{"id":"1cafc153-6187-4d59-bb62-5169ca268e2d","arxiv_id":"1908.07437","paper_version":4,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Geometric signatures, fixed by orientation and a gauge ray direction, completely characterize totally non-negative edge-signature systems on plabic networks in the disk.","lead":"The paper defines geometric signatures on planar bicolored directed networks in the disk and proves these are exactly the edge signatures that make Lam's system of relations produce a totally non-negative positroid cell for all positive weights. It also gives rational formulas for edge vectors at internal edges, extending Postnikov and Talaska boundary results, with explicit transformation rules under moves and gauge changes.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The PBDTP restriction is load-bearing: Theorem 7.16's iff characterization covers only graphs where every edge lies on a boundary-to-boundary path, and Remark 7.15 concedes failure outside it, yet no explicit counterexample is given to prove necessity.","rationale":"The reader's weakest assumption correctly identifies the PBDTP condition as the load-bearing scope restriction of the central completeness theorem. My read confirms that Theorem 7.16's only-if direction relies on Lemma 7.14, and that lemma's construction of a gauge transformation requires every edge to lie on a boundary-to-boundary path. The authors explicitly flag the failure mode in Remark 7.15, so the restriction is not hidden; however, no explicit counterexample is supplied, leaving open whether the PBDTP condition is truly necessary or merely an artifact of the proof. This is the most significant concern about the central claim because the paper presents itself as a complete solution to Lam's question, and the completeness is precisely what is restricted. I do not find an internal inconsistency in the main proof: the weight-scaling arguments in Theorem 7.16 are sound, the geometric-signature construction is explicit, and the equivalence and invariance results are plausible. The separate abstract claim about dimer partition functions is unsupported in the body, but it is not load-bearing for Theorem 7.16. Since the reader's conditional verdict already accounts for the PBDTP limitation and the missing dimer support, my review does not move the verdict; it endorses the conditionality and suggests a concrete test to determine whether the PBDTP restriction is genuine.","tokens_in":65434,"tokens_out":20395,"duration_ms":197057,"concrete_test":"Take the smallest non-PBDTP graph: a path from a boundary source to a boundary sink with a directed 2-cycle attached 'sideways' so the cycle edges are not contained in any boundary-to-boundary path (for example, attach the cycle to the path at a single vertex and also to a boundary sink only). Enumerate all 2^m signatures on the cycle and path edges, and for each signature compute the Lam system determinant and the boundary matrix as rational functions of positive weights. If two non-equivalent signatures both give identically non-zero determinant and a totally non-negative boundary matrix for all positive weights, PBDTP is necessary for the iff; if not, the 'optimal setting' claim in the introduction is not established.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central iff claim (Theorem 7.16) is stated only for PBDTP graphs, and the proof of the 'only if' direction passes through Lemma 7.14, whose equality-of-parities implies gauge-equivalence step uses PBDTP essentially: an edge not lying on any boundary-to-boundary path never enters the boundary measurement, so signatures on it are unconstrained by the total non-negativity hypothesis. The authors acknowledge this in Remark 7.15, but they provide no concrete family of non-PBDTP graphs in which a non-geometric signature actually satisfies full rank plus total non-negativity for all positive weights. Consequently, the paper's advertised 'complete' answer to Lam's question has an unmeasured boundary: it is complete only on the PBDTP subclass, and it is not demonstrated that this subclass is the correct or optimal domain. Separately, the abstract asserts that boundary-measurement images and dimer partition functions do not coincide for non-bipartite graphs; no proof or location in the body is given, so this claim should be removed or supported.