{"id":"bf5ddaf8-3d61-4561-9a15-c7d79a4e6a00","arxiv_id":"2608.01393","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"An exact support-function inequality is derived that characterizes shortest escape paths from arbitrary triangular forests and dual triangle covers for Moser's worm problem.","lead":"This paper derives a single support-function inequality that characterizes shortest escape paths from triangular forests, and the dual triangle-cover problem for Moser's worm. If correct, it offers a general computational route to a long-standing open problem in discrete geometry for arbitrary triangle shapes.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 4's support-function equivalence is sound; the unsupported compactness step in Theorem 5 is the load-bearing gap.","rationale":"The reader's verdict and this stress-test agree that the paper is CONDITIONAL, but the weakest assumption identified by the reader is the external TSPN equivalence from [14,15]. In my reading, Theorem 4 does not actually depend on the TSPN reduction: the support-function derivation in Section 2.2 and the proof in Section 2.4 are self-contained once the distance-simplex surjectivity is granted, and that surjectivity is true (the Lean appendix sketches it, though the code is corrupted). The more load-bearing internal gap is Theorem 5's existence proof, which cites nonemptiness and boundedness but not the compactness needed for attainment. Since the convergence theorems 6 and 8 rely on the existence of an optimal continuous path, this is a genuine missing step. It is fixable by a standard Arzelà–Ascoli and lower-semicontinuity argument, so the verdict remains CONDITIONAL rather than REJECT. The numerical reproducibility concerns (no code, data, or certificates) are also legitimate but secondary to the mathematical convergence claim; the author explicitly defers certificates to future work, which supports keeping the paper conditional rather than fully accepted.","tokens_in":28676,"tokens_out":35786,"duration_ms":437352,"concrete_test":"Write out the missing compactness lemma: for any B, the set of feasible paths with length at most B, parameterized by arclength on [0,1], is equicontinuous; show the feasible set is closed under uniform convergence (the support constraints are continuous in the uniform norm) and that J is lower semicontinuous. If the lemma is supplied and verified, Theorem 5 holds and Theorems 6 and 8 follow; if any step fails, the minimizer statement must be weakened to an infimum-only statement.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The paper's central equivalence, Theorem 4, is internally consistent: the weighted support inequality is exactly equivalent to robust escape for all starting positions and orientations, and the distance-simplex surjectivity used in the converse is true (though only asserted, not proved, in the main text). The vulnerable point is Theorem 5, whose proof asserts existence of an optimal continuous path but supplies no compactness argument. The proof only cites Finch's diameter bound to show that the feasible set is nonempty and bounded in length; it then says 'the proofs provide the compactness and liminf inequality' without actually proving compactness of the feasible set or lower semicontinuity of the length functional. Without an explicit Arzelà–Ascoli step (e.g., arclength reparameterization to obtain equicontinuity, closedness of the support constraints under uniform convergence, and lower semicontinuity of total variation), the minimum v may be an infimum rather than an attained optimum. This gap propagates to Theorem 6 and Theorem 8, where convergence statements are framed as convergence to an optimal continuous path. The gap is likely repairable by standard techniques, but as written the existence and convergence proofs are incomplete. Additional missing support includes the unexplained bound formula Eq (22) and the absence of certificates for the numerical solutions, but the compactness omission is the most direct threat to the convergence claims.