{"id":"136549c0-37cb-4bef-a1eb-4df06c638c31","arxiv_id":"2412.15344","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A length-weighted, boundary-conditioned betweenness centrality predicts which struts carry the most stress in 2D lattices better than standard centrality.","lead":"The paper introduces a graph-based measure that accounts for strut length and loading direction, and shows it pinpoints stressed struts in two-dimensional lattices more accurately than standard graph measures. This could offer a fast, low-cost way to predict stress hotspots in lightweight structures and mechanical metamaterials.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"ROC evidence for EBC_LB superiority is under-specified: no AUC values, no ground-truth label rule, and no ablation separating length weighting from boundary restriction.","rationale":"We read the paper as a heuristic proposal: EBC_LB is a graph-theoretic proxy, not a constitutive model. The strongest claim is empirical superiority. Therefore the decisive evidence is the ROC comparison in Figure 4. The absence of AUC values and a ground-truth definition directly undermines the claim, and the conflation of length weighting with boundary restriction makes the attribution to geometry unverified. The reader's weakest_assumption correctly notes that rigid-jointed experiments differ from the pin-jointed motivation (P ~ EA/l), but we see the missing quantitative validation as the more immediate load-bearing gap: even if the pin-jointed idealization is accepted, the reported evidence would not quantitatively establish the predictive advantage. Our proposed check—recomputing ROC with explicit labels, ablating to EBC_B, and permuting edge lengths—would settle both the validation and the attribution. We therefore keep the conditional verdict unchanged.","tokens_in":9314,"tokens_out":5742,"duration_ms":53808,"concrete_test":"Re-run the Figure 4 analysis using the released StructuralGT package and the original birefringence/FEA data. (1) Define and publish the ground-truth stressed-edge label explicitly (e.g., strut whose normalized birefringence intensity or FEA axial-plus-bending stress is above the 75th percentile). (2) Report AUC with 95% bootstrap CIs for EBC_LB, EBC_G, and a boundary-only unweighted version EBC_B for all four lattices. (3) Repeat with edge lengths randomly permuted while preserving topology. If EBC_LB significantly exceeds both EBC_G and EBC_B in all four cases, the central claim is supported; if EBC_LB approximately equals EBC_B, the claimed geometric improvement is not evidenced.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that EBC_LB outperforms EBC_G as a stress predictor rests on the ROC comparisons in Figure 4, yet the paper never reports AUC values, confidence intervals, or the binary ground-truth rule used to score each strut as 'stressed.' The only definition offered is that stressed edges are predicted when EBC_LB > average, but the ROC procedure requires a ground-truth label from birefringence/FEA, and no threshold (e.g., brightness percentile, FEA stress percentile) is stated. Without this, the curves in Figure 4a3-d3 cannot be reproduced or quantitatively compared. Additionally, EBC_LB (Eq. 2) differs from EBC_G (Eq. 1) in two coupled ways: the source-target set is restricted to boundary nodes, and paths are length-weighted. The ROC comparison conflates these changes, so the paper's stated conclusion that geometry (length weighting) matters is not actually isolated. The kite FEA example (Figure 3b) is the only length-perturbation test and yields only a step-function match, not quantitative agreement; the Discussion further concedes that regular lattices are sensitive to load-uniformity assumptions. Therefore the empirical support for the headline claim is currently under-specified.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes a graph-theoretic descriptor, length-weighted edge betweenness centrality restricted to source nodes on the loaded boundary and target nodes on the opposite boundary (EBC_LB, Eq. 2), as a fast, parameter-free predictor of which struts carry high stress in 2D strut lattices. The claim is validated by comparing EBC_LB with standard geodesic edge betweenness centrality (EBC_G, Eq. 1) against birefringence imaging and finite element analysis for four lattice families, including stochastic, auxetic, Archimedean, and rhombic networks. The paper also discusses the continuity problem for graph representations of lattices with gradually varying geometry, and reports computational speedups from restricting the source-target set to boundary nodes.","tokens_in":9544,"tokens_out":3406,"duration_ms":33219,"significance":"If