{"id":"f85f11a5-9539-49ca-8597-0281d5d2f37e","arxiv_id":"2508.15445","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"In adaptive hypergraphs, group-level rewiring driven by infection load raises epidemic thresholds and can change discontinuous outbreaks into continuous ones.","lead":"This paper models how groups (hyperedges) break apart and re-form as infections spread, and shows that such higher-order adaptivity can suppress explosive outbreaks and bistability, unlike simple pairwise adaptivity. The study provides a theoretical framework for co-evolving higher-order networks and spreading dynamics.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Routh–Hurwitz sign error: Eq. (7) does not follow from the Jacobian; predicted threshold is ~16% off for Fig. 3(a) parameters.","rationale":"Reading in good faith, the paper proposes a meaningful extension of adaptive network theory to hypergraphs and supports its qualitative claims with simulations on real and synthetic data. The reader's weakest_assumption—constant-E reformation—is a modeling limitation that is explicitly acknowledged and is not internally inconsistent; it does not itself undermine the central claim within the stated model. However, the derivation of Eq. (7) is more load-bearing. The Routh–Hurwitz table in the supplement has sign errors: the stated conditions would classify the disease-free state as unstable at β=0, which is impossible. The paper then selects a 'dominant inequality' by numerical comparison rather than proof, and the selected inequality differs from the exact linear-stability boundary for the printed Jacobian. For Fig. 3(a) parameters, the discrepancy is about 16%, which is large enough to matter for the threshold lines and the quantitative claims. This does not automatically falsify the qualitative conclusion that higher-order adaptivity raises thresholds and suppresses bistability, but it means the central theoretical result is not yet established as written. Since the reader already assigned CONDITIONAL, my concern sharpens that condition: the authors should correct the Routh–Hurwitz derivation or explicitly present Eq. (7) as an approximation validated only by simulation. I therefore keep the verdict at CONDITIONAL rather than moving it.","tokens_in":12738,"tokens_out":38197,"duration_ms":375615,"concrete_test":"Compute the eigenvalues of the 3x3 Jacobian in Eq. (S5) as a function of β for the Fig. 3(a) parameters (N=3000, E=6000, D=6, µ=0.2, r=0.1, v=3, h=0, λ=1) and find the smallest β where the largest real part crosses zero; compare with Eq. (7). Equivalently, deterministically integrate Eqs. (S1) from a single infected seed (ρ0=1/N) and locate the outbreak threshold. If the threshold matches the eigenvalue/determinant value (≈0.0195) rather than Eq. (7) (≈0.0226), the paper's threshold formula and the white lines in Fig. 3 need correction. Repeat with h=4 to check whether the qualitative claim about the transition type survives the corrected β_c.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The threshold formula Eq. (7) rests on the Routh–Hurwitz analysis in Supplementary Sec. IV.B, but that analysis is internally inconsistent. For the Jacobian block in Eq. (S5), the characteristic polynomial is P(s)=det(A-sI)=-s^3+tr(A)s^2-Ms+det(A). Stability requires the first column of the Routh array, (-1, tr, det/tr-M, det), to be all negative: tr<0, det<0, and det/tr-M<0. The supplement's inequalities (S8) instead use -det/b - y <0 with y=-M, i.e. M-det/tr<0, and -det<0 (det>0), the opposite signs. At β=0, det=(-µ-r)(-2µ-2^h r)(-3µ)<0, so their conditions already fail for the clearly stable disease-free state. The subsequent 'dominant inequality' (S9) is asserted by 'comparison with numerical solutions,' not derived from the Routh–Hurwitz conditions. Independently solving the exact determinant condition for the Fig. 3(a) parameters (D=6, µ=0.2, r=0.1, v=3, h=0) gives β_c≈0.0195, while Eq. (7) gives β_c≈0.0226, a ~16% discrepancy. Since Fig. 3 uses Eq. (7) for the threshold lines and the abstract's quantitative threshold claim relies on it, the central theoretical result is not closed. The qualitative direction may survive, but the stated derivation does not support Eq. (7).","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a co-evolution model of SIS-type spreading on hypergraphs in which hyperedges break at rate π(j)=r j^h and are immediately reformed, keeping the total number of hyperedges constant. For 3-uniform hypergraphs the authors derive a mean-field ODE system and, via linearization and a Routh-Hurwitz analysis, a closed-form outbreak threshold Eq. (7). They report that pairwise-like adaptivity (h=0) enlarges the bistable region and promotes discontinuous transitions, whereas higher-order adaptivity (h>0, together with selective rewiring λ) increases the threshold, shrinks or eliminates bistability, and shifts transitions to continuous. These claims are supported by Gillespie simulations on synthetic and real hypergraphs.","tokens_in":13128,"tokens_out":27737,"duration_ms":259165,"significance":"If correct, the central qualitative finding—that nonlinear (higher-order) dependence of hyperedge breaking on infection level reverses the effects of