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a support-function formulation for Bellman's lost-in-a-forest problem in a triangular forest and for the dual Moser worm problem of covering unit arcs by a triangle. The path is kept fixed at the origin while the triangle translates and rotates; the paper derives a scalar weighted support inequality (Theorem 4) claimed to be exactly equivalent to robust escape for all starting positions and orientations. It further claims existence of optimal continuous escape paths (Theorem 5), convergence of polygonal and collocation discretizations (Theorems 6 and 8), and an extension to arbitrary convex polygons (Section 2.5). Numerical results for triangles with various base angles are reported in Figures 2-4, together with a formula for a Moser-worm upper bound in Eq (22). A Lean formalization appendix is included.","tokens_in":28881,"tokens_out":11443,"duration_ms":150337,"significance":"If Theorem 4 is correct, it is a valuable reduction: the infinite family of escape constraints for all translations and rotations is collapsed into one scalar inequality per orientation, enabling a computational approach to arbitrary triangular forests and triangles as worm covers. The paper also reports what appear to be the first numerical escape paths for non-isosceles triangles, and it attempts to support the main theorems with machine-checked Lean proofs, which is commendable. However, the existence proof for the continuous optimum is incomplete, the numerical formulations do not exactly match the theorem's support function, and the worm upper-bound formula is unproved. These gaps currently prevent full confidence in the paper's central claims, although the algebraic core of Theorem 4 appears sound and repairable.","major_comments":[{"comment":"The proof of Theorem 5 does not establish existence of a minimizer. Non-emptiness of the feasible set and boundedness of length via Finch's diameter bound do not imply that the infimum is attained; an explicit compactness argument is needed. The statement 'the proofs provide the compactness and liminf inequality' refers to an argument that is not supplied in the main text, and the appendix's abstract lemma compact_subsequence_is_optimal only lists hypotheses without verifying that the feasible set of escape paths is compact or that the length functional is lower semicontinuous on that set. Since Theorems 6 and 8 formulate convergence to an optimal continuous path, this gap is load-bearing and needs a concrete Arzelà-Ascoli step, including arclength reparameterization, closedness of the support constraints under uniform convergence, and lower semicontinuity of total variation.","section":"§2.4, Theorem 5"},{"comment":"The support function used in the numerical formulations is not the same as the support function in Theorem 4. Equation (5) defines h over r([0,2π]), while Theorem 4 defines h over eΓr, which includes the initial segment from the origin to r(0). The discretized problems in Eq (9) and Eq (12) take maxima over the discretized vertices only, do not explicitly include the origin as a point in the max, and Eq (12) restricts the max to 1≤i≤K even though Theorem 6 defines hK with q0=0. Unless r(0)=0 is intended, the solved optimization problems have a different feasible set and objective from the characterized problem, so the numerical lengths in Figures 2-4 are not certified to be escape-path lengths for the original problem. This inconsistency must be resolved, for example by defining the discrete support as max(0, max_i ...) and including the origin in the supporting set.","section":"§2.2, Eqs (5), (8), (9); §2.4, Eq (12)"},{"comment":"The worm upper-bound formula 1/2 L^2 (1/tan α + 1/tan β) is introduced without proof or derivation. It is not a standard quoted result in the references, and it is the entire basis for the numerical Moser-worm cover areas in Figure 3. The paper should either prove the formula from the cited forest-worm duality or clearly label the worm areas as heuristic estimates rather than established upper bounds.","section":"§3.1, Eq (22)"},{"comment":"The Lean code in the appendix is not readable as supplied: identifiers, operators, and binders are replaced by the placeholder glyph '￿' throughout, including in the statements of traceEscapes3_iff_support and robustEscape3_iff_weightedSupport. Consequently the claimed machine-checked proofs cannot be verified from the manuscript. Please provide a clean, compilable version of the Lean code, especially because the appendix is invoked to fill the compactness gap in Theorem 5.","section":"Appendix"}],"minor_comments":[{"comment":"There are several typos and infelicities: 'apths' should be 'paths', 'efficiently' should be 'efficiently', and the phrase 'the topology of r(p) is strictly defined as open curve' is unclear.","section":"§2.2, p.6"},{"comment":"The claimed extension to arbitrary convex polygons is stated without proof. The phrase 'proof ... is very similar' is not sufficient, because the weighted-simplex surjectivity and the existence of the weights λj require assumptions that are not stated for general m-gons.","section":"§2.5, Eqs (20)-(21)"},{"comment":"The numerical results are reported with no data tables, no certificates of global optimality, and no verification that the reported solutions satisfy the continuum support constraints. The paper itself acknowledges that 'numerical global optimality requires a certified global solver', but Figures 3 and 4 are nevertheless presented as quantitative results; this should be made conditional.","section":"§3.1"},{"comment":"Reference [16] has a duplicated year '(2026). (2026).', and some arXiv identifiers in the references are inconsistently formatted.