{"id":"56377534-dcc5-44b1-a15d-be683eb631c0","arxiv_id":"1908.04467","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For directed signed networks with two nonidentical topologies, sign-consistent graphs yield polarization iff their union is structurally balanced and neutralization otherwise, while sign-inconsistent graphs always yield neutralization.","lead":"This paper studies groups of agents with mixed first-order and second-order dynamics whose interactions are described by two different signed directed graphs. It shows when the group polarizes into two opposing camps and when it neutralizes to zero, depending on whether the two graphs agree on interaction signs and whether their union is structurally balanced.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1's iff classification is internally inconsistent as stated because polarization allows θ=0; balanced networks with zero conserved quantity also neutralize, and unbalanced networks also polarize.","rationale":"The reader's weakest_assumption is the strong-connectivity condition, but strong connectivity is an explicit and scoped hypothesis of the paper rather than a hidden gap; disconnected unions are outside the stated contribution. The more load-bearing problem is the θ≥0 definition, which makes Theorem 1's if-and-only-if statements logically false as written: in any balanced case with zero conserved quantity the network neutralizes, and in any unbalanced case it also polarizes with θ=0. The reader did notice this issue in the rationale, but did not elevate it to the weakest assumption. Because the fix is a one-line change (θ>0) and the underlying Lyapunov/M-matrix convergence analysis appears internally consistent, the appropriate verdict remains CONDITIONAL, matching the reader's conditional verdict. The delegated results from [27] and [30] are a secondary concern, but the definitional inconsistency is more direct and more clearly load-bearing for the stated central claim.","tokens_in":17132,"tokens_out":17681,"duration_ms":168665,"concrete_test":"Run the two-agent example of Theorem 1's structurally balanced case: Bc=Bd=[[0,1],[1,0]], k large enough (k>μ), initial states x(0)=[1,-1]^T, y(0)=[0,0]^T. The formula (22) with ν=[1/2,1/2], D=I gives ν^T D(x(0)+k^{-1}y(0))=0, hence the predicted limit is x(t)→0 and y(t)→0. Simulating (3) will confirm neutralization in a structurally balanced topology, directly falsifying Theorem 1.2 as worded. Then repeat the same initial data after redefining polarization to require θ>0 (or θ=|ν^T D(x(0)+k^{-1}y(0))|>0); the counterexample disappears and the classification is consistent.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is a crisp if-and-only-if classification (Theorem 1) resting on the definition of polarization in Section II: lim_t |x_i(t)| = θ and lim_t y_i(t) = 0 with θ ≥ 0. With θ ≥ 0, neutralization (x→0, y→0) is a special case of polarization (θ=0). This makes both directions of Theorem 1 false as written. In a structurally balanced network satisfying the hypotheses, the paper's own converged solution (22) is lim x(t) = [ν^T D(x(0)+k^{-1}y(0))] D 1_n. For any initial condition with ν^T D(x(0)+k^{-1}y(0)) = 0 (e.g., n=2 with Bc=Bd=[[0,1],[1,0]], D=I, ν=[1/2,1/2], x(0)=[1,-1]^T, y(0)=[0,0]^T), the limit is x→0 and y→0, i.e., neutralization occurs although G(B) is structurally balanced, contradicting statement 2). Conversely, in any structurally unbalanced case the Lyapunov analysis gives x→0, so by the θ≥0 definition the system also achieves polarization, contradicting statement 1). The 'only if' implications are therefore false under the stated definition. Replacing θ ≥ 0 by θ > 0 restores consistency and matches the intended bipartite-consensus notion; the Lyapunov/M-matrix convergence analysis itself is not affected. This is a formal defect in the headline theorem, not a numerical or experimental issue.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies convergence of directed signed networks with mixed first- and second-order dynamics, where the communication topology is described by two signed digraphs G(Bc) and G(Bd) that may be sign-consistent or sign-inconsistent and have a strongly connected union. The main results are Theorem 1, which claims that under sign-consistent topologies and a sufficiently large damping gain k, polarization occurs if and only if the union graph is structurally balanced and neutralization occurs if and only if it is structurally unbalanced, together with an explicit limit formula for the polarized state; and Theorem 2, which claims that under sign-inconsistent topologies and sufficiently large k, neutralization always occurs. The proofs use a nonsingular transformation, a Lyapunov function with an M-matrix approach, and matrix properties delegated to prior work by the