{"id":"8135866f-3ad2-4e62-855c-e67e56975dc0","arxiv_id":"2607.09010","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"Han administrative persistent entropy collapses to zero at 220 CE while geographic entropy rises to 3.065, with Wasserstein and ICT early warnings 45–50 years prior and Three Kingdoms as β1 cycles at 190 CE.","lead":"Persistent homology applied to Han Dynasty road and prefecture networks shows administrative connectivity collapsing while geographic connectivity rises after 184 CE, a signature of internal fragmentation. Early-warning topological indicators fire 45–50 years before the formal end of the dynasty.","discovery_kind":"new_application","skeptic_critique":{"model":"grok-4.5","headline":"Channel 5 encodes the known Three Kingdoms domains, so the positive HGeo slope and the three long-lived β1 cycles are largely model-imposed rather than recovered from topography alone.","rationale":"The Reader correctly isolates Channel 5 (and the related Hcomb calibration to H*) as the weakest assumption. The geospatial pipeline, LCP validation on the Hexi Corridor, and the Admin collapse itself are carefully constructed and do not require Channel 5. The early-warning chronology on HAdmin is also largely independent of it. What does require Channel 5 is the central narrative claim that TDA recovers the Three Kingdoms as long-lived β1 cycles and that the Admin–Geo divergence is the topological signature of internal fragmentation. Because the domain assignment and the minter schedule are taken from the known historical outcome, those particular results are confirmatory rather than predictive. An ablation that turns Channel 5 off (or replaces the historical domain map with a purely geographic or random partition) is the single concrete check that decides whether the strongest claim survives. Until that check is reported the CONDITIONAL verdict is appropriate; the paper remains publishable once the free historical multipliers are ablated or the early-warning suite is shown to fire on a held-out polity whose end-state was not used in model construction.","tokens_in":23646,"tokens_out":730,"duration_ms":7226,"concrete_test":"Recompute the full HGeo series, ΔH(d), W×(Admin,Geo), G3(d) and the top-3 β1 lifetimes for d ≥ 184 CE with Channel 5 disabled (minter = mintra = 1 for all edges) while retaining the extended node set of Eq. (20). If the positive phase-G slope of HGeo, the discontinuous W× jump, and a G3 gap ≥ 10\times all disappear or fall below the pre-184 background, the claimed topological signature of internal fragmentation is an artifact of the historically informed cost schedule rather than an emergent recovery.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The strongest claim is that ΔH \to +3.065, W× jumps to 737 at 190 CE, and three long-lived β1 cycles matching Wei/Shu/Wu appear thirty years before formal partition, constituting a topological signature of internal fragmentation. That signature is produced by Channel 5 (Eq. 17 and Table VII), which is applied only to HGeo after 184 CE and multiplies inter-domain edges by minter up to 25 while leaving intra-domain edges near baseline. The domain map Z(r,d) in Eq. (16) is taken from the San Guo Zhi and assigns nodes to Wei, Shu, Wu (and Neutral) using the historically known post-Yellow-Turban frontiers. Without those historically informed multipliers the two networks remain identical by construction until node-set divergence, and the paper supplies no ablation showing that the same three long-lived bars and the positive phase-G slope of HGeo still appear from LCP costs and node persistence alone. The early-warning suite (ICT, ṖW2, ξ) is less dependent on Channel 5, but the headline fragmentation signature and the “prediction” of the Three Kingdoms are.