{"id":"9096fca2-80d7-4972-b9cd-c7cc667d16bf","arxiv_id":"2507.08157","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A programmable Gaussian boson sampling photonic processor extracts k-cliques and topological features from complex-weighted networks.","lead":"Researchers built a programmable photonic quantum computer that samples dense subgraphs in weighted networks, and used it to find cliques, estimate topological features, and track percolation. The work is a proof-of-principle that Gaussian boson sampling hardware can assist topological data analysis, though it does not demonstrate a quantum speedup.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Betti-number estimates from threshold-selected GBS cliques are not validated against exact Betti numbers; the 'estimate' may describe a sampling-biased subcomplex, not the input network.","rationale":"The paper reports a substantial experimental build (Babbage) and an interesting application of GBS's dense-subgraph bias. The clique-search improvement over uniform and squashed-state sampling (Fig. 1c) is plausible and consistent with prior GBS dense-subgraph results; the entropy-percolation correlation (Fig. 4) is qualitative but presented as such. The weakest point is the Betti-number estimation, which the abstract presents as a headline capability. The Methods describe a threshold 'determined by sorting the probability distribution of GBS sampling results'; the selected cliques are therefore not an independent random sample of the clique complex. Because GBS preferentially finds high-weight cliques, the complex built from those cliques is systematically different from the full clique complex; low-weight cliques can be essential for filling or creating cycles. The formula beta_k = m_k - r_k - r_{k+1} is correct for a given simplicial complex, but the complex being used is GBS-selected. The paper even acknowledges complete clique search is hard and GBS must be combined with heuristics (Methods, 'Advantage and complexity analysis'), and no exact Betti numbers are reported for comparison. A single small-network validation against exhaustive enumeration would settle whether the sampling bias changes the Betti numbers; until then, the claim should be read as 'Betti numbers of the GBS-selected subcomplex.' This matches the reader's weakest assumption, so the CONDITIONAL verdict is appropriate; no change is needed, but the requested additions (exact Betti comparison, fixed thresholds, error bars, released data) are essential.","tokens_in":16449,"tokens_out":3493,"duration_ms":37474,"concrete_test":"Test the Betti pipeline on a small network (e.g., N=12, the same type of dual-layer complex-weight model as Fig. 1b): (1) simulate GBS exactly using the paper's decomposition A' = cA + dI and Hafnian probabilities; (2) apply the paper's threshold-selection rule to choose 5-cliques, reconstruct the graph, and compute beta_0-beta_3 from beta_k = m_k - r_k - r_{k+1}; (3) classically enumerate all cliques and compute the true Betti numbers of the same network filtered by the same density threshold, and also of the full unfiltered network. If the GBS-derived Betti numbers equal the density-filtered exact values, the bias is benign; if they differ (or if either differs from the full-network values), the 'estimate' is not a network Betti number and the claim needs qualification.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The Betti-number pipeline (Methods, 'Boundary matrix and Betti number') is the core of the claim that the processor 'estimates Betti numbers.' The threshold used to select k-cliques is 'determined by sorting the probability distribution of GBS sampling results,' so the same data that defines the selected clique set also sets the cut. GBS is biased toward high-weight, dense subgraphs, so the reconstructed graph is a sampling-biased subcomplex, not the input network. The paper provides no comparison with exact Betti numbers (computable classically for these small graphs), and it explicitly concedes that complete k-clique search 'remains a challenging task' and that GBS must be combined with classical heuristics. Without ground truth, beta_k computed via beta_k = m_k - r_k - r_{k+1} may describe holes and components of the GBS-selected subcomplex rather than of the network. This is load-bearing because the abstract's concrete deliverable is Betti-number estimation; the other main results (clique-search improvement, entropy-percolation correlation) do not rescue this step.