same authors.","tokens_in":17402,"tokens_out":5124,"duration_ms":55356,"significance":"If the classification were correct as stated, the paper would provide a complete convergence characterization for a broad class of mixed-order directed signed networks with nonidentical topologies, going beyond identical-topology results. The introduction of sign-consistency as a structural property of pairs of signed digraphs is a useful concept, and the explicit converged solution in (22) as well as the M-matrix treatment of sign-inconsistent topologies are valuable contributions. The paper also gives falsifiable predictions: the sign pattern of the two topologies determines whether the network polarizes or neutralizes, with explicit lower bounds on the damping gain. These strengths make the underlying analysis worth considering, but the validity of the headline theorem is compromised by the definitional issue discussed below.","major_comments":[{"comment":"The iff classification in Theorem 1 is false under the stated definitions. Polarization is defined by lim_{t→∞} |x_i(t)| = θ and lim_{t→∞} y_i(t) = 0 with θ ≥ 0. Since neutralization requires lim_{t→∞} x_i(t) = 0 and lim_{t→∞} y_i(t) = 0, every neutralized trajectory also satisfies the polarization definition with θ = 0. Consequently, in any structurally unbalanced case where Theorem 1(2) and Case ii) of Appendix A establish neutralization, polarization is also achieved, contradicting the 'only if' direction of Theorem 1(1). Conversely, in a structurally balanced case, the converged solution (22) gives lim_{t→∞} x(t) = {ν^T D[x(0)+k^{-1}y(0)]} D 1_n. For initial conditions with ν^T D[x(0)+k^{-1}y(0)] = 0, the limit is zero, so neutralization is achieved despite structural balance, contradicting the 'only if' direction of Theorem 1(2). A concrete example is n = 2, Bc = Bd = [[0,1],[1,0]], D = I, ν = [1/2,1/2], x(0) = [1,-1]^T, y(0) = [0,0]^T. To repair the theorem, polarization should be defined with θ > 0 (excluding the zero limit), and the statements must be qualified to hold for generic initial conditions or for initial conditions with a nonzero conserved quantity, since the zero-conserved-quantity case inevitably produces neutralization even in balanced networks.","section":"Section II (definition of polarization) and Theorem 1"},{"comment":"The proof of the necessity directions is not adequate. The text states only that 'the necessity results of this theorem follow directly by the mutually exclusive relationship between the structural balance and unbalance of G(B)'. This is not a valid proof of the two 'if and only if' claims, especially because the definitions of polarization and neutralization overlap when θ = 0. After the definitional issue is fixed, the necessity arguments must be supplied explicitly: for balanced graphs one must rule out neutralization (except for the zero-measure initial-condition set), and for unbalanced graphs one must rule out nontrivial polarization.","section":"Appendix A, Proof of Theorem 1"},{"comment":"The proof of Lemma 5 relies on the positive stability of Ξ = I - (L_{Bc+ + Bd+} + Δ|Bc- + Bd-|)^{-1}(Bc- + Bd-), which is asserted to follow from '[27, Lemma 4.2 and Corollary 4.2]' without reproducing the argument. Since this property is load-bearing for the conclusion that L_{Bc} + L_{Bd} is positive stable and hence for Theorem 2, the proof should either state the exact result from [27] or give a self-contained derivation. The same applies to the existence of W and H in Lemma 3, which is delegated to [27, Theorems 4.1 and 4.2].","section":"Lemma 5 and its proof in Appendix B"}],"minor_comments":[{"comment":"The thresholds μ and ζ are defined for any δ > 1, so they are not single constants but families of constants indexed by δ. The theorems state 'let k > μ be selected for any δ > 1', which is logically ambiguous; it would be clearer to say 'for any δ > 1, let k > μ(δ)' or to fix δ first.","section":"Section III, Eq. (7) and (12)"},{"comment":"The vector ν in the converged solution (22) is described only as satisfying ν^T(DL_BD) = 0 and ν^T 1_n = 1. It would be helpful to state explicitly that ν is the normalized left null vector of the Laplacian L_{|B|}, whose existence and uniqueness follow from strong connectivity.","section":"Theorem 1, limit formula"},{"comment":"The paper does not include numerical simulations. While not required for a theoretical result, a small simulation illustrating the sign-consistent balanced case, the sign-consistent unbalanced case, and the sign-inconsistent case would help the reader see the practical meaning of the thresholds μ and ζ.