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.5","summary":"The paper applies persistent homology to three Han-era networks (administrative, geographic, and Silk Road trade) built from CHGIS v6 prefectures, a 1 km DEM with least-cost paths, and five historical friction channels. It reports that HAdmin collapses to zero by 220 CE (phase-G slope −0.0261 yr−1, R²=0.689) while HGeo rises to 3.065, producing ΔH=+3.065 as a signature of internal fragmentation; W×(Admin,Geo) jumps from 0 to 737 at 190 CE; three long-lived β1 cycles under a post-184 fragmentation model are identified with Wei, Shu, and Wu; and ICT, Ẇ2, and ξ give early-warning signals 45–50 years before formal dissolution. A combined index Hcomb is calibrated so that Hcomb(220)=H*=0.5241 from prior Roman–Byzantine work.","tokens_in":24045,"tokens_out":1239,"duration_ms":15858,"significance":"If the Admin–Geo divergence and early-warning chronology can be shown to be robust to the historically informed Channel 5 domain map, the paper would supply a quantitative, cross-civilizational signature distinguishing internal fragmentation from territorial loss, and a falsifiable early-warning suite independent of scalar entropy. Strengths include primary geospatial sources (CHGIS v6), explicit LCP validation on the Hexi Corridor, reported OLS/Chow/bootstrap statistics, and a clean three-network control design with a stable trade network. The work is a natural extension of the authors’ Roman–Byzantine series and would matter for historical complex-systems and TDA applications if circularity in the fragmentation model and H* calibration is resolved.","major_comments":[{"comment":"§III.C.4, Eq. (16)–(17) and Table VII: Channel 5 (minter up to 25 after 184 CE) and the domain map Z(r,d) are taken from the San Guo Zhi and assign nodes to Wei/Shu/Wu. The positive phase-G slope of HGeo (+0.0169 yr−1), the three long-lived β1 cycles, and the claim that they “match” Wei/Shu/Wu thirty years before partition are therefore largely imposed by the cost model rather than recovered from topography and node persistence alone. The manuscript needs an ablation with C5 off (or with domain-agnostic penalties) showing whether ΔH, the G3 gap (Table XVII), and the three long bars still appear; without it the headline “prediction” of the Three Kingdoms is not supported.","section":"§III.C.4 / §VI.E"},{"comment":"§V.D after Eq. (28) and Table XI: α0 is fixed by the requirement Hcomb(220)=H*=0.5241. The abstract and introduction frame the work as testing whether H*=0.5241 generalizes to a mechanistically distinct empire; calibration by construction cannot constitute that test. Either treat H* as an external fixed scale and report the uncalibrated Hcomb(220), or reframe Hcomb as a Han-specific resilience index and drop the generalization claim for the numerical threshold.","section":"§V.D / Abstract"},{"comment":"§VI.D–E and Table XXIV: W× jumps and ΔH open only after 184 CE because VGeo is extended and C5 is switched on by construction (Eqs. 19–20). The discontinuous W×=737 at d=190 CE is therefore partly definitional. Clarify which components of the early-warning suite (especially Admin ICT, Ẇ2, ξAdmin) are independent of C5 and node-set extension, and report those series with C5 disabled so the 45–50 year lead time is not confounded with the fragmentation model.","section":"§VI.D / Table XXIV"}],"minor_comments":[{"comment":"Abstract and §I claim the three β1 cycles emerge “thirty years before formal partition” under the post-184 model; the wording should state that the model encodes known domains rather than implying unsupervised recovery.","section":"Abstract / §I"},{"comment":"Table IX note: H(220) described as “numerically −0.000” is confusing; state H=0 exactly when β1=1 and p1=1.","section":"Table IX"},{"comment":"Fig. 6 and Table XXIII: bootstrap CIs are incomplete for several terminal decades; fill or mark as unavailable consistently.","section":"Fig. 6 / Table XXIII"},{"comment":"Notation: HAdmin / Hadm / H_Admin and HGeo / Hgeo appear interchangeably; standardize.","section":"Throughout"},{"comment":"§II.G validation is strong; a short sensitivity check on the 600 km neighbor cutoff and altitude penalties (Eq. 5) would help readers assess LCP robustness.","section":"§II.F–G"}],"recommendation":"major_revision","confidential_remarks":"The circularity around Channel 5 and the forced H* calibration are the load-bearing issues; both are fixable with ablations and reframing. Fit for physics.soc-ph is reasonable if the authors treat the work as historical complex-systems methodology rather than a parameter-free discovery of successor polities. Novelty relative to the authors’ own arXiv:2605.27200 and 2607.05695 should be stated more sharply in the cover letter."