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a theory of edge vectors on planar bicolored directed trivalent perfect (plabic) networks in the disk that parametrize totally non-negative positroid cells. The central construction assigns to each edge a vector whose components are signed sums over directed paths from that edge to boundary sinks, with signs determined by a winding index relative to a gauge ray direction and by intersections with gauge rays. The authors prove that the resulting vertex relations have full rank (Theorem 3.11), that the edge-vector components are rational functions of the weights with subtraction-free denominators and explicit flow expansions (Theorem 3.12), and that the edge vectors transform predictably under changes of orientation, gauge ray direction, vertex and weight gauge, and Postnikov moves. In Section 7 the vertex relations are recast as Lam's half-edge relations, and geometric signatures are introduced. The main completeness result, Theorem 7.16, states that for a PBDTP graph representing an irreducible positroid cell, a signature gives a full-rank system with totally non-negative image for every choice of positive weights if and only if it is equivalent to the geometric signature, in which case the boundary solution is the Postnikov boundary measurement matrix and the image is the expected positroid cell. The paper also gives a face formula for geometric signatures and an interpretation of the relations as totally non-negative amalgamation of small Grassmannians.","tokens_in":65600,"tokens_out":10683,"duration_ms":126595,"significance":"If Theorem 7.16 is correct, it answers Lam's signature question for the PBDTP subclass of plabic networks and gives a concrete bridge among geometric relations, boundary measurement matrices, and amalgamation of totally non-negative Grassmannians. The paper's strengths are its explicit and constructive character: the full-rank statement in Theorem 3.11 is proved by identifying the determinant with the sum of conservative-flow weights, the edge-vector components are given by closed Talaska-type flow formulas, and the invariance and vertex-consistency checks are carried out in the appendices. The main caveat is that the advertised completeness is proved only under the PBDTP hypothesis, and the relation between this hypothesis and the stated claim of completeness needs to be made precise in the published version.","major_comments":[{"comment":"The completeness theorem is proved only for PBDTP graphs, and the PBDTP condition is load-bearing in Lemma 7.14: the step in which equal parities on all boundary-to-boundary paths and cycles are converted into gauge equivalence of signatures uses the fact that every edge lies on such a path. Remark 7.15 explicitly concedes that outside this class there is extra gauge freedom and that the statement of Theorem 7.16 must be modified. The abstract and the introduction nevertheless advertise a 'complete and explicit characterization' and describe the setting as 'optimal.' Since the paper gives no concrete non-PBDTP graph for which a non-geometric signature still has full rank and totally non-negative image for all positive weights, it does not establish that PBDTP is the optimal domain. The authors should either provide such an example or an argument that none exists, or rephrase the advertised completeness claim as a characterization on the PBDTP subclass.","section":"Definition 2.5; Lemma 7.14; Remark 7.15; Theorem 7.16"},{"comment":"The abstract states that 'the image of the boundary measurement map and the dimer partition functions do not coincide if the graph is not bipartite.' I could not find this assertion stated or proved anywhere in the body. Section 7.3 only conjectures a Kasteleyn-type interpretation for PBDTP graphs and notes that for reduced bipartite graphs the geometric signature realizes Speyer's variant; no non-bipartite comparison is proved. This claim should be removed from the abstract or supported by a proof in the text.","section":"Abstract; Section 7.3 (Remark 7.24, Conjecture 7.25)"},{"comment":"The asymptotic sign argument in Step 3 of the proof of Theorem 7.16 draws conclusions about the sign of the matrix entry A_{ij} from total non-negativity. Total non-negativity is a condition on maximal minors, not directly on individual entries of the reduced row echelon form. The argument should explicitly use the fixed-sign relation between A_{ij} and the maximal minor obtained by replacing the pivot column i_r with column j, or else state the sign convention that makes this implication immediate. This is a local proof gap rather than an error in the statement, but it is part of the proof of the central theorem and should be repaired.","section":"Theorem 7.16, Step 3 (Eqs. (7.18)–(7.19))"}],"minor_comments":[{"comment":"The phrase 'complete and explicit characterization' appears before the PBDTP hypothesis is introduced; state the hypothesis at the first mention of completeness so that the scope of the theorem is visible to the reader.","section":"Abstract and Section 1"},{"comment":"The word 'optimal' in the description of the PBDTP setting is stronger than what is proved. Replacing it with 'the setting in which we prove completeness' would better match the content of Theorem 7.16 and Remark 7.15.","section":"Section 1"},{"comment":"Example 3.15 gives a clear demonstration that null edge vectors can occur on reducible networks. It would help the reader if the example were referenced again at the beginning of Section 6, where the phenomenon is discussed systematically.","section":"Example 3.15 and Section 6"},{"comment":"The notation int(e) in Eq. (3.5) refers to the edge incident to a boundary sink, but the same symbol is used earlier for the intersection number of an arbitrary edge in a path. Clarify that the boundary edge has its own intersection number, since the distinction is used in several later formulas.","section":"Definition 3.4 and Eq. (3.5)"}],"recommendation":"major_revision","confidential_remarks":"The main derivation appears sound and the paper is candid about the PBDTP restriction in Remark 7.15. My recommendation of major revision is driven by the gap between the abstract's 'complete' claim and the actual PBDTP-scoped theorem, and by the unsupported dimer claim in the abstract. I do not see circularity or a fatal error in the core geometric-signature construction."