{"id":"d1dd06c4-d33c-413b-8cc7-9dc6587b6884","arxiv_id":"1908.07664","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For K4-free graphs, the graphic arrangement group embeds in a product of free groups, so it is residually free, torsion-free, and acts properly on a CAT(0) cube complex.","lead":"A graph determines a group of braids in which some strands are allowed to cross, and this group is called a graphic arrangement group. The paper proves that for graphs with no four mutually connected vertices, the group embeds into a product of free groups, giving clean proofs of residual properties and a CAT(0) cube complex action.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Corollary 6.5 is false: two octahedra joined at a vertex give P_Γ ≅ P_O×P_O, of type FP_7-not-FP_8, contradicting the claimed FP_15-not-FP_16.","rationale":"Reading in good faith, the core injectivity theorem (Theorem 5.5) and its consequences appear sound. I checked the case |{i,j,r,s}|=3 omitted in Proposition 5.2: the pure braid relation (4)(iii) indeed forces the remaining two generators to commute when one triangle edge is deleted, and the crossing case in |{i,j,r,s}|=4 follows from Remark 2.1, so the reader's stated weakest assumption is a presentational gap, not a mathematical error. The real soft spot is in Section 6. The statement 'Γ connected and no maximal 2-cliques ⇒ Λ_X connected' is false, and the two-octahedron example is a concrete counterexample to Corollary 6.5(ii). This changes the verdict from CONDITIONAL to REJECT: the paper contains a false advertised theorem, even though the injectivity and residual-property results may be correct and repairable by adding a connectedness-of-Λ or block-decomposition hypothesis.","tokens_in":17188,"tokens_out":26538,"duration_ms":264554,"concrete_test":"Take Γ = Octa ∪_v Octa. Step 1: list maximal cliques; they are the 16 triangular faces, so the hypotheses of Corollary 6.5 hold with m=16. Step 2: use Proposition 4.4 with X={v}; since P_1 is trivial, the pullback is the direct product, so P_Γ ≅ P_O × P_O. Step 3: apply Theorem 6.4/Corollary 6.5 to a single octahedron to get P_O of type FP_7-not-FP_8. Step 4: conclude the product is FP_7-not-FP_8 and compare with the predicted FP_15-not-FP_16; the contradiction shows the stated theorem is false.","verdict_should_be":"REJECT","load_bearing_attack":"The proof of Theorem 6.4 asserts that 'since Γ is connected and has no maximal 2-cliques, Λ_X is connected.' This is false: the incidence graph of edges and 3-cliques is disconnected whenever two triangles meet only at a vertex. The failure is not a harmless presentation issue—Corollary 6.5 as stated is false. Let Γ be the 1-skeleton of two octahedra identified at one vertex. Γ is connected, K4-free, and every maximal clique is one of the 16 triangular faces; Λ_Γ is the disjoint union of the two octahedron incidence graphs, so b1(Λ)>0 and Λ has no isthmuses. Corollary 6.5(ii) with m=16 predicts P_Γ is FP_{15} but not FP_{16}. But because the two octahedra intersect in a 1-clique, Proposition 4.4 (generalized parallel connection, pullback over P_1=1) gives P_Γ ≅ P_O×P_O. By Theorem 6.4 applied to one octahedron, P_O is FP_7 and not FP_8; a direct product of two such groups is FP_7 and not FP_8 (finiteness length is the minimum, and FP_n is inherited by retracts). This contradicts the predicted finiteness type. The residual-freeness portion (Theorem 5.5) is not affected, but the advertised homological finiteness result is wrong for cut-vertex gluings.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the graphic arrangement groups P_Γ associated to a finite simple graph Γ, defined as the fundamental group of the complement of the graphic hyperplane arrangement. The main results are: (1) for K_4-free Γ, the product of deletion maps to the maximal cliques is injective, embedding P_Γ into a product of free groups, yielding residual freeness, torsion-freeness, linearity, residual torsion-free nilpotence, and a free proper action on a CAT(0) cube complex; (2) an extension of this embedding to graphs whose 4-cliques are almost disjoint, with an example showing the map may be injective even when this condition fails; (3) a homological finiteness type theorem (Theorem 6.4 and Corollary 6.5) asserting that for connected Γ with all maximal cliques of size 3, P_Γ is of type FP_{m-1} but not FP_m when the incidence graph of edges and 3-cliques has a cycle and no isthmuses, where m is the number of 3-cliques; and (4) an