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The new empirical core is the Admin–Geo split after 184 CE: HAdmin falls to zero while HGeo rises to 3.065, W× jumps from 0 to 737 at 190 CE, and ICT / Ẇ2 / ξ fire 45–50 years early. That is a clean, quantitative signature of internal fragmentation, and it is not in the Roman–Byzantine papers. The CHGIS + 1 km DEM + Hexi Corridor LCP validation is careful; the three-network split and the stable Silk Road control are useful; OLS, Chow, and bootstrap reporting are present.\n\nThe soft spot is load-bearing and exactly where the stress-test puts it. Channel 5 (minter up to 25) and the domain map Z(r,d) are taken from the known Wei–Shu–Wu geography and applied only to HGeo after 184. Without that historically informed cost penalty the two networks stay identical by construction until node-set divergence, and the paper never shows the three long-lived β1 bars or the positive phase-G slope of HGeo from LCP costs and node persistence alone. Hcomb(220) is also forced to the prior H* = 0.5241 by solving for α0. So the headline “prediction” of the Three Kingdoms thirty years early is model-imposed, not an independent recovery. The early-warning suite is less dependent on Channel 5 and is the more defensible claim.\n\nMath and citation pattern look fine for a methods extension; free multipliers are numerous but mostly documented. No code or full bootstrap tables ship, which is a practical minus.\n\nThis is for people already working historical networks or TDA-on-graphs who want a second imperial case and a clear internal-vs-external contrast. It deserves a serious referee, not a desk reject, provided the authors are forced to ablate Channel 5 and to separate calibrated Hcomb from the uncalibrated Admin–Geo observables. I would read the revision; I would not cite the Three Kingdoms claim until the ablation exists.","headline":"Solid Han geospatial TDA with a real Admin–Geo divergence, but the Three Kingdoms β1 “prediction” is largely imposed by Channel 5 rather than recovered from topography.","tokens_in":24678,"tokens_out":527,"would_cite":false,"duration_ms":6849,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.5","headline":"Han administrative networks collapse to zero entropy while geographic networks reorganize into Three Kingdoms cycles thirty years before formal partition.","keywords":["persistent homology","imperial collapse","Han Dynasty","topological data analysis","early-warning indicators","network fragmentation","Silk Road","Wasserstein distance"],"falsifier":"Recompute the geographic network after 184 CE with Channel 5 turned off or with domain multipliers assigned at random rather than by the documented Wei–Shu–Wu frontiers; if the positive phase-G slope, the ΔH = 3.065 divergence, the W× jump, and the three long-lived β₁ cycles with gap ratio ~11 all vanish, the internal-fragmentation signature is an artifact of that historically informed cost penalty.","tokens_in":24504,"feed_emoji":"🗺️","tokens_out":1023,"duration_ms":24985,"temperature":0.7,"pith_summary":"This paper asks whether the same persistent-homology measures used on the Roman–Byzantine world also detect collapse in the Han Dynasty, a system that fragmented from within after the Yellow Turban Rebellion. The authors build three weighted networks—administrative prefectures, the same places kept alive after administrative dissolution, and a fixed Silk Road trade graph—whose edge costs come from least-cost paths on a real elevation model plus five historical friction channels. Administrative persistent entropy falls to zero by 220 CE while geographic entropy rises, producing a clean divergence that the paper calls the topological signature of internal fragmentation. Three long-lived one-dimensional cycles matching the Wei, Shu and Wu domains appear at 190 CE, and three independent early-warning indicators fire 45–50 years before formal dissolution. A reader cares because the method supplies a quantitative, mechanism-agnostic language for comparing how different empires fall apart and for spotting reorganization before the history books declare it.","feed_headline":"Topology flags Han collapse 45 years early","feed_subtitle":"Admin entropy hits zero while geography reorganizes into Wei–Shu–Wu cycles decades before partition.","key_machinery":"Three-network decomposition (H_Admin, H_Geo, H_Trade) measured by persistent entropy of the β₁ barcode, together with the Admin–Geo divergence ΔH and the cross-network Wasserstein