pairwise-like adaptivity—would be a valuable contribution to adaptive higher-order network theory. The paper offers a tractable mean-field framework, explicit formulas, and extensive simulations on multiple empirical hypergraphs. A notable strength is that the threshold formula contains no fitted parameters. However, the theoretical derivation of Eq. (7) contains a clear sign error in the Routh-Hurwitz conditions, so the quantitative threshold and the phase-diagram lines derived from it are not currently supported. The qualitative simulation results may survive a corrected analysis, but the theoretical framework as written is not reliable.","major_comments":[{"comment":"The Routh-Hurwitz conditions in Eq. (S8) have reversed signs. For the characteristic polynomial in Eq. (S6), stability of the 3x3 block requires tr<0, det<0, and det/tr - M <0 (equivalently M > det/tr). The paper instead requires the second and third inequalities of Eq. (S8) to be negative, which is the opposite. At β=0, det=(-μ-r)(-2μ-2hr)(-3μ)<0 for μ,r,h>0, so the disease-free state is stable, yet the paper's condition -det<0 would require det>0 and would deem it unstable. Thus Eq. (S9) and Eq. (7) do not follow from the Jacobian (S5). Independently solving det(A)=0 for the Fig. 3(a) parameters (N=3000, E=6000, μ=0.2, r=0.1, v=3, h=0) gives β_c≈0.018, not the value ≈0.016 from Eq. (7). The exact discrepancy may depend on parameters, but the derivation is invalid.","section":"Supplementary Sec. IV.B, Eq. (S9)"},{"comment":"The statement that the 'last formula of Eq. (S8) dominates the outbreak threshold' is not derived; it is asserted through calculation and comparison with numerical solutions of the same ODE system. This is not a substitute for a correct linear stability analysis, especially because the preceding Routh-Hurwitz inequalities are wrong. Since the phase diagrams in Fig. 3 use Eq. (7) as the outbreak threshold, the quantitative theoretical predictions of the paper are unsupported until the threshold is re-derived from the correct stability conditions.","section":"Supplementary Sec. IV.B, Eq. (S9)"},{"comment":"The model assumes that every broken hyperedge is immediately replaced by a new one, so the total number of hyperedges E is constant. This is explicitly introduced for tractability, and the threshold formula depends on <d>=3E/N being fixed. The central qualitative claims are only tested under this immediate-reformation rule; no robustness analysis is provided for delayed, probabilistic, or incomplete reformation, which would change the co-evolutionary dynamics qualitatively. The title and abstract imply broader applicability, but the demonstrated scope is narrower.","section":"Model section, constant-E assumption"}],"minor_comments":[{"comment":"'This session' should be 'This section' in both places.","section":"Supplementary Sec. V.A and V.B"},{"comment":"The typeset formula is ambiguous: it should be made clear that the denominator applies to the sum Ω + sqrt(...), i.e., β_c = (Ω + sqrt(...)) / (2^{v+2}<d>). The OCR/plain-text rendering currently suggests Ω + [sqrt(...)/denom].","section":"Main text, Eq. (7)"},{"comment":"Several references have corrupted characters (e.g., ref. [8] 'D¡ ¯Lima', ref. [10] 'E¡ ª'); the bibliography needs to be cleaned.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The Routh-Hurwitz sign error is unambiguous; I verified the determinant condition independently and found the same qualitative failure. The authors should be asked to re-derive the threshold from the correct stability conditions and to update Eqs. (7), (S9), (S10) and the corresponding phase diagrams. The qualitative message may survive, but the theoretical framework as written is not sound. This is a fixable issue within the scope of a revision, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth a serious look, but the threshold formula in Eq. (7) doesn't follow from the derivation as written. The Routh-Hurwitz step in the supplement has a sign inconsistency: condition (S8) effectively requires det>0 and M - det/tr < 0, while the disease-free state has det<0 and stability requires the opposite signs. So the claimed derivation of Eq. (S9) and Eq. (7) is not sound. The stress-test's ~16% mismatch for the Fig. 3(a) parameters is consistent with this.\n\nThat said, the paper does real work. The idea of higher-order adaptivity—power-law hyperedge breaking and fitness-based formation—is a natural and useful extension of pairwise adaptive networks, and as far as I know it's new. The qualitative claim that higher-order adaptivity shrinks or eliminates bistability and turns discontinuous transitions continuous, opposite to pairwise-like adaptivity, is supported by simulations on synthetic and real hypergraphs. The hyperedge-based mean-field framework is clearly laid out, and the threshold formula has no fitted parameters, which is a point in its favor even if the closed form is wrong.