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The manuscript relies heavily on the author's earlier preprints [14] and [15] for the TSPN reduction, so the current paper is not self-contained; the editor may wish to verify that those preprints are available and that the equivalence invoked from them is indeed established. The numerical section is extensive but lacks reproducibility data and certificates, which may be a concern for the journal's standards. The appendix's Lean code is corrupted in the submitted text, so the claimed formal verification cannot currently be assessed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: the weighted support inequality in Theorem 4 is the real contribution, and I think it holds. The paper extends Temerev's isosceles constraint to arbitrary base angles, gives a clean derivation from the edge-length equilibrium identity and the distance-simplex surjectivity, and does it with no fitted parameters. The polygonal-chain and closed-curve variants are natural but useful extensions. That is worth a serious referee.\n\nI largely agree with the reader's conditional verdict and with the stress-test note, with one emphasis: the load-bearing gap is not Theorem 4's equivalence but Theorem 5's existence claim. The proof cites Finch's diameter bound to show the feasible set is nonempty and bounded, then asserts compactness and lower semicontinuity without proving either. An explicit Arzelà–Ascoli step, or a reformulation that only needs infima and approximate minimizers, is required. The gap is repairable with standard techniques, but it propagates to Theorems 6 and 8 as written.\n\nTwo softer issues. First, the TSPN path-fixing reduction from the author's earlier papers is invoked without re-proof; if that equivalence is flawed, the support inequality characterizes a different problem. Second, the numerical section has no code, data, or certificates, which the author explicitly defers. Equation (22), the worm-area bound, appears without derivation or citation; it needs support.\n\nThe Lean appendix deserves mention. It claims formal proofs of Theorems 4–8, but no repository or commit hash is provided and the pasted code looks like it has encoding artifacts. I would not count the formalization as verifiable as-is. If the author posts clean files, that would materially raise my confidence.\n\nWho this is for: people working on Bellman's lost-in-a-forest, Moser's worm, or support-function discretizations of geometric optimization. They will get a usable formulation and useful numerical maps, but should treat the convergence theorems as sketched rather than proved. My recommendation: send it to peer review. A good referee can check the equivalence, ask for the compactness argument to be written out, and request code and certificates. The paper deserves that engagement despite the current gaps.","headline":"The paper's support-function equivalence for arbitrary triangles is a genuine advance and probably correct; the existence and convergence proofs are the soft spot, not the central inequality.","tokens_in":29403,"tokens_out":2651,"would_cite":true,"duration_ms":35176,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["49K30","49Q10","52A40"],"pacs":[],"model":"deepseek-v4-flash","headline":"A path escapes every translated, rotated triangle exactly when one weighted support-function inequality holds for every phase angle.","keywords":["Bellman's lost-in-a-forest problem","Moser's worm problem","support function","arbitrary triangle","escape path","polygonal chain","universal cover","functional minimization"],"falsifier":"Choose a non-isosceles triangle, say base angles $20^\\circ$ and $50^\\circ$, and compute a path that satisfies the inequality at every phase. Then test the original problem directly: sample starting points densely inside the triangle and orientations $\\theta\\in[0,2\\pi)$, and check whether the path intersects the translated and rotated triangle boundary in every case. A single sampled pair for which the path stays strictly inside the triangle while the inequality holds would refute Theorem 4, since the theorem asserts that no