the central claim holds, EBC_LB would be a practically valuable heuristic for stress-hotspot identification in architected materials, complementing expensive finite element simulations. The paper's strengths are its combination of experimental birefringence data, finite element analysis, graph-theoretic modeling, and an open-source software package (StructuralGT), together with a concrete computational-scaling argument. However, the quantitative support for the headline claim is currently under-specified, and the comparison conflates two separate modifications to the centrality measure, so the significance is conditional on the requested revisions.","major_comments":[{"comment":"The ROC comparison in Figure 4 is the central quantitative evidence for the claim that EBC_LB outperforms EBC_G, but the manuscript reports no AUC values, no confidence intervals, and no binary ground-truth labeling rule. The text only states that 'the greater the area under the curve, the better the performance' and that EBC_LB outperforms EBC_G in all cases. Without a stated threshold for converting birefringence or FEA stress maps into a stressed/not-stressed label for each edge, the ROC curves cannot be reproduced or quantitatively compared. Please report the exact binarization rule (e.g., brightness percentile or FEA stress percentile), the AUC values with uncertainties or confidence intervals, sample sizes, and a paired significance test for the four networks.","section":"Predictions of stressed edges, Figure 4 (a3-d3)"},{"comment":"The comparison between EBC_G and EBC_LB conflates two distinct modifications: length weighting of shortest paths and restriction of source-target pairs to boundary nodes. The text concludes that 'the inclusion of the boundary conditions and length weighting' jointly improve prediction, but it never isolates the contribution of each ingredient. An ablation is needed: report ROC comparisons for EBC_L (length-weighted, all source-target pairs) and EBC_B (boundary-restricted, unweighted) separately. Without this, the paper cannot support the specific claim that geometric length weighting, as opposed to merely restricting paths to boundary nodes, improves stress prediction.","section":"Eq. (1) vs Eq. (2), Relating stress and structure for strut lattices"},{"comment":"The theoretical motivation is based on pin-jointed axial force scaling P ~ EA/l, but the experiments and FEA use rigid-jointed laser-cut lattices, where bending moments contribute. Figure 3b shows only a step-function approximation to the FEA stress ratio Sr/Sl, not quantitative agreement. As written, this validates only the qualitative direction of the length effect. Please clarify which stress measure EBC_LB is intended to predict (axial stress, von Mises stress, or the principal-stress difference that produces birefringence), and provide a quantitative error metric for the kite example and at least one full lattice, rather than visual ROC curves alone.","section":"Figure 3, kite example and mechanism discussion"},{"comment":"The paper diagnoses the continuity problem clearly, showing in Figures 1c-e that graph representations change discontinuously when lattice angle or strut thickness changes and nodes are added or removed. However, length weighting does not remove this discontinuity, because adding or removing nodes changes both topology and path lengths. The manuscript does not provide a quantitative measure of continuity or a test showing that EBC_LB varies smoothly under continuous geometric perturbation. Since the abstract states that the continuity challenge is addressed, this claim needs explicit support, such as a plot of a chosen centrality parameter versus θ for the rhombic-lattice family.","section":"Introduction and Figure 1 (c-e), Discussion on continuity"}],"minor_comments":[{"comment":"The caption says 'Stresses, Sl and Sl represent the stresses experienced by the left and right members,' but the text correctly refers to Sl and Sr; the second 'Sl' should be 'Sr'.","section":"Figure 3 caption"},{"comment":"There is a typo: 'Regular struts latices are ubiquitous' should be 'Regular strut lattices are ubiquitous'.","section":"Introduction, first sentence"},{"comment":"The notation for the shortest-path count is inconsistent: the text defines σ_st as the number of shortest paths, but Eq. (1) writes σ_st(e)/σ_st without clearly distinguishing the numerator and denominator; please make the notation uniform and define all symbols explicitly in one place.","section":"Eq. (1) and surrounding text"},{"comment":"The Methods mention StructuralGT and the SI, but there is no data