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my read. The main result is Theorem 7.16: on a PBDTP plabic graph representing an irreducible positroid cell, a signature gives full rank and a totally non-negative image for every positive weight iff it is gauge-equivalent to the geometric signature; then the boundary solution is Postnikov's boundary measurement matrix. That is a real answer to Lam's question, on the class where every edge lies on a boundary-to-boundary path. The proof is structurally sound. They reduce to the geometric signature by flipping signs of weights, then use a formal Talaska-style expansion to force parity identities along paths and cycles, and Lemma 7.14 upgrades parity to gauge equivalence. The appendices do the heavy sign bookkeeping, and it looks consistent. I did not machine-check the sign calculations, but the architecture is convincing.\n\nWhat is genuinely new are the internal edge vector formulas (Theorem 3.12), the transformation rules under moves, orientations, and gauges, and the completeness result itself. The paper extends Postnikov/Talaska from boundary sources to internal edges with subtraction-free rational denominators. The comparison with the concurrent vector-relation configuration in [7] is explicit and useful.\n\nSoft spots, in proportion. First, the PBDTP condition is load-bearing, and the authors say so in Remark 7.15. But they don't provide a concrete non-PBDTP example where a non-geometric signature actually satisfies the hypotheses, so the reader can't see exactly how the iff fails. That doesn't make the theorem wrong, but it does mean the 'complete answer' is only complete on a subclass. The abstract and intro could be clearer on this. Second, the abstract claims boundary measurement images and dimer partition functions do not coincide for non-bipartite graphs. I couldn't find a proof in the body; there are remarks connecting to Speyer/Kasteleyn and a conjecture, but not a theorem. That claim should be removed, proved, or attributed to the companion paper [1]. It is a side remark, so minor.\n\nThe citation pattern is fine: Postnikov, Talaska, and Lam are the right anchors, and the distinction from [7] is handled honestly.\n\nWho is this for: people working on totally non-negative Grassmannians, plabic networks, cluster algebras, and KP solitons. It is a technical but careful paper. I would send it to a serious referee, and I would be comfortable with accept-with-revisions. My own verdict is conditional accept, with the abstract fixed and the PBDTP boundary better advertised.","headline":"A solid, self-aware completeness theorem for geometric signatures on PBDTP plabic graphs; the main proof holds, but the advertised 'complete' answer is only complete on that subclass and one abstract claim lacks support.","tokens_in":66148,"tokens_out":3111,"would_cite":true,"duration_ms":34279,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14M15","05C10","05C22"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that, on boundary-to-boundary plabic graphs in the disk, a Lam system of relations is full rank with totally non-negative image for all positive weights exactly when its edge signature is geometric, and then its boundary…","keywords":["totally non-negative Grassmannians","positroid cells","plabic networks","edge signatures","geometric signature","boundary measurement map","amalgamation","edge vectors"],"falsifier":"Test the sign obstruction in equation (7.16) on a small PBDTP graph: pick a signature that differs from the geometric one in parity on some boundary-to-boundary path or some conservative cycle, and choose positive weights concentrated on that path or cycle with all other weights of order $\\delta \\ll 1$. The theorem predicts that some boundary matrix entry, or its denominator, changes sign and breaks total non-negativity for sufficiently small $\\delta$; finding a non-geometric signature whose matrix stays totally non-negative for all positive weights on such a graph would disprove Theorem 7.16.","tokens_in":65202,"feed_emoji":"🕸️","tokens_out":7937,"duration_ms":79778,"temperature":0.7,"pith_summary":"Lam asked which edge signatures make a system of relations on a plabic graph full rank with a totally non-negative image for every choice of positive weights. The paper answers this for planar bicolored trivalent directed perfect graphs in the disk that are PBDTP, meaning every edge