such pair can exist.","tokens_in":28425,"feed_emoji":"📐","tokens_out":9535,"duration_ms":104396,"temperature":0.7,"pith_summary":"The paper tries to establish that the shortest path which guarantees escape from a triangular forest of any shape is characterized exactly by one family of scalar inequalities built from the path's support function, and that the dual worm problem of covering every unit curve by a triangle obeys the same condition. Keeping the escape curve fixed at the origin while the triangle translates and rotates compresses the unknown starting position and orientation into three phase-shifted support values. Theorem 4 states that a continuous path starting at the origin escapes for every starting point and every orientation if and only if $\\sin\\beta\\,h(t+\\pi+\\alpha)+\\sin\\alpha\\,h(t+\\pi-\\beta)+\\sin(\\alpha+\\beta)\\,h(t) \\ge \\sin\\alpha\\sin\\beta$ for every phase $t$, where $h$ is the support function of the path. The paper then turns this into a constrained variational problem, proves existence of an optimal escape path and convergence of polygonal and discretized approximations, and reports numerical solutions for arbitrary non-isosceles triangles, which earlier literature did not provide.","feed_headline":"One inequality decides escape from any triangle forest","feed_subtitle":"Support-function test compresses every starting position and orientation into one scalar condition.","key_machinery":"The support function of the complete path, $h(\\varphi)=\\max_p r(p)\\cdot(\\cos\\varphi,\\sin\\varphi)$, taken over the curve together with the segment from the origin to its start so that $h(\\varphi)\\ge 0$. The argument is carried by the weighted slack identity: with side lengths $L_1,L_2,L_3$ and distances $d_1,d_2,d_3$ from the origin to the three supporting lines of a translated triangle, the weighted sum $L_1d_1+L_2d_2+L_3d_3$ equals twice the triangle's area for every starting point. This identity, together with the phase-shifted normal directions, converts the requirement that the path reach at least one of the three lines for every translation and rotation into a single scalar inequality involving $h$ at the three phases. The intermediate value theorem supplies the transition from inequality to actual boundary crossing because the path is continuous and starts at the origin, which lies inside the translated triangle.","core_discovery":"The central claim is that escape from a triangle is not a family of geometric impossibilities spread over continuously many positions and orientations; it is one weighted inequality per phase. For the normalized triangle with unit base and base angles $\\alpha,\\beta$, the three sides have outward normals whose rotation phases are $t+\\pi+\\alpha$, $t+\\pi-\\beta$, and $t$, with side lengths $\\sin\\beta/\\sin(\\alpha+\\beta)$, $\\sin\\alpha/\\sin(\\alpha+\\beta)$, and $1$. The equilibrium identity $L_1 n_1+L_2 n_2+L_3 n_3=0$ forces the weighted sum of the three supporting-line offsets to equal the constant $\\sin\\alpha\\sin\\beta/\\sin(\\alpha+\\beta)$, twice the triangle's area. Hence a path whose weighted support sum reaches that constant must, by continuity and the intermediate value theorem, cross at least one boundary line of every translated and rotated triangle; conversely, if the inequality fails, the proof constructs a translated triangle whose interior contains the whole path, so escape fails. This exact characterization is the load-bearing result, and the existence, polygonal convergence, and discretized convergence theorems all hang on it.","pith_inferences":["Because the underlying equivalence with Bellman's problem is inherited from earlier papers and not re-proved here, the strongest test of the paper's contribution is a direct numerical or formal check of Theorem 4 for a few non-isosceles triangles against the original grid formulation.","The polygon formulas in Section 2.5 are stated with a proof described only as very similar; convergence for arbitrary polygons is therefore a plausible extension rather than an established result of this paper.","The same phase-shifted support certificate could in principle certify global optimality for the worm problem at every angle, not just the isosceles and 30-60-90 cases singled out in the numerical band; the paper hints at this but does not compute the certificates.","A natural next step would be to run the same support-function optimization at high precision for individual angle pairs, producing bounds that tighten the known universal-cover upper bound."],"forward_implications":["For any triangle, the shortest escape path can be written as a minimum of