availability statement for the birefringence images, FEA models, or computed centrality values. Given that the ROC analysis is central to the claims, providing these data or a repository link is important for reproducibility.","section":"Data and code availability"}],"recommendation":"major_revision","confidential_remarks":"The paper is within scope for a materials-science journal and the proposed centrality measure is potentially useful, but the central empirical claim is not yet quantitatively supported. The ROC evidence should be made complete with AUC values, ground-truth labeling rules, and confidence intervals. I would also encourage the editor to verify that the supplementary information contains the promised details on the spring-network equivalence and the analytical solution, since those are cited as supporting the main text's mechanistic claims."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's your quick take. The paper proposes a boundary-restricted, length-weighted edge betweenness centrality (EBC_LB) for predicting stressed struts in 2D lattices. That specific combination is new. Prior work used standard betweenness on all node pairs; this paper restricts sources/targets to loaded boundaries and weights paths by edge length. It also shows that the restricted definition is cheaper to compute, scaling as O(n^1.5 log n) vs O(n^2). The idea is sensible, and the small kite FEA example gives a physical motivation: longer struts carry less stress, which length weighting captures qualitatively.\n\nThe experimental validation is under-specified. The main quantitative evidence is the ROC curves in Figure 4, claiming EBC_LB outperforms EBC_G in all four lattices. The paper never reports AUC values, confidence intervals, or the exact rule used to convert birefringence intensities into binary stressed-edge labels. Without that, the curves cannot be reproduced or critically compared. The stress-test note is on target. It also correctly flags that the comparison conflates two changes: length weighting and boundary restriction. You cannot tell which ingredient matters from these figures. The kite test isolates length weighting but only yields a step-function match, not quantitative agreement, and the Discussion admits regular lattices are sensitive to load-uniformity assumptions. So the central claim is plausible but not yet nailed down.\n\nThe continuity problem is only partially addressed. Length weighting does reduce sensitivity to node placement, but the paper does not demonstrate a formal continuity guarantee, just an example. That's okay for a heuristic, but the abstract claims to \"address\" it, which overshoots what is shown.\n\nWhat the paper does well: the continuity discussion is thoughtful, and the use of an open-source package (StructuralGT) for graph extraction is a plus. The authors state limitations honestly. The computational scaling advantage is real and useful.\n\nWhere I'd push: give AUC numbers with error bars, state the ground-truth threshold, and run an ablation that tests length weighting alone and boundary restriction alone. That would turn a conditional result into a solid one.\n\nBottom line: this paper deserves a serious referee, not a desk reject. It introduces a useful variant and backs it with real experiments and FEA, but the evidence for the headline claim needs to be quantified. I'd bring it to a reading group if you care about network mechanics or heuristic stress prediction. I'd cite it as the source for boundary-conditioned centrality, though I'd like the ROC statistics in place first.","headline":"A genuinely new centrality variant for strut lattices, with under-reported ROC evidence — worth a close look, but the quantitative claims need tightening before publication.","tokens_in":10063,"tokens_out":3065,"would_cite":true,"duration_ms":27023,"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":"A boundary-aware, length-weighted betweenness centrality predicts stress-bearing struts in 2D lattices more accurately than standard graph centrality.","keywords":["graph theory","edge betweenness centrality","strut lattices","stress prediction","birefringence","finite element analysis","mechanical metamaterials","boundary conditions"],"falsifier":"Perform finite element analysis on a rigid-jointed lattice whose struts have strongly varying thickness while keeping lengths comparable, and compare measured axial stresses to EBC_LB rankings; a short thick strut carrying high stress while ranked low by EBC_LB would falsify the claim that length-weighted path counts alone predict stress.","tokens_in":9122,"feed_emoji":"📐","tokens_out":6287,"duration_ms":54695,"temperature":0.7,"pith_summary":"The