lies on a directed path from the boundary to the boundary. On such a graph the admissible signatures are exactly the geometric signatures, one per graph up to a simple gauge transformation, and the solution at boundary sources is Postnikov's boundary measurement matrix, with image the corresponding positroid cell. This matters because those relation systems are the algebraic transcription of amalgamation, the gluing of small positive Grassmannians that underlies cluster theory and scattering-amplitude computations.","feed_headline":"Only geometric edge signatures keep plabic networks non-negative","feed_subtitle":"Lam's relation systems work exactly for geometric signatures, reproducing Postnikov's boundary measurements.","key_machinery":"The central object is the geometric signature: a bit $\\epsilon_{U,V}$ assigned to each edge of the oriented network, computed from two geometric indices — the local winding number of an ordered pair of edges around a fixed gauge ray direction $l$, and the number of intersections of the edge with gauge rays emitted from the boundary sources. The Lam system writes these bits into relations $z_{U,e}=(-1)^{\\epsilon_{U,V}}w_{U,V}z_{V,e}$ at edges, equality at black vertices, and a zero sum at white vertices. The signature carries the whole argument because Theorem 7.16 says it is the only choice, up to gauge equivalence, that guarantees full rank and total non-negativity for all positive weights; its face-level counterpart (Theorem 7.23) says the total signature around any face is just the parity of the number of white vertices on it.","core_discovery":"The central claim, Theorem 7.16, is a complete characterization. Let $G$ be a PBDTP plabic graph representing an irreducible positroid cell $S^{TNN}_M$. A signature $\\epsilon_{U,V}$ on the edges induces a Lam system of relations that has full rank and a totally non-negative image for every choice of positive weights if and only if $\\epsilon_{U,V}$ is equivalent to the geometric signature; in that case the boundary-source solution is exactly the Postnikov boundary measurement matrix and the image is $S^{TNN}_M$. The geometric signature is unique up to gauge equivalence, and the paper shows that changes of orientation, gauge-ray direction, and vertex position act as gauge transformations of the signature. The paper also gives rational formulas for the edge-vector components at internal edges, extending Talaska's flow formula to the interior of the graph.","pith_inferences":["If Theorem 7.16 is right, it gives a practical criterion: to test whether a proposed relation system on a disk plabic network is admissible, one only needs to compare path and cycle parities with the geometric signature, avoiding direct checks over all positive weights.","The PBDTP condition may be the right generality for a slightly weaker statement: without it, signatures are still constrained on boundary-to-boundary paths and cycles, but the leftover edge gauge degrees of freedom likely form a finite-dimensional family; classifying that family would extend the theorem to arbitrary plabic graphs.","The face-signature formula suggests a direct bridge to dimer models: one could try to realize the geometric signature explicitly as a Kasteleyn sign matrix on any PBDTP graph, not only on reduced bipartite graphs, and use it to count perfect matchings with boundary conditions.","A natural testable extension, mentioned as an open problem in the paper, is the same construction on networks in the annulus or on other surfaces with boundary; the gauge-ray machinery should adapt with a modified Talaska formula, but the signature classification may pick up new topological invariants."],"forward_implications":["Lam's program for plabic networks in the disk is settled: the signatures that preserve total non-negativity under amalgamation of $\\mathrm{Gr}^{TP}(1,3)$ and $\\mathrm{Gr}^{TP}(2,3)$ are precisely the geometric ones.","The boundary measurement matrix is recovered as the unique solution of the geometric linear system at boundary sources, so internal edge vectors give a consistent extension of Postnikov's and Talaska's boundary formulas into the graph.","On acyclically orientable graphs, all edge-vector components are subtraction-free rational functions of the positive weights, and null edge vectors cannot occur.","For graphs that are not PBDTP, the completeness statement must be modified; the extra gauge freedom at edges not lying on boundary-to-boundary paths is untrivial and can even support null edge vectors on reducible networks.","The face-level signature formula connects the geometric signature to Kasteleyn and dimer sign conditions, and the paper identifies its master signature on Le-networks as geometric."],"supporting_citations":[{"why":"Supplies the plabic networks, boundary measurement map, moves and reductions, and Le-graph parametrization of positroid cells that the whole construction starts from.","marker":"[48]"},{"why":"Provides the flow and conservative-flow formula that Theorem 3.12 extends from boundary sources to internal