{"id":"b6635b8b-14d2-436c-a5fa-1b67333f279a","arxiv_id":"2506.15049","paper_version":2,"verdict":"REJECT","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"The base-cobase graph of the regular matroid R10 is bipartite, giving the first negative answer to the Farber-Richter-Shank Hamiltonian connectivity problem, while wheels and whirls are shown Hamiltonian connected; the separate series-parallel lattice path result relies on a false lemma.","lead":"This paper studies graphs built from matroid bases, asking whether a single path can visit every vertex exactly once. It reports new families where this is possible and a counterexample where it is not, but one of the main proofs is invalid as written.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 2.4(ii)⇒(iii) is false for a rank-2 triangle-with-doubled-edge block matroid; the spex(LPM) Mat/Ham theorem and Theorem 5.1's (ii)⇒(i) are unproven.","rationale":"The reader's weakest assumption pinpoints a genuine false step, and I reproduce it with a minimal concrete matroid. The counterexample satisfies the hypotheses of Lemma 2.4: it is connected, block, and has dim(P_{M,M*}) < |E|-1, yet the conclusion (iii) fails because M*/F is U_{0,2}. This is not an artifact of a missing edge case; the same matroid directly falsifies Lemma 3.2's rank claim and base-cobase equality. Since Theorem 3.6 is a central advertised contribution and depends on Lemma 3.2 through Lemma 3.4, the proof of that theorem collapses as written. The wheel/whirl Hamiltonian-connectivity results and the R10 bipartiteness result are separate constructions and appear not to rely on Lemma 2.4, so they may well survive, but the paper cannot be accepted with a false lemma supporting its first main theorem. I therefore concur with the reader's REJECT verdict and recommend no change. The R10 neighbor-list typo noted by the reader is real but secondary; it does not affect the bipartiteness conclusion and is not the load-bearing issue.","tokens_in":23536,"tokens_out":9325,"duration_ms":95324,"concrete_test":"In SageMath or by hand, construct M as the rank-2 matroid on {e,f,g,h} with circuits {e,f}, {e,g,h}, {f,g,h}. Compute (1) the affine hull of the four base-cobase incidence vectors and verify it has dimension 2; (2) all flacets of M and, for F={e,f}, compute M*/F and verify it is U_{0,2}; (3) compute M\\{e,f}, its rank, and the base-cobase set of U_{1,2}⊕(M\\{e,f}). If these match the values above, Lemma 2.4(ii)⇒(iii) and Lemma 3.2 are refuted, so the proof of Theorem 3.6 fails; if not, the objection is answered.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing defect is in the polytopal route to the first main theorem. Lemma 2.4 asserts (ii)⇒(iii): if dim(P_{M,M*}) < |E|-1 for a connected block matroid, then some flacet F has M_F and M*/F block and P_{M,M*} = P_{M_F⊕M*/F, ...}. This is false. Take M to be the rank-2 graphic matroid of a triangle with one edge doubled, E={e,f,g,h}, with circuits {e,f}, {e,g,h}, {f,g,h}. M is connected and block; the four base-cobases are {e,g}, {e,h}, {f,g}, {f,h}, and all satisfy x_e+x_f=1 and x_g+x_h=1, so dim(P_{M,M*})=2 < 3 = |E|-1. But the only non-trivial flacet is F={e,f}; in M* this is a 2-cocircuit, so M*/F = U_{0,2}, which is not a block matroid. Thus (ii) does not imply (iii). The failure propagates to Lemma 3.2: for C={e,f}, the lemma claims rank(M\\C)=r-1=1, but M\\C = U_{2,2} has rank 2, and U_{1,2}⊕(M\\C) has no base-cobases although M does. Since Lemma 3.4 and Theorem 3.6 invoke Lemma 3.2, the advertised Mat/Ham result for spex(LPM) is currently unsupported; Theorem 5.1's (ii)⇒(i) also relies on it. The R10 and wheel/whirl sections are independent and are not invalidated by this defect.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies base-cobase graphs G(M,M*) of block matroids and their Hamiltonian connectivity. It proposes: (1) a polytopal criterion \"Mat\" and uses it to prove that series-parallel extensions of lattice path matroids satisfy Mat, hence Ham and all properties from Problem 1.1; (2) complete structural descriptions of wheel and whirl base-cobase graphs, used to prove Hamiltonian connectivity for those classes; and (3) an explicit description of the base-cobase graph of R10, showing it is bipartite and therefore not Hamiltonian connected, which would be the first refutation of any property in Problem 1.1. The wheel/whirl and R10 parts are largely independent, but the proof of the spex(LPM) result relies on a polytopal lemma whose key implication is