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a programmable photonic Gaussian boson sampling (GBS) processor, 'Babbage,' with temporal-polarization-hybrid encoding, and applies it to topological network analysis of a random 40-node dual-layer network with complex edge weights. The authors claim three main results: (i) GBS-assisted heuristic search finds high-weight k-cliques with improved success rate over classical uniform sampling and squashed-state sampling; (ii) Betti numbers of the network can be estimated by thresholding GBS sampling probabilities, filtering cliques, and computing boundary-matrix ranks; and (iii) the normalized Rényi entropy of GBS sampling patterns tracks k-clique percolation and can detect topological phase transitions through the Euler characteristic.","tokens_in":16609,"tokens_out":3023,"duration_ms":36239,"significance":"If validated, the work would demonstrate a practical, near-term use of GBS for network science, especially for complex-weighted graphs where classical dense-subgraph methods are less developed. The paper's strengths include a fully programmable photonic platform, use of the externally established GBS bias toward heavy dense subgraphs, and a genuinely interesting proposal to use sampling entropy as a percolation indicator without exhaustive clique enumeration. However, the central quantitative claims currently rest on a single network realization, small sample counts, and threshold choices derived from the same data used to define the reconstructed complex, with no ground-truth validation for the Betti-number estimates.","major_comments":[{"comment":"The Betti-number pipeline is not validated against exact Betti numbers. The threshold used to select 5-cliques is 'determined by sorting the probability distribution of GBS sampling results,' so the same data that defines the selected clique set also sets the cut. Because GBS is biased toward high-weight dense subgraphs, the reconstructed graph is a sampling-biased subcomplex rather than the input network, and βk = mk − rk − rk+1 computed from it may describe holes and components of that subcomplex. For a 40-node network, exact Betti numbers are classically computable, so the authors should compare their estimates with ground truth across the threshold range and report the discrepancy. Without this comparison, the abstract's concrete deliverable of 'estimate Betti numbers' is unsupported.","section":"Methods, 'Boundary matrix and Betti number'; Fig. 2b"},{"comment":"The claimed clique-search enhancement is demonstrated on one random dual-layer network with approximately 3000 samples and no error bars. The improvement ratio p_GBS/p_classical and p_GBS/p_squashed is a single realization and could be dominated by sampling noise or by the details of the post-processing (greedy shrinking and local search). The authors should report results over multiple random network instances, give standard errors or confidence intervals, and specify whether the identical post-processing was applied to all three input types. Without these, the 'improvement' in Fig. 1c is not statistically established.","section":"Fig. 1c and Methods, 'Improvement in the success rate of GBS compared to classical sampling'"},{"comment":"The network encoding uses A′ = cA + dI with rescaling constants c and d chosen to ensure that all eigenvalues tanh r_i lie between 0 and 1. These constants are free parameters that directly control the sampling probability distribution and therefore affect every downstream quantity: clique identification, Betti numbers, Euler characteristics, and Rényi entropy. The manuscript does not specify how c and d are chosen, nor how sensitive the results are to this choice. The authors should provide an explicit prescription and a sensitivity analysis, since otherwise the experimental mapping from network weights to GBS probabilities is underdetermined.","section":"Methods, 'Details about the TPH-encoding photonic quantum processor'"},{"comment":"The claim that normalized Rényi entropy H̃2 'closely mirrors' the percolation order parameter Φ is supported only by visual inspection. No quantitative agreement metric (e.g., correlation coefficient, root-mean-square error, or threshold-estimation error) is given, and the 5-photon entropy curve is based on only about 3000 samples, so statistical uncertainty is non-negligible. The order α = 2 is also chosen without justification or robustness testing. The authors should quantify the agreement, provide error bars, and show that the qualitative conclusion is stable under variations of α and sample size.","section":"Fig. 4 and Methods, 'Details about using Rényi entropy to detect clique percolation'"}],"minor_comments":[{"comment":"The sentence 'Although a quantum computational advantage in estimating Betti numbers not exhibit' is missing a verb and should be rewritten (e.g., 'Although we do not exhibit a quantum computational advantage...').","section":"Main text, Discussion"},{"comment":"The phrase 'there is currently no efficient classical method can simulate the outcomes of GBS experiments' should read 'there is currently no efficient classical method that can simulate...'