extension of the finiteness results to the graphic full braid group B_Γ.","tokens_in":17502,"tokens_out":10107,"duration_ms":87647,"significance":"The embedding theorem for K_4-free graphs, if fully proved, gives a large class of groups with strong residual properties, and the method via retractive families is elegant and clearly presented. The counterexample in Example 5.11 shows that the injectivity phenomenon is subtle and interesting. However, the advertised finiteness type theorem is false in the stated generality, as demonstrated by the two-octahedra counterexample below; this substantially weakens the paper's contribution. The proof of the key proposition underlying the embedding theorem also contains an omitted case. The paper extends a line of work by the same authors and Randell, and the dependence on [6] is heavy but visible.","major_comments":[{"comment":"The proof asserts that 'Since Γ is connected and has no maximal 2-cliques, Λ_X is connected.' This assertion is false. Let Γ be the 1-skeleton of two octahedra identified at a single vertex. Then Γ is connected and K_4-free, and every maximal clique is one of the 16 triangular faces. The incidence graph Λ_Γ is the disjoint union of the incidence graphs of the two octahedra, so it is disconnected; it has first Betti number 10 and no isthmuses. Corollary 6.5(ii) with m=16 then predicts that P_Γ is of type FP_15 but not FP_16. However, by Proposition 4.4 with X a 1-clique (the identified vertex), P_Γ is the direct product P_O × P_O, where O is a single octahedron. Applying Corollary 6.5(ii) to O (m=8), P_O is FP_7 and not FP_8. A direct product of two such groups is FP_7 and not FP_8, since FP_n is inherited by retracts and the finiteness length is the minimum of the factors. This contradicts the claimed finiteness type. The theorem needs an additional hypothesis such as connectedness of Λ_Γ, or a separate treatment of graphs that decompose along cut vertices.","section":"Section 6, Theorem 6.4 and Corollary 6.5"},{"comment":"The proof for the case |{i,j,r,s}| = 3 is omitted with the words 'We leave the case-by-case verification to the reader.' This proposition is used to verify condition (i) of Theorem 3.4, which is load-bearing for the injectivity theorem (Theorem 5.5). As written, the proof is incomplete; the case analysis should be supplied explicitly.","section":"Section 5, Proposition 5.2"}],"minor_comments":[{"comment":"In the first sentence of the proof, 'transverse to Y' should be 'transverse to X(Γ)'.","section":"Section 5, Proposition 5.2"},{"comment":"In the |{i,j,r,s}| = 4 case, the phrase 'Since S is not a 4-clique' is imprecise; the intended meaning is that the vertex set {i,j,r,s} is not a 4-clique.","section":"Section 5, Proposition 5.2"},{"comment":"The verification of injectivity relies on [6, Thm. 3.2.13] and several 'one can check' identities; the example is only sketched, and the reader is asked to trust a lengthy case analysis. This is acceptable for an example, but the exposition would benefit from more details.","section":"Section 5, Example 5.11"},{"comment":"The paper relies heavily on [6] (cited as 'to appear') for Theorem 3.4, Propositions 6.1–6.3, and Example 5.11. Since [6] is not yet published and is by the same authors, the editor may wish to verify its availability, and the authors should consider stating the quoted results more explicitly.","section":"References, [6]"}],"recommendation":"major_revision","confidential_remarks":"The false claim in Theorem 6.4 is serious and should be corrected before publication; the two-octahedra counterexample shows that the incidence graph connectivity assertion fails, and the advertised finiteness type is wrong in that case. The dependence on the authors' own prior paper [6] is substantial; while citation to previous work is legitimate, the authors should ensure that the key quoted results are either stated or clearly available to readers. The main embedding theorem (Theorem 5.5) appears plausible, but the omitted case in Proposition 5.2 must be supplied."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things you should know. The K4-free embedding theorem (5.5) is a genuine contribution and looks right. The finiteness-type theorem (6.4/Cor. 6.5) is false as stated; I have a concrete counterexample.