distance W×. After 184 CE, Channel 5 raises the cost of inter-domain edges only in H_Geo, so geographic nodes persist while long-range loops reorganize into domain-scale cycles that carry most of the persistence weight.","core_discovery":"The topological signature of Han internal fragmentation is the Admin–Geo divergence: administrative persistent entropy collapses to zero at 220 CE (phase-G slope −0.0261 yr⁻¹) while geographic entropy simultaneously rises to 3.065, with the cross-network Wasserstein distance jumping from exactly zero to 737 at 190 CE. Under the post-184 fragmentation model, three long-lived β₁ cycles corresponding to the Wei, Shu and Wu domains emerge thirty years before formal partition, and three independent indicators (Wasserstein velocity, correlation length, Integrated Change Tracker) signal onset 45–50 years before dissolution while the Silk Road trade network stays stable.","pith_inferences":["Admin–Geo divergence may serve as a diagnostic for other internally fragmenting polities once comparable hierarchical settlement data exist.","The county-level β₀ atomization that precedes regional β₁ consolidation suggests a two-scale process testable on other empires with fine-grained place catalogues.","Stable trade topology amid political collapse implies commercial circuits can outlast the states that once secured them; the pattern could be checked on other long-distance historical trade systems.","If the early-warning suite is portable, archival geospatial datasets alone could generate collapse chronologies without first accepting narrative periodizations."],"forward_implications":["Internal fragmentation can be distinguished from simple territorial loss by opposite trends in administrative versus geographic persistent entropy.","Independent topological indicators can flag systemic risk 45–50 years before formal political dissolution.","Successor polities can appear as long-lived β₁ cycles decades before they are declared as states.","Long-distance trade topology can remain stable across political collapse, acting as a structural control.","The collapse threshold H* = 0.5241 and three-network method generalize from Roman–Byzantine to a mechanistically distinct imperial system."],"fun_headline_variants":["Admin-geo divergence marks Han fragmentation decades before partition","Wasserstein jump at 190 CE signals Han internal collapse onset","Three β1 cycles flag Wei-Shu-Wu domains 30 years early","Topology tracks Han admin entropy collapse as geography reorganizes","Early-warning indicators detect Han dissolution 45-50 years prior"],"cache_read_input_tokens":16512,"weakest_assumption_plain":"The rise in geographic entropy and the three kingdom cycles depend on a friction channel that multiplies the cost of routes crossing the historically known post-184 political domains, so the model already encodes the later map of Wei, Shu and Wu.","fun_headline_variants_meta":{"raw":{"variants":["Admin-geo divergence marks Han fragmentation decades before partition","Wasserstein jump at 190 CE signals Han internal collapse onset","Three β1 cycles flag Wei-Shu-Wu domains 30 years early","Topology tracks Han admin entropy collapse as geography reorganizes","Early-warning indicators detect Han dissolution 45-50 years prior"]},"model":"grok-4.5","effort":"low","cost_usd":0.006724,"raw_usage":{"total_tokens":1778,"prompt_tokens":986,"num_sources_used":0,"completion_tokens":89,"cost_in_usd_ticks":67240000,"prompt_tokens_details":{"text_tokens":986,"audio_tokens":0,"image_tokens":0,"cached_tokens":0},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":703,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":986,"tokens_out":89,"duration_ms":7100,"temperature":1.0,"reasoning_tokens":703,"cache_read_input_tokens":0,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-13T01:01:19.172481+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Recompute the geographic network after 184 CE with Channel 5 turned off or with domain multipliers assigned at random rather than by the documented Wei–Shu–Wu frontiers; if the positive phase-G slope, the ΔH = 3.065 divergence, the W× jump, and the three long-lived β₁ cycles with gap ratio ~11 all vanish, the internal-fragmentation signature is an artifact of that historically informed cost penalty.","supporting_citations":[],"review_version":1}