edges.","marker":"[57]"},{"why":"Formulates the relation-system and edge-signature problem on bipartite graphs that Theorem 7.16 resolves.","marker":"[41]"},{"why":"Introduces amalgamation of cluster varieties, the operation whose totally non-negative specialization these relation systems represent.","marker":"[18]"},{"why":"Supplies the gauge-ray method for measuring winding and counting boundary-source crossings used to define geometric signatures.","marker":"[27]"},{"why":"Defines the master signature on Le-networks and establishes its geometric type, which the paper re-derives as a special case.","marker":"[4]"},{"why":"Provides the alternative vector-relation construction on reduced bipartite graphs whose signature is not geometric, serving as a contrast case.","marker":"[7]"},{"why":"States the Kasteleyn-theorem variant that the face-signature formula (Theorem 7.23) is conjectured to realize.","marker":"[56]"},{"why":"Proves the Kasteleyn interpretation for reduced bipartite graphs, which supports the conjecture for general PBDTP graphs.","marker":"[1]"}],"fun_headline_variants":["Geometric signatures alone keep plabic networks non-negative","Edge signatures unlock non-negative Grassmannian cells","Geometric signatures match Postnikov boundary measurements","Characterizing non-negative plabic networks via edge vectors"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is the PBDTP condition: every edge of the plabic graph must lie on at least one directed path from the boundary back to the boundary; if any edge fails this, path and cycle parities no longer determine a signature and the 'if and only if' statement gains extra gauge freedom.","fun_headline_variants_meta":{"raw":{"variants":["Geometric signatures alone keep plabic networks non-negative","Edge signatures unlock non-negative Grassmannian cells","Geometric signatures match Postnikov boundary measurements","Characterizing non-negative plabic networks via edge vectors"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000288,"raw_usage":{"total_tokens":1721,"prompt_tokens":1010,"completion_tokens":711,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":626,"completion_tokens_details":{"reasoning_tokens":651}},"tokens_in":626,"tokens_out":711,"duration_ms":5776,"temperature":1.0,"reasoning_tokens":651,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:17:57.034652+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Test the sign obstruction in equation (7.16) on a small PBDTP graph: pick a signature that differs from the geometric one in parity on some boundary-to-boundary path or some conservative cycle, and choose positive weights concentrated on that path or cycle with all other weights of order $\\delta \\ll 1$. The theorem predicts that some boundary matrix entry, or its denominator, changes sign and breaks total non-negativity for sufficiently small $\\delta$; finding a non-geometric signature whose matrix stays totally non-negative for all positive weights on such a graph would disprove Theorem 7.16.","supporting_citations":[{"cited_title":"A Formula for Pl¨ ucker Coordinates Associated with a Planar Network","cited_arxiv_id":null,"evidence_quote":"Provides the flow and conservative-flow formula that Theorem 3.12 extends from boundary sources to internal edges."},{"cited_title":"Press, Somerville, MA, 2016","cited_arxiv_id":null,"evidence_quote":"Formulates the relation-system and edge-signature problem on bipartite graphs that Theorem 7.16 resolves."},{"cited_title":"Cluster X –Varieties, Amalgamation and Poisson-Lie Groups","cited_arxiv_id":null,"evidence_quote":"Introduces amalgamation of cluster varieties, the operation whose totally non-negative specialization these relation systems represent."},{"cited_title":"Shapiro, and A","cited_arxiv_id":null,"evidence_quote":"Supplies the gauge-ray method for measuring winding and counting boundary-source crossings used to define geometric signatures."},{"cited_title":"Reducible M-curves for Le-networks in the totally-nonnegative Grassmannian and KP–II multiline solitons","cited_arxiv_id":null,"evidence_quote":"Defines the master signature on Le-networks and establishes its geometric type, which the paper re-derives as a special case."},{"cited_title":"Vector-relation configurations and plabic graphs","cited_arxiv_id":"1908.06959","evidence_quote":"Provides the alternative vector-relation construction on reduced bipartite graphs whose signature is not geometric, serving as a contrast case."},{"cited_title":"Variations on a theme of Kasteleyn, with application to the totally nonnegative Grassmannian","cited_arxiv_id":null,"evidence_quote":"States the Kasteleyn-theorem variant that the face-signature formula (Theorem 7.23) is conjectured to realize."},{"cited_title":"Kasteleyn theorem, geometric signatures and KP-II divisors on planar bipartite networks in the disk","cited_arxiv_id":"2012.13797","evidence_quote":"Proves the Kasteleyn interpretation for reduced bipartite graphs, which supports the conjecture for general PBDTP graphs."}],"review_version":1}