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The paper's core Lyapunov/M-matrix analysis appears sound and the explicit convergence formulas are valuable, but the theorem statements as written are formally false because of the θ ≥ 0 definition of polarization. This is not a deep technical flaw in the stability analysis, but it requires a nontrivial reformulation of the main claims, including a genericity qualification for initial conditions in the balanced case. I recommend major revision rather than rejection because the underlying convergence results are likely salvageable with corrected definitions and qualified statements."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe paper extends earlier work on signed networks to directed graphs with nonidentical topologies and mixed-order dynamics. The genuinely new pieces are the sign-inconsistent case and the M-matrix route to proving neutralization. The thresholds μ and ζ are derived from Lyapunov equations, not tuned to reach the conclusion. That part looks honest and mostly sound.\n\nThe problem is the stated if-and-only-if in Theorem 1. Polarization is defined with θ ≥ 0, so neutralization (x→0, y→0) is a special case of polarization with θ=0. That makes both directions of the iff false as written. For a structurally unbalanced network satisfying the hypotheses, the Lyapunov argument gives x→0, so the system also 'polarizes' by the definition. For a balanced network with initial conditions such that ν^T D(x(0)+k^{-1}y(0))=0, the paper's own limit formula gives x→0 and y→0, so neutralization occurs even though the union is balanced. The stress-test example with n=2 and Bc=Bd=[[0,1],[1,0]] works. Replacing θ ≥ 0 by θ > 0 fixes the classification and matches the intended bipartite-consensus notion. The convergence analysis itself is unaffected; it is the statement and the one-sentence necessity argument that need repair.\n\nOther soft spots are minor. Lemma 3 and the positivity of Ξ are delegated to earlier papers by the same group, which makes verification slower but is not fatal. The strong-connectivity assumption on the union is stated but the disconnected case is never discussed; the classification likely fails component-wise there. The paper is also not fully self-contained, but the delegation is to prior published work.\n\nFor the right reader — someone working on signed networks or heterogeneous multi-agent systems — this is a useful extension and the sign-inconsistent neutralization result is a real addition. It deserves a serious referee, but the referee should ask for the definition fix and a proper proof of the necessity direction before acceptance.\n\nRecommendation: send to peer review, but flag the θ≥0 issue explicitly.","headline":"The main convergence classification is plausible and useful, but as stated the headline theorem is formally false because the paper's own definition of polarization includes neutralization as a special case.","tokens_in":17973,"tokens_out":3805,"would_cite":false,"duration_ms":37336,"reading_group":"maybe","serious_thinker":"unclear","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["93D05","93A14","05C50","34D20"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that directed signed networks with two nonidentical interaction graphs are fully classified by sign-consistency: if the two graphs agree on signs, polarization or neutralization is decided by structural balance of their…","keywords":["signed networks","nonidentical topologies","structural balance","polarization","neutralization","directed signed digraphs","mixed-order dynamics","M-matrix method"],"falsifier":"Simulate system (2) on a concrete instance of two sign-consistent signed digraphs whose union is strongly connected and structurally balanced, choose $k>\\mu$ from (7), and let the positions run; if the final $|x_i|$ do not all approach one common value $\\theta$ while velocities decay, Theorem 1 fails. For the sign-inconsistent case, simulate any pair with strongly connected union and $k>\\zeta$; observing persistent velocities or positions that do not decay to zero would refute Theorem 2.","tokens_in":16876,"feed_emoji":"⚖️","tokens_out":9064,"duration_ms":83615,"temperature":0.7,"pith_summary":"This paper studies directed networks in which some agents follow first-order dynamics and others second-order, and in which each interaction is cooperative or antagonistic, so the topology is a signed digraph. The new element is that the agents use two different signed digraphs at the same time, one for position coupling and one for velocity coupling, rather than a single topology. The paper claims a complete classification: when the two signed digraphs are sign-consistent, the agents polarize into two opposing groups exactly when the union of the two digraphs is structurally balanced, and they neutralize to zero exactly when the union is structurally unbalanced. When the two digraphs are sign-inconsistent, the paper claims the network always