\n\nThe soft spots are the ones you'd expect. The constant-E assumption (broken hyperedge immediately replaced) is load-bearing for the mean-field closure; delay or failure of reformation would change the dynamics qualitatively. The \"dominant inequality\" step is asserted from numerical comparison, not proven. No code is provided, so the simulations are described but not independently runnable. And because theory and simulation come from the same model, the agreement is internal consistency rather than external validation. None of these flaws necessarily kills the qualitative conclusion, but they need to be addressed.\n\nThe paper is for network scientists studying adaptive higher-order systems. I'd send it to peer review, but the reviewer should be asked to check the stability analysis carefully and the authors should provide the actual derivation of Eq. (7) or a corrected formula. I wouldn't cite the threshold in my own work until that's fixed, but the concept and the simulation results are worth keeping in mind.","headline":"A promising new model of higher-order adaptivity whose qualitative story likely survives, but Eq. (7) is not supported by the supplied stability analysis.","tokens_in":13555,"tokens_out":3516,"would_cite":false,"duration_ms":35757,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper shows that when a group's breakup rate grows with the number of infected members, epidemic thresholds rise and abrupt discontinuous transitions become continuous, the opposite of what pairwise-like rewiring produces.","keywords":["higher-order adaptivity","hypergraphs","co-evolving networks","epidemic spreading","bistability","discontinuous phase transition","mean-field approximation","cusp points"],"falsifier":"Measure, in a real co-location or contact setting, how a group's breakup probability scales with the number of infected members j: if the scaling exponent is close to zero, the predicted suppression of bistability should disappear. Alternatively, run the same model with a delay before new hyperedges form; as delay grows, hyperedges are lost, and the epidemic threshold should deviate from Eq. (7) while fragmentation-induced abrupt transitions may reappear.","tokens_in":12642,"feed_emoji":"🦠","tokens_out":7479,"duration_ms":89412,"temperature":0.7,"pith_summary":"This paper argues that the way groups break apart during an epidemic can change the qualitative shape of an outbreak, not just its final size. Earlier adaptive-network models let every contact involving an infected person rewire at the same rate; the paper calls this pairwise-like adaptivity and shows it widens the region where extinction and outbreak coexist. It then introduces higher-order adaptivity, in which a group's breakup rate grows as r times j to the power h, where j is the number of infected members, and shows this raises the epidemic threshold but shrinks or removes the bistable region and turns discontinuous transitions into continuous ones. The practical upshot is that the same protective instinct, enacted with awareness of how many infected people are inside a group, can have the opposite dynamical effect from blanket avoidance. The claim is supported by an analytical threshold formula for 3-uniform hypergraphs and by simulations on synthetic and real human-contact hypergraphs.","feed_headline":"Breaking groups by infection count smooths epidemic transitions","feed_subtitle":"On hypergraphs, group breakup that grows with infected members raises thresholds and kills bistable regions.","key_machinery":"The load-bearing object is the power-law hyperedge breaking rate π(j) = r j^h. The exponent h is the switch between two regimes: h = 0 dissolves groups at a rate independent of infection count, reproducing pairwise-like adaptivity; h > 0 makes group breakup nonlinearly sensitive to the number of infected members, the defining feature of higher-order adaptivity. The supporting machinery is a hyperedge-based mean-field approximation that tracks l_{m,k}, the number of m-hyperedges containing k infected nodes, assuming the total number of hyperedges E stays constant. Linearizing the resulting ODE system at the disease-free state and applying the Routh–Hurwitz criterion yields the closed-form out","core_discovery":"The paper claims that higher-order adaptivity—group dissolution that accelerates nonlinearly with the number of infected members—produces qualitatively different epidemic behavior from pairwise-like adaptivity. In the model, an infected person in a group transmits at rate β times j^v, where j is the number of infected members, and a hyperedge containing j infected and at least one susceptible member breaks at rate r times j^h. When h = 0, this reduces to the usual adaptive-rewiring picture, which amplifies or induces bistability and discontinuous transitions. When h > 0, the same adaptivity raises the outbreak threshold but narrows or eliminates the bistable region and changes discontinuous","pith_inferences":["If the assumption of immediate hyperedge reformation is dropped, the total number of hyperedges becomes time-dependent; one would expect fragmentation to reintroduce abrupt extinction and make thresholds deviate from the paper's Eq. (7). This is testable by adding a reformation delay to the same agent-based model.","The same nonlinear-breakup rule may apply beyond disease: in opinion cascades, cooperation, or contagion on higher-order structures, a breakup rate that grows with the number of 'active' members could similarly suppress explosive