paper argues that standard graph-theoretic descriptors miss two features that dominate stress distribution in strut lattices: strut length and the direction of loading. Its central proposal is a modified edge-betweenness centrality, EBC_LB, that counts only length-weighted shortest paths between nodes on the loaded boundary and the opposite boundary, and then flags struts whose EBC_LB exceeds the average. Across four lattice types, ordered rhombic and Archimedean lattices, an auxetic network, and a nanofiber-derived network, this parameter outperforms geodesic EBC_G in predicting which struts actually light up under birefringence imaging, with the largest gains in the most ordered, anisotropic structures. The authors also identify a continuity problem: small changes in intersection angle or strut thickness can force a different graph abstraction, and length weighting restores a continuous dependence on geometry. They note the method targets high-void-fraction lattices, and that regular lattices can be sensitive to assumptions about load uniformity.","feed_headline":"Boundary-aware betweenness metric pinpoints stressed struts","feed_subtitle":"Counting length-weighted shortest paths from loaded to opposite edge beats standard centrality on four lattice types.","key_machinery":"The load-bearing object is the length-weighted edge boundary betweenness centrality, defined as EBC_LB = (1/(2T(S-1))) sum over s in S, t in T of $\\sigma$^w_st(e)/$\\sigma$^w_st, where S and T are the nodes touching the loaded and opposite boundaries, $\\sigma$^w_st is the number of Euclidean-length-weighted shortest paths between s and t, and $\\sigma$^w_st(e) is the subset passing through edge e. This definition connects directly to the pin-jointed axial-force formula P ~ EA/l: longer edges receive fewer weighted shortest paths and are predicted to carry less axial load. Restricting the sum to boundary pairs also cuts the number of source-target pairs from roughly $N^{2}$ to S times T, lowering the observed computational scaling to about O($n^{{1.5}}$ log n) even though Dijkstra's algorithm replaces breadth-first search.","core_discovery":"The central claim is that stress in a strut lattice concentrates along the length-weighted shortest paths connecting the compressed boundary to the opposite boundary, and that a betweenness centrality computed only over those source-target pairs, EBC_LB, therefore identifies the struts that carry high stress. In every tested network, EBC_LB outperformed the standard geodesic edge betweenness centrality EBC_G, and the improvement grew as lattices became more ordered and anisotropic, because boundary conditions lift topological degeneracies that EBC_G cannot see. For a force applied vertically to a rhombic lattice, for example, the high-EBC_LB edges distribute in an hourglass pattern matching the observed stress pattern, even though every edge is topologically and geometrically equivalent in the infinite lattice.","pith_inferences":["Beyond the paper's 2D demonstrations, EBC_LB could serve as a fast surrogate in iterative topology-optimization loops, replacing many finite element solves with a graph calculation.","The boundary-restricted shortest-path idea transfers naturally to electrical and thermal transport through percolating networks, where the analogue of the loaded edge is the contact with an electrode or heat reservoir.","For rigid-jointed lattices where bending moments dominate, an EBC-type parameter that encodes strut angles or bending stiffness, rather than only length, would be a natural next test; the paper's own kite comparison shows only step-function agreement with FEA stress ratios.","Because the centrality prediction hinges on identifying source and target boundaries, samples must be cut consistently relative to the loading direction; reorienting the same lattice changes predicted hotspots, matching the paper's observation that rotation requires a new cut sample."],"forward_implications":["For a 2D strut lattice under uniaxial compression, struts with EBC_LB above the lattice average can be flagged as stress-bearing without running a finite element simulation.","The advantage of EBC_LB over EBC_G grows as lattices become more ordered and anisotropic, so the parameter matters most for precisely the architectures where symmetry makes standard centrality degenerate.","Because only boundary-pair shortest paths are needed, EBC_LB is cheaper to compute than EBC_G, by roughly an order of magnitude for square samples.","Stress classification can be tuned by lowering the EBC threshold, and ROC curves let designers choose the trade-off between