false.","tokens_in":23907,"tokens_out":11085,"duration_ms":113020,"significance":"The R10 result, if correct, would be a significant contribution: it would answer a question of Farber, Richter, and Shank in the negative within the class of regular matroids. The explicit descriptions of the wheel and whirl base-cobase graphs, and the concrete description of G(R10,R10*), are valuable and may serve future work. The paper is clearly written and the arguments are mostly self-contained, and there is no data fitting or circularity in the proofs. However, the central lemma used for the spex(LPM) theorem is false, so the first main theorem and its corollaries are unsupported in the present version. The R10 and wheel/whirl sections are not affected by this defect, but the paper as a whole cannot be accepted without replacing the false lemma and re-proving the results that depend on it.","major_comments":[{"comment":"Lemma 2.4 is false as stated. Consider the rank-2 graphic matroid M on E={e,f,g,h} obtained from a triangle by duplicating one edge into parallel edges e and f; its circuits are {e,f}, {e,g,h}, and {f,g,h}. M is connected and block, and its base-cobases are {e,g}, {e,h}, {f,g}, and {f,h}. The incidence vectors of all base-cobases satisfy x_e+x_f=1 and x_g+x_h=1, so dim(P_{M,M*})=2<3=|E|-1, giving condition (ii). The only nontrivial tight set is F={e,f}, but F is not a flacet: E\\F={g,h} is not a flat of M, and {e,f} is not a flat of M*. Hence there is no nontrivial flacet for which M_F and M*/F are both block matroids, so (ii) does not imply (iii). In fact (i) also does not imply (iii) in this example. The proof of (ii)⇒(iii) is unjustified at the step where a support hyperplane containing P_{M,M*} is asserted to be a facet of P_M; in the example the hyperplane x_e+x_f=1 cuts a 2-dimensional face of the 3-dimensional base polytope but is not facet-defining.","section":"Section 2, Lemma 2.4"},{"comment":"Lemma 3.2 is also false, and its proof invokes the false Lemma 2.4. For the same matroid M with circuit C={e,f}, M\\C is U_{2,2} on {g,h}; it is not a block matroid and has rank 2, not r-1=1. Moreover M'=U_{1,2}\\oplus(M\\C) has no base-cobases, whereas M has four. Thus the reduction used in Lemma 3.4 fails, and Theorem 3.6, which applies Lemma 3.4 to spex(LPM), is unsupported. Since Theorem 5.1 uses Theorem 3.6 for the direction (ii)⇒(i), that direction is likewise unsupported. The authors would need a correct replacement for Lemma 2.4 or a different proof strategy for the spex(LPM) result.","section":"Section 3, Lemmas 3.2 and 3.4, Theorem 3.6"}],"minor_comments":[{"comment":"The list of neighbors of [abcde]S2 in Theorem 5.4 contains [abcde]S2 twice; the proof indicates that one of the entries should be [acbde]S2. The statement of the bipartition in Corollary 5.5 also appears to misstate Y: it should be the D5-vertices of even sign and the S2-vertices of odd sign.","section":"Theorem 5.4 and Corollary 5.5"},{"comment":"The base cases handled \"computationally\" for n≤6 in Theorem 4.8 and for n=3 in Theorem 4.9 are not documented; please provide the code, a table of verifications, or an explicit finite check so the reader can reproduce them.","section":"Theorems 4.8 and 4.9"},{"comment":"Several steps in the proofs of Theorems 4.8 and 4.9 are described as \"suitable stitching\" or \"analogous\" without full details. Given the complexity of the case analysis, the authors should make these concatenations explicit or provide sufficiently detailed figures.","section":"Section 4.3"},{"comment":"There are several typographical errors, including \"Hamiltoninan\" and \"Haimiltonian\" for \"Hamiltonian\", and Proposition 4.4 appears to have a rendering issue with the floor function in the distance formula. These should be corrected in a final revision.","section":"Throughout"}],"recommendation":"reject","confidential_remarks":"The false Lemma 2.4 is a load-bearing error: it invalidates Lemma 3.2, Lemma 3.4, Theorem 3.6, and the (ii)⇒(i) direction of Theorem 5.1. The R10 section is independent and seems promising after minor corrections, and the wheel/whirl sections may be salvageable with more detail, but the current manuscript cannot be published without a substantially different proof of the first main result."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things you should know. The base-cobase graph of R10 is bipartite, so Hamiltonian connectivity fails—this is the first refutation of any property in Farber–Richter–Shank's Problem 1.1, and it's the real news. The other thing is that the paper's advertised main theorem, Mat for series-parallel extensions of lattice path matroids, is currently unsupported: Lemma 2.4(ii)⇒(iii) is false.