neutralizes, no matter the structure. The theorems matter because they turn an apparently intricate mixed-order, two-topology problem into a single graph-theoretic condition, and they give explicit damping thresholds that guarantee the predicted outcome.","feed_headline":"One condition decides whether signed networks polarize or die out","feed_subtitle":"When the two communication graphs agree in sign, structural balance decides; when they clash, the network always dies out.","key_machinery":"The machinery is the Laplacian algebra of signed digraphs plus a block-coordinate Lyapunov analysis. The nonsingular transformation (4) rewrites the mixed-order system as a block matrix whose diagonal blocks involve $kI+L_{B_c}$ and whose coupling involves $L_{B_c}+L_{B_d}$. The sign-consistency property (Definition 1) ensures that $L_{B_c}+L_{B_d}=L_{B_c+B_d}$, so the union of the two topologies has a genuine Laplacian; structural balance then gives a diagonal sign matrix $D$ with $DL_B D=L_{|B|}$. For a strongly connected union, Lemma 3 supplies positive-definite Lyapunov matrices $W$ (balanced case) or $H$ (unbalanced case), and the Schur complement argument turns the requirement of a negative-definite derivative into the explicit gain bound $k>\\mu$. For sign-inconsistent topologies that Laplacian identity fails, so the paper factorizes $L_{B_c}+L_{B_d}$ from (11) as the product of an $M$-matrix and a nonnegative matrix, proves positive stability from Gershgorin circles and positivity of the determinant, and runs a second Lyapunov argument with threshold $k>\\zeta$.","core_discovery":"The central discovery is that the long-run behavior of system (3) is governed by one new property, sign-consistency, together with the classical notion of structural balance. Theorem 1 considers sign-consistent topologies: with a strongly connected union and any damping gain $k>\\mu$ (where $\\mu$ is defined by (7)), polarization occurs if and only if the union graph $G(B_c+B_d)$ is structurally balanced, with the explicit limit $\\lim_{t\\to\\infty} x(t)=\\{\\nu^T D[x(0)+k^{-1}y(0)]\\}D1_n$ and $\\lim_{t\\to\\infty} y(t)=0$; neutralization occurs if and only if the union is structurally unbalanced. Theorem 2 considers sign-inconsistent topologies: with a strongly connected union and any damping gain $k>\\zeta$ (defined by (12)), neutralization is always achieved. Together the two theorems assert that sign-inconsistency is not a nuisance to be smoothed over but a decisive feature: it forces collapse to zero, whereas sign-consistency lets the network read its fate from the structural balance of the union digraph.","pith_inferences":["A natural untested extension is whether the dichotomy survives for arbitrary small $k>0$; the paper proves the threshold only for the specific Lyapunov constructions, so there may be sign-consistent pairs where intermediate gains cause oscillatory or even divergent behavior outside the theorem's remit.","The disconnected-union case is left open; when the union splits into several strongly connected components, one would expect the polarization-iff-balanced classification to hold component-wise after removing the extra zero eigenvalues, but the paper does not establish this.","The M-matrix factorization (11) is likely to generalize to more than two nonidentical topologies or to switching topologies, suggesting that sign-inconsistency's collapse mechanism is a robust phenomenon rather than an artifact of the two-graph setup.","Because the sign-consistency condition is stated entrywise, a practical check is cheap: inspect every pair of nonzero weights in $B_c$ and $B_d$; if any pair has opposite signs, Theorem 2 applies and predicts neutralization."],"forward_implications":["A single scalar test, structural balance of the union digraph, completely decides polarized versus neutralized behavior for sign-consistent nonidentical topologies, for any large enough damping gain.","Sign-inconsistent pairs with a strongly connected union cannot polarize: for sufficiently large $k$ every trajectory decays to zero regardless of the underlying signed structure.","When polarization occurs, the final configuration is explicit: agents split according to the balanced partition $D1_n$, with the common magnitude determined by $\\nu^T D[x(0)+k^{-1}y(0)]$.","Because single-integrator and double-integrator signed networks are the two extremes of the mixed-order model (2), the two theorems bridge those previously separate settings through the nonidentical-topology lens.","The explicit thresholds $\\mu$ and $\\zeta$ convert the qualitative classification into a checkable design rule: choose $k$ above the threshold and the claimed convergence is guaranteed."],"supporting_citations":[{"why":"Defines signed-digraph structural balance and the Laplacian matrix that