arc length under a scalar support constraint, eliminating the permutation variables of the earlier TSPN formulation.","The same formulation covers closed curves and closed polygonal chains by simply adding the closing segment to the objective.","Solving the discretized problem yields escape paths for arbitrary non-isosceles triangles; the paper states these are the first such results, while matching known isosceles cases.","The finite polygonal and angular-collocation problems converge to the continuous optimum as the segment count and grid resolution grow, with convergence proved in Theorems 6 and 8.","Through the forest-worm duality, the same inequality yields an upper bound on the area of a triangle covering all unit arcs, namely $\\tfrac{1}{2}L^2(\\cot\\alpha+\\cot\\beta)$ for optimal escape length $L$.","The optimizer can be extended to arbitrary convex polygons by replacing the three side normals with a weighted combination of all polygon normals, yielding formulas for universal polygon covering curves and polygonal chains."],"supporting_citations":[{"why":"The earlier paper establishes the path-fixed translated-triangle reformulation that is the starting point of Section 2.2.","marker":"[14]"},{"why":"The earlier paper supplies the TSPN equivalence, the compactness and arc-length reparameterization argument, and the convergence framework invoked in the proofs of Theorems 4 through 8.","marker":"[15]"},{"why":"Finch's survey supplies the diameter escape path used to prove existence of an optimal continuous path and the forest-worm duality used for the cover-area upper bound.","marker":"[1]"},{"why":"The unit-strip paper is the earlier application of the same support-function technique, motivating the constraint form used here.","marker":"[16]"},{"why":"Temerev's independent derivation of the same isosceles constraint supports Corollary 1's claim that the formula reproduces recent results.","marker":"[11]"},{"why":"Gibbs's isosceles escape paths are the baseline the numerical results are compared against, supporting the claim of agreement for isosceles triangles.","marker":"[12]"}],"fun_headline_variants":["Escape any triangle: one inequality rules them all","Single inequality decides escape from any triangle","Triangle escape solved by one scalar test","One weighted sum determines triangular forest escape","Support-function test: one inequality for all triangles"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"Everything in the paper depends on the equivalence, proved in the author's earlier papers and invoked here, between Bellman's original escape problem and the reformulation in which the path is fixed at the origin while the triangle translates and rotates; if that equivalence has a flaw, the support inequality describes a different condition from the original problem.","fun_headline_variants_meta":{"raw":{"variants":["Escape any triangle: one inequality rules them all","Single inequality decides escape from any triangle","Triangle escape solved by one scalar test","One weighted sum determines triangular forest escape","Support-function test: one inequality for all triangles"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000196,"raw_usage":{"total_tokens":1339,"prompt_tokens":900,"completion_tokens":439,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":516,"completion_tokens_details":{"reasoning_tokens":373}},"tokens_in":516,"tokens_out":439,"duration_ms":5976,"temperature":1.0,"reasoning_tokens":373,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T00:56:40.716398+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Choose a non-isosceles triangle, say base angles $20^\\circ$ and $50^\\circ$, and compute a path that satisfies the inequality at every phase. Then test the original problem directly: sample starting points densely inside the triangle and orientations $\\theta\\in[0,2\\pi)$, and check whether the path intersects the translated and rotated triangle boundary in every case. A single sampled pair for which the path stays strictly inside the triangle while the inequality holds would refute Theorem 4, since the theorem asserts that no such pair can exist.","supporting_citations":[{"cited_title":"R., & Wetzel, J","cited_arxiv_id":null,"evidence_quote":"Finch's survey supplies the diameter escape path used to prove existence of an optimal continuous path and the forest-worm duality used for the cover-area upper bound."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gibbs's isosceles escape paths are the baseline the numerical results are compared against, supporting the claim of agreement for isosceles triangles."}],"review_version":2}