\n\nThe paper does some things well. Lemma 2.5 gives a clean characterization of Mat in terms of identically self-dual matroids and the 1-skeleton of the base-cobase polytope. The structure theorems for wheels and whirls (Propositions 4.2 and 4.3) are concrete and useful, and the Hamiltonian connectivity proofs are detailed. The R10 analysis is the highlight: the description of the base-cobase graph as a Cayley-like graph on S5/S2 and S5/D5, with the sign-based bipartition, is elegant and convincing. There is a small typo in Theorem 5.4's neighbor list—[abcde]_{S2} appears twice—but the surrounding proof makes the intended list clear.\n\nThe soft spot is not small. Lemma 2.4 claims that for a connected block matroid, dim(P_{M,M*}) < |E|−1 forces a nontrivial flacet F with both M_F and M*/F block, and with P_{M,M*} equal to the base-cobase polytope of M_F ⊕ M*/F. The stress-test example is a rank-2 triangle with a doubled edge: E = {e,f,g,h}, circuits {e,f}, {e,g,h}, {f,g,h}. The base-cobases are the four 2-sets choosing one of {e,f} and one of {g,h}, so dim = 2 < 3. The only nontrivial flacet is F = {e,f}; M_F is U_{1,2}, which is block, but contracting F in the dual gives U_{0,2}, which is not block. So (ii) holds and (iii) fails. The proof's rank calculation about M*/F is wrong.\n\nThis breaks Lemma 3.2, which is used in Lemma 3.4 to lift properties to series-parallel extensions. Consequently Theorem 3.6 and the (ii)⇒(i) direction of Theorem 5.1 are unproven. The wheel/whirl and R10 sections do not depend on Lemma 2.4 and are not affected.\n\nOne more minor issue: the wheel/whirl proofs say n≤6 \"can be dealt with computationally\" but give no details. That's a documentation gap, not a mathematical error.\n\nBottom line: the paper has a significant independent result in the R10 counterexample, and the wheel/whirl work looks solid, but the current version can't be accepted as is. It deserves a serious referee, and a major revision should either repair Lemma 2.4 or restrict the claims that rely on it.","headline":"The R10 counterexample is real and gives the first negative answer to Farber–Richter–Shank, but the paper's main polytopal theorem rests on a false lemma and needs major rework.","tokens_in":24441,"tokens_out":9699,"would_cite":true,"duration_ms":88420,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["05B35","05C45"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves Hamiltonian connectivity of base-cobase graphs for lattice path extensions, wheels, and whirls, and shows the regular matroid R10 refutes the general question.","keywords":["base-cobase graph","Hamiltonian connectivity","matroids","lattice path matroids","regular matroids","wheels and whirls","base-cobase polytope","R10 matroid"],"falsifier":"Take the connected block matroid on four elements given by a triangle with one doubled edge. Its base-cobase polytope has dimension 2, strictly less than $|E|-1=3$, yet the only non-trivial flacet $F=\\{e,f\\}$ yields $M^*/F=U_{0,2}$, which is not a block matroid; this is exactly the configuration Lemma 2.4 says cannot occur, and it would knock out Lemma 3.2 and the series-parallel reduction built on it.","tokens_in":23297,"feed_emoji":"🔗","tokens_out":12645,"duration_ms":113991,"temperature":0.7,"pith_summary":"Matroids have base graphs whose vertices are bases and that are known to be hypercubes or Hamiltonian connected; the paper asks whether the induced graph on base-cobases—sets that are bases of a matroid and of its dual—keeps that property. It answers yes for three families: series-parallel extensions of lattice path matroids, wheels, and whirls all have Hamiltonian-connected base-cobase graphs. It also shows when the polytopal proof method can work at all: a block matroid's base-cobase graph is the 1-skeleton of its base-cobase polytope exactly when the graph is itself a base graph, which among regular matroids happens only for direct sums of the two-element uniform matroid. The negative result is that the regular matroid $R_{10}$ has a 72-vertex base-cobase graph that is bipartite and therefore not Hamiltonian connected, giving the first counterexample to a property in the 1985 problem list reproduced as Problem 1.1. The paper thus delimits where Hamiltonian connectivity of base graphs extends to base-cobase graphs.","feed_headline":"R10's base-cobase graph is bipartite, so Ham fails","feed_subtitle":"The