missed hotspots and false alarms.","The authors propose extending the approach to time-dependent EBC_LB(t) for wave propagation and to three-dimensional networks."],"supporting_citations":[{"why":"Establishes the baseline practice of using a graph centrality's relative magnitude to predict which edges carry stress in disordered lattices.","marker":"[20]"},{"why":"Shows betweenness centrality predicts forces in granular packings, the precedent for using centrality as a mechanical-load indicator.","marker":"[23]"},{"why":"Supplies the algorithm used to generate the auxetic lattice test case with aligned edge orientations.","marker":"[30]"},{"why":"Motivates the length dependence of axial force in strut networks (P ~ EA/l) used to justify length weighting.","marker":"[4]"},{"why":"Cited for the classical beam and truss theory underlying the axial-force dependence on strut length and stiffness.","marker":"[36]"},{"why":"Documents the node-representation ambiguity for nanofiber intersections that motivates the continuity discussion.","marker":"[35]"},{"why":"Provides prior evidence that edge betweenness centrality can act as a failure predictor in disordered materials.","marker":"[34]"},{"why":"The graph-extraction software used in the paper to convert network images into graphs without node merging.","marker":"[39]"}],"fun_headline_variants":["Boundary-aware betweenness beats standard stress metrics","Counting boundary-to-boundary paths predicts strut stress","Stress hotspots found by boundary-aware centrality","Edge betweenness tuned to boundaries maps stress","New centrality captures strut stress patterns"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the stress a strut carries under uniaxial compression is well approximated by the fraction of length-weighted shortest paths between the loaded and opposite boundaries that pass through that strut; this is motivated by axial-force scaling P ~ EA/l, but the experiments and finite element analysis use rigid-jointed lattices where bending moments also contribute.","fun_headline_variants_meta":{"raw":{"variants":["Boundary-aware betweenness beats standard stress metrics","Counting boundary-to-boundary paths predicts strut stress","Stress hotspots found by boundary-aware centrality","Edge betweenness tuned to boundaries maps stress","New centrality captures strut stress patterns"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000207,"raw_usage":{"total_tokens":1351,"prompt_tokens":848,"completion_tokens":503,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":464,"completion_tokens_details":{"reasoning_tokens":436}},"tokens_in":464,"tokens_out":503,"duration_ms":10866,"temperature":1.0,"reasoning_tokens":436,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T11:30:27.063981+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Perform finite element analysis on a rigid-jointed lattice whose struts have strongly varying thickness while keeping lengths comparable, and compare measured axial stresses to EBC_LB rankings; a short thick strut carrying high stress while ranked low by EBC_LB would falsify the claim that length-weighted path counts alone predict stress.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the baseline practice of using a graph centrality's relative magnitude to predict which edges carry stress in disordered lattices."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shows betweenness centrality predicts forces in granular packings, the precedent for using centrality as a mechanical-load indicator."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the algorithm used to generate the auxetic lattice test case with aligned edge orientations."},{"cited_title":"P ., Liu, A","cited_arxiv_id":null,"evidence_quote":"Motivates the length dependence of axial force in strut networks (P ~ EA/l) used to justify length weighting."},{"cited_title":"& Young, D","cited_arxiv_id":null,"evidence_quote":"Cited for the classical beam and truss theory underlying the axial-force dependence on strut length and stiffness."},{"cited_title":"A., Mahler, S., Hammig, M","cited_arxiv_id":null,"evidence_quote":"Documents the node-representation ambiguity for nanofiber intersections that motivates the continuity discussion."},{"cited_title":"& Moretti, P","cited_arxiv_id":null,"evidence_quote":"Provides prior evidence that edge betweenness centrality can act as a failure predictor in disordered materials."},{"cited_title":"GitHub https://github.com/compass-stc/StructuralGT","cited_arxiv_id":null,"evidence_quote":"The graph-extraction software used in the paper to convert network images into graphs without node merging."}],"review_version":1}