\n\nThe new content: Cohen and Falk apply their retractive-family machinery ([6]) to graphic arrangement groups. For K4-free graphs, they prove the product of deletion maps into products of pure braid groups on cliques is injective, so P_Gamma is residually free, residually torsion-free nilpotent, linear, and acts freely and properly on a CAT(0) cube complex. The argument is mostly careful. The generalization to almost-disjoint 4-cliques is plausible. The omitted case check in Prop. 5.2 is a presentational gap, not a visible error. The heavy reliance on [6] is fine, since the new applications are real.\n\nThe problem is Section 6. The proof of Thm. 6.4 asserts that because Gamma is connected and has no maximal 2-cliques, the incidence graph Lambda_Gamma of edges and 3-cliques is connected. That's false: two triangles meeting only in a vertex yield two disconnected components of Lambda_Gamma. The counterexample is two octahedra identified at a single vertex. Gamma is connected and K4-free, every maximal clique is a triangle, m=16, b1(Lambda)>0, no isthmuses. Cor. 6.5(ii) predicts P_Gamma is FP_15 not FP_16. But by Prop. 4.4 the two octahedra meet in a 1-clique, so P_Gamma cong P_O x P_O, and finiteness length of a product is the minimum of the factors. Since each octahedron has m=8, P_O is FP_7-not-FP_8, so P_Gamma is FP_7-not-FP_8, contradicting the claimed result. The false connectivity assumption is load-bearing: it is used to get injectivity after projectivization (Prop. 6.1) and to compute the cokernel (Prop. 6.2). So Thms. 6.4, Cor. 6.5, and Thm. 7.7 need to be withdrawn or at least restricted to graphs where Lambda_Gamma is connected. Cor. 6.6 about non-K(pi,1) is also suspect.\n\nThe reader's take was too kind on soundness; the real issue is not the omitted case in 5.2 but the connectivity claim in 6.4. Who is this for: arrangement group people will want the embedding theorem. The finiteness claims are wrong and should not be used. It deserves a serious referee because the valid part is important and the flaw is subtle enough to need expert confirmation. My recommendation: send to peer review, but the authors will have to substantially revise or remove Section 6 before publication. The paper as it stands cannot be accepted.","headline":"The K4-free embedding theorem is a real contribution, but the finiteness-type theorem (6.4/6.5) is false as stated: two octahedra joined at a vertex give a direct product with the wrong finiteness length.","tokens_in":18045,"tokens_out":6482,"would_cite":true,"duration_ms":738291,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["20F36","32S22","52C35","20E26"],"pacs":[],"model":"deepseek-v4-flash","headline":"For finite simple graphs with no K4, the associated graphic arrangement group embeds in a product of free groups via the product of deletion maps, yielding residual freeness, linearity, and a CAT(0) action.","keywords":["pure braid group","hyperplane arrangement","graphic arrangement","homological finiteness type","K4-free graph","residually free","CAT(0) cube complex","graphic braid group"],"falsifier":"Compute, from the Artin presentation of $P_\\Gamma$ (relations (4)), the commutator $[a_{ij},a_{rs}]$ for a specific $K_4$-free graph where the two edges share exactly one vertex and the three vertices do not form a clique; the paper asserts it is trivial in all such cases, so finding a single graph where it is nontrivial would refute Theorem 5.5.","tokens_in":16957,"feed_emoji":"🪢","tokens_out":12905,"duration_ms":109888,"temperature":0.7,"pith_summary":"The paper studies the groups $P_\\Gamma$ obtained from pure braid groups by allowing pairs of strands to cross whenever the corresponding edge is missing from a graph $\\Gamma$. Its main result is that if $\\Gamma$ has no four mutually adjacent vertices (is $K_4$-free), the product of maps that delete sets of vertices embeds $P_\\Gamma$ into a direct product of free groups. Because such properties are inherited by subgroups and products, $P_\\Gamma$ is then residually free, torsion-free, residually torsion-free nilpotent, residually finite, and linear, and it acts freely and properly on a CAT(0) cube complex. A second result computes homological finiteness type for graphs whose maximal cliques are all triangles: if the incidence graph of edges and triangles has a cycle and no isthmuses, then $P_\\Gamma$ is of type $FP_{m-1}$ but not $FP_m$, where $m$ is the number of triangles; the same holds for the natural extension $B_\\Gamma$ by the graph's automorphism group.","feed_headline":"K4-free graphs embed braid groups in products of free groups","feed_subtitle":"Deletion maps give these graphic arrangement groups residual freeness, linearity, and