first counterexample to a 1985 base-cobase problem: regular matroids need not be Hamiltonian connected.","key_machinery":"The base-cobase polytope $P_{M,M^*}=P_M\\cap P_{M^*}$ is the first workhorse: it lets the authors treat the exchange graph as the 1-skeleton of a $(0,1)$-polytope, where the dichotomy for such skeletons—hypercube or Hamiltonian connected—applies whenever the graph is the full skeleton. Mat is the property that this skeleton is again the base graph of some matroid, equivalently that the base-cobase set is the base set of an identically self-dual matroid related to $M$ by a special weak map; this is what transfers from a minor-closed class to its series-parallel extensions. For wheels and whirls, the load-bearing structure is the explicit decomposition of $G(M,M^*)$ into two copies of the hypercube $Q_n$, namely $Q_n^+$ and $Q_n^-$, with $0$ and $1$ removed for wheels and identified for whirls, glued along 'lean' vertices whose supports are cyclic intervals; Hamiltonian paths are assembled from hypercube path coverings of faulty hypercubes. For $R_{10}$, the machinery is the model of its elements as the 10 triples of $[5]$, with circuits exactly the 4-sets and their 6-set complements; this yields the 72 base-cobases organized as $(S_5/D_5)\\cup(S_5/S_2)$, with the five neighbours of each vertex read off from the triple sets, and a two-colouring by the sign of a permutation that proves bipartiteness.","core_discovery":"The central discovery is that Hamiltonian connectivity of base-cobase graphs is a real phenomenon but not a universal one. The paper proves the property Mat: when a block matroid's base-cobase graph is exactly the base graph of some matroid, then the graph is the 1-skeleton of the base-cobase polytope and Hamiltonian connectivity follows from the known dichotomy for 1-skeleta of $(0,1)$-polytopes. Mat is shown to be inherited by series-parallel extensions of minor-closed classes, and lattice path matroids are shown to have base-cobase sets that are again the bases of a lattice path matroid, so every series-parallel extension of a lattice path matroid satisfies Ham. For wheels and whirls, the paper gives an explicit decomposition of the base-cobase graph into two hypercubes, with the extreme vertices removed for wheels and identified for whirls, stitched along 'lean' vertices, and uses hypercube path-covering results to prove Hamiltonian connectivity in both families. Finally, working inside regular matroids, the paper describes the base-cobase graph of $R_{10}$ as 72 vertices indexed by cosets of the dihedral and two-element subgroups of $S_5$, proves the full edge structure, and observes a bipartition by permutation sign; because any bipartite graph with more than two vertices fails Hamiltonian connectivity, $R_{10}$ refutes the Hamiltonian part of Problem 1.1.","pith_inferences":["The bipartiteness of the $R_{10}$ base-cobase graph makes the paper's open question about equicardinal bipartition classes a natural next test: unequal colour classes would immediately yield a non-Hamiltonian base-cobase graph.","The hypercube-stitching description used for wheels and whirls may extend to other multipath matroids or to necklaces, offering a route toward Hamiltonian paths in base-cobase graphs of larger positroid classes.","Because Mat holds only in very special classes, the polytopal-skeleton route cannot be expected to prove Hamiltonian connectivity for all regular matroids; any positive result there would likely need direct exchange-path constructions rather than a base-graph reduction."],"forward_implications":["Every block matroid in the class of series-parallel extensions of lattice path matroids has a Hamiltonian-connected base-cobase graph, so it satisfies the connectedness, circuit, strong circuit, diameter, polynomial diameter, and Hamiltonian properties from Problem 1.1.","The base-cobase graphs of wheels and whirls are Hamiltonian connected even though they are not base graphs of any matroid.","The base-cobase graph of $R_{10}$ is bipartite, so no Hamiltonian path can connect vertices in opposite colour classes; this refutes the Hamiltonian-connectivity question for regular matroids.","Among regular matroids, the polytopal transfer method works only for direct sums of $U_{1,2}$, so the proof of Hamiltonian connectivity for any broader regular class would need a different mechanism."],"supporting_citations":[{"why":"Gives the dichotomy for 1-skeleta of (0,1)-polytopes, either hypercube or Hamiltonian connected, which is the