the paper builds on.","marker":"[12]"},{"why":"Gives the second-order baseline case $B_c=0$ whose convergence result the paper extends and replaces for nonidentical topologies.","marker":"[18]"},{"why":"Provides the interval bipartite consensus framework and the eigenvector limit computation used for the explicit polarization value.","marker":"[25]"},{"why":"Supplies the M-matrix and Lyapunov lemmas from which Lemma 3 and the stability arguments are drawn.","marker":"[27]"},{"why":"Introduces sign-consistency for nonidentical signed graphs in the undirected setting; this paper extends it to directed graphs and to sign-inconsistent cases.","marker":"[30]"},{"why":"Gives the M-matrix definition and the equivalence between positive stability and the nonnegative inverse property used in Lemmas 4 and 5.","marker":"[31]"},{"why":"Supplies the Schur complement lemma and simultaneous diagonalization used to derive the gain thresholds $k>\\mu$ and $k>\\zeta$.","marker":"[35]"}],"fun_headline_variants":["Sign-consistency decides: polarized or neutralized in signed networks","When union is balanced, networks polarize; else they die out","Sign-consistent topologies: balance yields polarization, imbalance neutralization","Sign-inconsistent graphs always force signed networks to zero","Mixed-order agents: sign clash mandates neutralization, harmony permits polarization"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The classification requires the union of the two signed digraphs to be strongly connected, meaning every agent influences every other through at least one of the two topologies; if the union splits into disconnected pieces, the stated if-and-only-if results are not proven and may fail piecewise.","fun_headline_variants_meta":{"raw":{"variants":["Sign-consistency decides: polarized or neutralized in signed networks","When union is balanced, networks polarize; else they die out","Sign-consistent topologies: balance yields polarization, imbalance neutralization","Sign-inconsistent graphs always force signed networks to zero","Mixed-order agents: sign clash mandates neutralization, harmony permits polarization"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001054,"raw_usage":{"total_tokens":4402,"prompt_tokens":897,"completion_tokens":3505,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":513,"completion_tokens_details":{"reasoning_tokens":3417}},"tokens_in":513,"tokens_out":3505,"duration_ms":27766,"temperature":1.0,"reasoning_tokens":3417,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:43:20.307371+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate system (2) on a concrete instance of two sign-consistent signed digraphs whose union is strongly connected and structurally balanced, choose $k>\\mu$ from (7), and let the positions run; if the final $|x_i|$ do not all approach one common value $\\theta$ while velocities decay, Theorem 1 fails. For the sign-inconsistent case, simulate any pair with strongly connected union and $k>\\zeta$; observing persistent velocities or positions that do not decay to zero would refute Theorem 2.","supporting_citations":[{"cited_title":"Consensus problems on networks with antag onistic inter- actions,","cited_arxiv_id":null,"evidence_quote":"Defines signed-digraph structural balance and the Laplacian matrix that the paper builds on."},{"cited_title":"Dynamic distributed control for networks wit h cooperative- antagonistic interactions,","cited_arxiv_id":null,"evidence_quote":"Gives the second-order baseline case $B_c=0$ whose convergence result the paper extends and replaces for nonidentical topologies."},{"cited_title":"Interval bipartite consensu s of networked agents associated with signed digraphs,","cited_arxiv_id":null,"evidence_quote":"Provides the interval bipartite consensus framework and the eigenvector limit computation used for the explicit polarization value."},{"cited_title":"Convergence analysis of directed signed netw orks via an M - matrix approach,","cited_arxiv_id":null,"evidence_quote":"Supplies the M-matrix and Lyapunov lemmas from which Lemma 3 and the stability arguments are drawn."},{"cited_title":"Bipartite consensus for second-o rder multi- agent systems over nonidentical signed graphs,","cited_arxiv_id":null,"evidence_quote":"Introduces sign-consistency for nonidentical signed graphs in the undirected setting; this paper extends it to directed graphs and to sign-inconsistent cases."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the M-matrix definition and the equivalence between positive stability and the nonnegative inverse property used in Lemmas 4 and 5."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Schur complement lemma and simultaneous diagonalization used to derive the gain thresholds $k>\\mu$ and $k>\\zeta$."}],"review_version":1}