transitions.","A direct empirical test would measure how a real group's breakup probability scales with the number of infected members; if the scaling exponent h is near zero in a given setting, the predicted smoothing effect should be absent there.","The threshold's independence from λ hints that investing in contact information may not delay the first outbreak, but may change whether the epidemic jumps or grows gradually—an intervention-relevant distinction."],"forward_implications":["Higher-order adaptivity raises the epidemic threshold, making outbreaks harder to trigger, for both nonlinear higher-order spreading and ordinary pairwise spreading.","The bistable region shrinks as h increases and can disappear completely, with cusp points marking where the number and type of phase transitions change.","Higher-order adaptivity can also eliminate bistability and discontinuous transitions induced by pairwise-like adaptivity alone, even when the nonlinear infection mechanism is removed (v = 1).","The fitness parameter λ does not affect the outbreak threshold but does shrink the bistable region and alter the type of phase transition, meaning better information about who is susceptible changes qualitative behavior without changing onset.","The theoretical results are consistent across synthetic uniform hypergraphs and three real human-co-location hypergraphs, so the qualitative contrast is not an artifact of one network structure."],"supporting_citations":[{"why":"Supplies the adaptive pairwise-network model and pair-approximation method that the hyperedge-based equations generalize, and defines the h = 0 baseline.","marker":"[8]"},{"why":"Grounds the nonlinear infection kernel θ(j) = βj^v as a recognized higher-order spreading mechanism.","marker":"[18]"},{"why":"Establishes that higher-order spreading alone can induce bistability and discontinuous transitions, the phenomena the paper studies under adaptivity.","marker":"[23]"},{"why":"Provides the fitness-based selection rule used for susceptible-node choice when a broken hyperedge is reformed.","marker":"[39]"},{"why":"Provides the Thiers13 high-school co-location dataset from which a real hypergraph is derived for validation.","marker":"[40]"},{"why":"Provides additional workplace and school co-location datasets used to test the findings on multiple real hypergraphs.","marker":"[41]"},{"why":"Supplies the numerical continuation tool used to locate cusp points and characterize changes in phase-transition type.","marker":"[43]"}],"fun_headline_variants":["Infection-scaled group breakup smooths epidemic phase transitions","Hyperedge breakup that scales with infected members kills bistable region","Infection-scaled hyperedge breakup eliminates epidemic bistability","When group breakup scales with infections, epidemic jumps fade"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"The derivation keeps the total number of hyperedges fixed by assuming a broken group is immediately replaced by a new group containing a susceptible node; if reformation is delayed, incomplete, or not anchored to susceptible nodes, the threshold formula and the phase-transition predictions no longer follow.","fun_headline_variants_meta":{"raw":{"variants":["Infection-scaled group breakup smooths epidemic phase transitions","Hyperedge breakup that scales with infected members kills bistable region","Infection-scaled hyperedge breakup eliminates epidemic bistability","When group breakup scales with infections, epidemic jumps fade"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000676,"raw_usage":{"total_tokens":2883,"prompt_tokens":685,"completion_tokens":2198,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":429,"completion_tokens_details":{"reasoning_tokens":2129}},"tokens_in":429,"tokens_out":2198,"duration_ms":20901,"temperature":1.0,"reasoning_tokens":2129,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T17:53:44.784762+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure, in a real co-location or contact setting, how a group's breakup probability scales with the number of infected members j: if the scaling exponent is close to zero, the predicted suppression of bistability should disappear. Alternatively, run the same model with a delay before new hyperedges form; as delay grows, hyperedges are lost, and the epidemic threshold should deviate from Eq. (7) while fragmentation-induced abrupt transitions may reappear.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the adaptive pairwise-network model and pair-approximation method that the hyperedge-based equations generalize, and defines the h = 0 baseline."},{"cited_title":"Boccaletti, P","cited_arxiv_id":null,"evidence_quote":"Grounds the nonlinear infection kernel θ(j) = βj^v as a recognized higher-order spreading mechanism."},{"cited_title":"Neuh¨ auser, A","cited_arxiv_id":null,"evidence_quote":"Establishes that higher-order spreading alone can induce bistability and discontinuous transitions, the phenomena the paper studies under adaptivity."},{"cited_title":"Pastor-Satorras, C","cited_arxiv_id":null,"evidence_quote":"Provides the fitness-based selection rule used for susceptible-node choice when a broken hyperedge is reformed."}],"review_version":1}