base-graph theorem the paper tries to extend.","marker":"[51]"},{"why":"Raises Problem 1.1, including the Hamiltonian-connectivity question that $R_{10}$ refutes.","marker":"[26]"},{"why":"Supplies the hypercube path-covering lemmas with prescribed endpoints used to build Hamiltonian paths for wheels and whirls.","marker":"[16]"},{"why":"Introduces base-cobase graphs and polytopes and provides the identity $P_{M,M^*}=P_M\\cap P_{M^*}$ that anchors the polytopal arguments.","marker":"[21]"},{"why":"Provides lattice path matroid duality and minor-closure facts used to show the base-cobases of a lattice path matroid form another lattice path matroid.","marker":"[14]"},{"why":"Introduces tight sets and the reconfiguration framework from which Lemma 2.4's dimension characterization is drawn.","marker":"[6]"},{"why":"Used in Theorem 5.1 to decompose an identically self-dual matroid as $M/F\\oplus M_F$, the key step showing Mat is rare among regular matroids.","marker":"[43]"},{"why":"Classifies connected binary identically self-dual matroids as $U_{1,2}$, the endpoint of the regular-matroid classification.","marker":"[42]"}],"fun_headline_variants":["R10 refutes Hamiltonian base-cobase conjecture","Wheels, whirls, lattice paths: Hamiltonian base-cobase","R10's bipartite base-cobase graph breaks Hamiltonian connectivity","Hamiltonian base-cobase graphs: yes for many, no for R10"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is Lemma 2.4, which says that a dimension deficit in the base-cobase polytope of a connected block matroid forces a non-trivial flat whose restriction and contraction are again block matroids, and this splitting step is what carries the series-parallel extension proof.","fun_headline_variants_meta":{"raw":{"variants":["R10 refutes Hamiltonian base-cobase conjecture","Wheels, whirls, lattice paths: Hamiltonian base-cobase","R10's bipartite base-cobase graph breaks Hamiltonian connectivity","Hamiltonian base-cobase graphs: yes for many, no for R10"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.0015,"raw_usage":{"total_tokens":6068,"prompt_tokens":1048,"completion_tokens":5020,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":664,"completion_tokens_details":{"reasoning_tokens":4943}},"tokens_in":664,"tokens_out":5020,"duration_ms":34752,"temperature":1.0,"reasoning_tokens":4943,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T19:50:08.916620+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the connected block matroid on four elements given by a triangle with one doubled edge. Its base-cobase polytope has dimension 2, strictly less than $|E|-1=3$, yet the only non-trivial flacet $F=\\{e,f\\}$ yields $M^*/F=U_{0,2}$, which is not a block matroid; this is exactly the configuration Lemma 2.4 says cannot occur, and it would knock out Lemma 3.2 and the series-parallel reduction built on it.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the dichotomy for 1-skeleta of (0,1)-polytopes, either hypercube or Hamiltonian connected, which is the base-graph theorem the paper tries to extend."},{"cited_title":"Farber, B","cited_arxiv_id":null,"evidence_quote":"Raises Problem 1.1, including the Hamiltonian-connectivity question that $R_{10}$ refutes."},{"cited_title":"Casta˜neda and I","cited_arxiv_id":null,"evidence_quote":"Supplies the hypercube path-covering lemmas with prescribed endpoints used to build Hamiltonian paths for wheels and whirls."},{"cited_title":"Cordovil and M","cited_arxiv_id":null,"evidence_quote":"Introduces base-cobase graphs and polytopes and provides the identity $P_{M,M^*}=P_M\\cap P_{M^*}$ that anchors the polytopal arguments."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides lattice path matroid duality and minor-closure facts used to show the base-cobases of a lattice path matroid form another lattice path matroid."},{"cited_title":"B´erczi, B","cited_arxiv_id":null,"evidence_quote":"Introduces tight sets and the reconfiguration framework from which Lemma 2.4's dimension characterization is drawn."},{"cited_title":"Lucas, Weak maps of combinatorial geometries , Trans","cited_arxiv_id":null,"evidence_quote":"Used in Theorem 5.1 to decompose an identically self-dual matroid as $M/F\\oplus M_F$, the key step showing Mat is rare among regular matroids."},{"cited_title":"Lindstr¨om, On binary identically self-dual matroids , Eur","cited_arxiv_id":null,"evidence_quote":"Classifies connected binary identically self-dual matroids as $U_{1,2}$, the endpoint of the regular-matroid classification."}],"review_version":2}