn one triangulation of a non-orientable surface into another, when the surface is built by placing $n$ marked points on a distinguished boundary component and keeping all other topology fixed. Its central claim is that this worst-case number, the diameter of the flip-graph (or of its quotient by boundary-fixing homeomorphisms), grows linearly in $n$ with a growth rate $c_\\Sigma$ lying between $5/2$ and $4$. The paper further proves that the lower endpoint is attained: for the unpunctured Möbius strip, the flip-graph diameter is exactly asymptotic to $5n/2$, satisfying $\\lfloor 5n/2 \\rfloor - 2 \\le \\operatorname{diam}(\\mathcal{F}(M_n)) \\le \\lfloor 5n/2 \\rfloor$ for every positive $n$. These results extend previously known orientable bounds to non-orientable surfaces and identify the simplest non-orientable surface as the extremal example. A third result bounds the diameter of the subgraph formed by simplicial triangulations of the Möbius strip between roughly $5n/2$ and $4n$ with additive constants.","feed_headline":"Möbius strip flip-graph diameter grows like 5n/2","feed_subtitle":"Same linear rate lies between 5/2 and 4 for every non-orientable filling surface, with 5/2 sharp.","key_machinery":"The lower bound is carried by two explicit families of triangulations, $A^-_n$ and $A^+_n$, built from a zigzag of boundary arcs together with fixed triangulations of a one-holed subsurface; every geodesic between them is forced to spend many flips on three distinguished boundary arcs, and lemmas on boundary-arc contraction convert those forced flips into recursive lower bounds that accumulate to $\\lfloor 5n/2 \\rfloor$. The upper bound on the Möbius strip is carried by the central triangle: a triangle bounded by at least one non-separating arc whose three bounding arc lengths sum to $n$. The paper proves that every triangulation of $M_n$ contains such a triangle, using a structural classification of triangles (each is bounded by at least two arcs, with a fixed parity pattern of non-separating arcs), and then routes any two triangulations through intermediate triangulations $C_u(v,w)$ that are all within distance about $n$ of the two endpoints.","core_discovery":"On the paper's own terms, the discovery is that the asymptotic-diameter theory of flip-graph moduli spaces for orientable surfaces carries over to non-orientable surfaces, and that the Möbius strip is the example that makes the general lower bound sharp. For any non-orientable filling surface $\\Sigma$, the limit $c_\\Sigma = \\lim_{n\\to\\infty} \\operatorname{diam}(\\mathcal{MF}(\\Sigma_n))/n$ exists and satisfies $5/2 \\le c_\\Sigma \\le 4$; for non-orientable one-holed surfaces of demigenus $g$ (the number of cross-caps inserted into a disk), the upper bound improves to $4 - 1/(2g)$ when $g\\ge 3$ and to $23/8$ when $g=2$. For the unpunctured Möbius strip $M$, whose mapping class group is trivial so that the modular flip-graph coincides with the ordinary flip-graph, the paper proves $\\lfloor 5n/2 \\rfloor - 2 \\le \\operatorname{diam}(\\mathcal{F}(M_n)) \\le \\lfloor 5n/2 \\rfloor$ for every $n$.","pith_inferences":["If the answer to the paper's Question 1.4 is that $c_\\Sigma$ is monotone nondecreasing in demigenus, then the Möbius strip's value $5/2$ would force every non-orientable one-holed surface to have $c_\\Sigma = 5/2$ only at demigenus 1; equality at higher demigenus would contradict monotonicity.","The two-family construction $A^-_n, A^+_n$ is more flexible than the earlier orientable-only versions, so tracking flips incident to a fourth or fifth boundary arc in the same recursion could in principle yield lower bounds strictly above $5/2$ for surfaces of higher demigenus.","Proposition 4.1 is asserted without proof; a direct enumeration of all triangulations of $M_5$ and $M_6$ would serve as a mechanical check of that classification before a full proof, and would also test whether the seven-flip bound to simplicial triangulations is tight.","One could ask whether the exact $5n/2$ behaviour transfers to a punctured Möbius strip; the paper's central-triangle machinery depends on the unpunctured structure, so the punctured case is a natural testbed for whether the constant $5/2$ is stable under adding interior punctures."],"forward_implications":["Because the limit $c_\\Sigma$ exists for every non-orientable filling surface, the asymptotic flip distance between triangulations is a well-defined topological invariant of the surface type.","The Möbius strip attains $c_\\Sigma = 5/2$, so the universal lower bound in Theorem 1.1 is sharp; no constant larger than $5/2$ can be valid uniformly across non-orientable filling surfaces.","For a non-orientable one-holed surface of demigenus $g\\ge 3$, the general upper bound $4$ improves to $4 - 1/(2g)$, and to $23/8$ for $g=2$, so the bound tightens as the non-orientable topology grows.","On the Möbius strip, the simplicial-triangulation flip graph has diameter at least $\\lfloor 5n/2\\rfloor - 16$ and at most $4n+K$, so restricting to simplicial triangulations does not change the linear order of worst-case flip distance.","Because the flip graph is a quasi-isometric model of the mapping class group, the linear bounds provide explicit constants for the quasi-isometry of non-orientable mapping class groups."],"supporting_citations":[{"why":"Supplies the orientable filling-surface flip-graph bounds, the boundary-arc contraction lemma, and the upper-bound lemma that the paper transposes to the non-orientable case.","marker":"[8]"},{"why":"Provides the one-holed lower-bound strategy via far-apart triangulations and the explicit paths behind the $4 - 1/(2g)$ and $23/8$ upper bounds.","marker":"[9]"},{"why":"Establishes strong convexity of flip-graph geodesics and the quasi-isometry with mapping class groups, which the paper extends in weakened form to arcs parallel to the privileged boundary.","marker":"[3]"},{"why":"Gives the rotation-distance bound for polygon triangulations and the lemma used to prove strong convexity for boundary-parallel arcs and to make vertices ears efficiently.","marker":"[15]"},{"why":"Introduces the simplicial flip-graph of Möbius-strip triangulations and proves its connectivity, providing the object studied in the paper's Theorem 1.3.","marker":"[4]"}],"fun_headline_variants":["Non-orientable flip-graph diameters: 5n/2 to 4n","Möbius strip flip-graph: sharp 5n/2 diameter","Flip-graph diameter: non-orientable surfaces tamed","From 5/2 to 4: non-orientable flip-graph growth","Exact 5n/2 diameter for Möbius strip triangulations"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premises are two structural rules asserted without proof: that every triangle in a Möbius-strip triangulation belongs to one of the three listed shape classes, and that a vertex-reduction lemma proven for orientable surfaces (Theorem 2.4, transposed from [8, Lemma 3.2]) still works without orientability. If either rule has an exception, the matching upper bound collapses.","fun_headline_variants_meta":{"raw":{"variants":["Non-orientable flip-graph diameters: 5n/2 to 4n","Möbius strip flip-graph: sharp 5n/2 diameter","Flip-graph diameter: non-orientable surfaces tamed","From 5/2 to 4: non-orientable flip-graph growth","Exact 5n/2 diameter for Möbius strip triangulations"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000543,"raw_usage":{"total_tokens":2642,"prompt_tokens":1032,"completion_tokens":1610,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":648,"completion_tokens_details":{"reasoning_tokens":1505}},"tokens_in":648,"tokens_out":1610,"duration_ms":10326,"temperature":1.0,"reasoning_tokens":1505,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T23:38:29.608242+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For $n=5$ or $n=6$, enumerate all triangulations of the unpunctured Möbius strip and check each triangle against Proposition 4.1: any triangle with fewer than two bounding arcs, any two-arc triangle with both or neither of its loops non-separating, or any three-arc triangle with exactly one non-separating arc would refute the structural claim behind the upper bound. Alternatively, compute the exact flip distance between the paper's $A^-_n$ and $A^+_n$ by exhaustive search; a distance below $\\lfloor 5n/2 \\rfloor - 2$ would contradict Theorem 1.2.","supporting_citations":[{"cited_title":"9, 2697–2737","cited_arxiv_id":null,"evidence_quote":"Supplies the orientable filling-surface flip-graph bounds, the boundary-arc contraction lemma, and the upper-bound lemma that the paper transposes to the non-orientable case."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the one-holed lower-bound strategy via far-apart triangulations and the explicit paths behind the $4 - 1/(2g)$ and $23/8$ upper bounds."},{"cited_title":"6, 3809– 3844","cited_arxiv_id":null,"evidence_quote":"Establishes strong convexity of flip-graph geodesics and the quasi-isometry with mapping class groups, which the paper extends in weakened form to arcs parallel to the privileged boundary."},{"cited_title":"3, 647–681","cited_arxiv_id":null,"evidence_quote":"Gives the rotation-distance bound for polygon triangulations and the lemma used to prove strong convexity for boundary-parallel arcs and to make vertices ears efficiently."},{"cited_title":"Edelman and Victor Reiner, Catalan triangulations of the M¨ obius band, Graphs and Combinatorics 13 (1997), no","cited_arxiv_id":null,"evidence_quote":"Introduces the simplicial flip-graph of Möbius-strip triangulations and proves its connectivity, providing the object studied in the paper's Theorem 1.3."}],"review_version":1}