torsion-freeness at once.","key_machinery":"The central mechanism is an injectivity criterion for retractive families of subsets of a group's generating set, proved in the authors' prior work (Theorem 3.4). A subset $S$ of the generating set $Y$ is called retractive when the quotient by the generators outside $S$ is injective on the subgroup generated by $S$; a family of pairwise incomparable subsets is retractive if all intersections and singletons are. The criterion says that if $Y$ is covered by a retractive family $\\mathcal{X}$ and two additional commutation conditions hold—transverse pairs of generators commute and elements outside a subset $S$ centralize the commutator subgroup of the target group—then the product of the quotient maps $\\prod\\rho_S$ is injective. The paper verifies both conditions for graphic arrangement groups using the clique structure of $\\Gamma$: cliques are retractive because they give semidirect product splittings (via the Fadell–Neuwirth bundle), and the commutation conditions follow from the Artin presentation of the pure braid group.","core_discovery":"We prove that for a $K_4$-free graph $\\Gamma$, the homomorphism $\\rho_{\\mathcal{X}(\\Gamma)}: P_\\Gamma \\to \\prod_{X\\in\\mathcal{X}(\\Gamma)} P_X$, where $\\mathcal{X}(\\Gamma)$ consists of the 3-cliques and maximal 2-cliques of $\\Gamma$, is injective. Since each $P_X$ is a product of free groups, $P_\\Gamma$ embeds in a right-angled Artin group, and consequently is residually free, torsion-free, residually torsion-free nilpotent, linear, residually finite, and acts freely and properly on a CAT(0) cube complex. The embedding is extended to graphs whose 4-cliques are almost disjoint (no two share a 3-clique), with $P_\\Gamma$ then embedding in a product of pure braid groups of rank at most four; the hypothesis is shown not to be necessary by example. For connected graphs in which every maximal clique is a triangle, with $m$ triangles and incidence graph $\\Lambda_\\Gamma$, we show $P_\\Gamma$ is of type $FP_{m-1}$ but not $FP_m$ when $\\Lambda_\\Gamma$ contains a cycle and has no isthmuses; the same dichotomy holds for the graphic braid group $B_\\Gamma$, defined as the extension of $P_\\Gamma$ by $\\mathrm{Aut}(\\Gamma)$.","pith_inferences":["The injectivity of $\\rho_{\\mathcal{X}(\\Gamma)}$ appears to fail exactly when overlapping cliques force Brunnian-type braids into the kernel; Problem 5.13 asks for a characterization, and the paper's Example 5.11 shows the sufficient hypotheses are not necessary, so testing graphs with two 4-cliques sharing a 3-clique but with a separating edge set is a natural next step.","Since $P_\\Gamma$ embeds in a product of free groups for $K_4$-free $\\Gamma$, the word and conjugacy problems for these groups are solvable by free-group algorithms, though the paper does not address complexity bounds; one could estimate the growth of the embedding's redundancy.","The finiteness-type dichotomy depends only on the incidence graph of edges and triangles, which suggests a purely combinatorial way to distinguish the homotopy types of graphic arrangement complements among $K_4$-free graphs, for instance by the first Betti number of $\\Lambda_\\Gamma$.","The construction of $B_\\Gamma$ as an extension by $\\mathrm{Aut}(\\Gamma)$ shows that residual freeness of $P_\\Gamma$ does not pass to $B_\\Gamma$ in general (it may have torsion), but the finiteness type does; a natural question is whether graph automorphisms preserve the linearity or CAT(0) action of $P_\\Gamma$."],"forward_implications":["Every $K_4$-free graphic arrangement group $P_\\Gamma$ is residually free, torsion-free, residually torsion-free nilpotent, residually finite, and linear; it acts freely and properly on a CAT(0) cube complex and hence has the Haagerup property.","For graphs whose 4-cliques are almost disjoint, $P_\\Gamma$ embeds in a product of pure braid groups of rank at most four, so it is torsion-free, linear, residually finite, residually torsion-free nilpotent, and acts freely and properly on a CAT(0) complex (not cocompactly in general).","When $\\Gamma$ is connected with every maximal clique of size 3 and the incidence graph of edges and 3-cliques has a cycle and no isthmuses, $P_\\Gamma$ is of type $FP_{m-1}$ but not $FP_m$, where $m$ is the number of 3-cliques; the same holds for the graphic braid group $B_\\Gamma$.","If such a graphic arrangement has an incidence graph containing a cycle, the arrangement is not a $K(\\pi,1)$ arrangement."],"supporting_citations":[{"why":"Supplies the injectivity criterion for retractive families (Theorem 3.4) and the finiteness-type result (Proposition 6.3) on which both main theorems rest.","marker":"[6]"},{"why":"Origin of Proposition 5.2, the commuting condition for transverse pairs of generators.","marker":"[21]"},{"why":"Proves that the pure braid group on four strands is not residually free, delimiting the $K_4$-free hypothesis.","marker":"[7]"},{"why":"Gives the Fadell–Neuwirth bundle whose section underlies the semidirect product splitting that makes cliques retractive.","marker":"[11]"},{"why":"Computes the $\\Sigma$-invariants of (right-angled) Artin groups used to derive the $FP_{m-1}$ but not $FP_m$ dichotomy.","marker":"[24]"},{"why":"Provides the definition of type $F_m$ and the finite-index transfer used for the graphic braid group $B_\\Gamma$.","marker":"[3]"},{"why":"Shows that braid groups on at most four strands act freely and properly on a CAT(0) complex, used in the almost-disjoint 4-clique extension.","marker":"[2]"}],"fun_headline_variants":["K4-free graph braid groups are residually free","Deletion maps embed graphic braid groups in free groups","K4-free graphs give residually free graphic braid groups","Graphic arrangement groups: deletion maps yield residual freeness","Residually free braid groups from K4-free graph arrangements"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The main injectivity theorem depends on the claim that two braid generators of $P_\\Gamma$ commute whenever their vertex sets are not contained together in any clique of $\\Gamma$; the proof leaves the three-vertex subcase to the reader, and if that subcase is wrong the theorem fails.","fun_headline_variants_meta":{"raw":{"variants":["K4-free graph braid groups are residually free","Deletion maps embed graphic braid groups in free groups","K4-free graphs give residually free graphic braid groups","Graphic arrangement groups: deletion maps yield residual freeness","Residually free braid groups from K4-free graph arrangements"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001657,"raw_usage":{"total_tokens":6648,"prompt_tokens":1082,"completion_tokens":5566,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":698,"completion_tokens_details":{"reasoning_tokens":5482}},"tokens_in":698,"tokens_out":5566,"duration_ms":38974,"temperature":1.0,"reasoning_tokens":5482,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:00:24.112472+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute, from the Artin presentation of $P_\\Gamma$ (relations (4)), the commutator $[a_{ij},a_{rs}]$ for a specific $K_4$-free graph where the two edges share exactly one vertex and the three vertices do not form a clique; the paper asserts it is trivial in all such cases, so finding a single graph where it is nontrivial would refute Theorem 5.5.","supporting_citations":[{"cited_title":"Discriminantal bundles, arrangement groups, and subdirect products of free groups","cited_arxiv_id":"1008.0417","evidence_quote":"Supplies the injectivity criterion for retractive families (Theorem 3.4) and the finiteness-type result (Proposition 6.3) on which both main theorems rest."},{"cited_title":"Malcolm, final report for MAT 485, Undergraduate Research , Northern Arizona University, 2015","cited_arxiv_id":null,"evidence_quote":"Origin of Proposition 5.2, the commuting condition for transverse pairs of generators."},{"cited_title":"213–230, Ed","cited_arxiv_id":null,"evidence_quote":"Proves that the pure braid group on four strands is not residually free, delimiting the $K_4$-free hypothesis."},{"cited_title":"Fadell and L","cited_arxiv_id":null,"evidence_quote":"Gives the Fadell–Neuwirth bundle whose section underlies the semidirect product splitting that makes cliques retractive."},{"cited_title":"Meier, H","cited_arxiv_id":null,"evidence_quote":"Computes the $\\Sigma$-invariants of (right-angled) Artin groups used to derive the $FP_{m-1}$ but not $FP_m$ dichotomy."},{"cited_title":"Brown, Cohomology of groups, Springer Verlag, Berlin Heidelberg New York, 1982, MR0672956","cited_arxiv_id":null,"evidence_quote":"Provides the definition of type $F_m$ and the finite-index transfer used for the graphic braid group $B_\\Gamma$."},{"cited_title":"Brady and J","cited_arxiv_id":null,"evidence_quote":"Shows that braid groups on at most four strands act freely and properly on a CAT(0) complex, used in the almost-disjoint 4-clique extension."}],"review_version":1}