.","section":"Main text, Discussion"},{"comment":"Reference [40] is listed as 'Phys. Mon. Phys. 80, 1275-1335 (2008)'; this should be 'Rev. Mod. Phys. 80, 1275-1335 (2008)'.","section":"References"},{"comment":"Reference [62] is cited as 'PRX Quantum 37, 100247 (2020)'; PRX Quantum does not use volume 37, and the title 'Big networks: A survey' suggests the journal and volume are incorrect. Please verify and correct this reference.","section":"References"},{"comment":"The caption mentions a '220 m lone-fibre'; this should presumably be '220 m long-fibre'.","section":"Fig. 1 caption"}],"recommendation":"major_revision","confidential_remarks":"The Betti-number validation gap is the main risk: the paper's headline application is not checked against exact classical computation, which is feasible at this scale. The authors should be pushed to add ground-truth comparisons and error bars before publication. I do not see evidence of deliberate misrepresentation; the manuscript candidly acknowledges the need for classical heuristics and the absence of demonstrated quantum advantage."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Let me give you the short version. This paper does something genuinely new: it encodes a complex-weighted network into a programmable photonic GBS device and uses the sampling statistics to guide clique finding, Betti-number estimation, and percolation detection. The hardware effort is real—the TPH-encoded processor with tunable squeezing and a reconfigurable interferometer is a substantial piece of engineering. The clique-search result is plausible: GBS finds high-weight cliques more efficiently than uniform or squashed-state sampling on a 40-node network, though it is one network with about 3000 samples and no error bars.\n\nThe strongest new piece, in my reading, is the Rényi-entropy percolation indicator. They compare it to the order parameter Φ computed from the actual clique structure, and the agreement looks decent. That is a legitimate empirical correlation, and it avoids exhaustive clique enumeration, which is the right kind of claim for a near-term device.\n\nThe soft spot is the Betti-number estimation. The method selects k-cliques above a probability threshold that is itself determined from the GBS sampling distribution, then reconstructs a network from those cliques and computes Betti numbers via boundary-matrix ranks. The problem is that this describes the topology of the GBS-selected subcomplex, not necessarily the input network, and no comparison to exact Betti numbers is made. For a network this small, exact computation is trivial—they could have generated ground-truth values in minutes. The paper concedes that complete clique search remains hard and that GBS needs classical heuristics, which is honest, but it weakens the abstract's claim that the processor 'estimates Betti numbers.' As written, the estimate is unvalidated.\n\nThere are a few other minor issues. No code or data is released, which is frustrating for a methods paper. The discussion of classical simulation of GBS is too optimistic—Ref. [65] directly challenges the claim that no efficient classical method can simulate the outcomes. And the 'large-scale' description is a stretch for a 40-node network.\n\nOverall, I think this deserves a serious referee. The experiment is real, the percolation result is interesting, and the combination of GBS with complex-weighted TDA is worth exploring. But the Betti-number section needs exact comparisons, fixed pre-registered thresholds, and honest error bars before the central claim is supportable. I would send it to review with a request for major revision.","headline":"A real experimental proof-of-principle for GBS-based network analysis, but the Betti-number estimate is unvalidated and needs exact comparison before the central claim can stand.","tokens_in":17247,"tokens_out":2690,"would_cite":false,"duration_ms":28795,"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 programmable photonic quantum processor can find heavy cliques, estimate Betti numbers, and flag percolation transitions in weighted networks.","keywords":["Gaussian boson sampling","topological data analysis","Betti numbers","k-clique identification","complex-weighted networks","clique percolation","Rényi entropy","programmable photonic processor"],"falsifier":"On a small weighted network whose exact Betti numbers and percolation threshold are known from exhaustive clique enumeration, one could encode the network in the photonic processor and compare the GBS-derived $\\beta_k$ and $\\tilde{H}_2$-based threshold with the exact values while varying the sampling-probability threshold and the number of samples. If the estimates converge to the topology of a sampling-selected subcomplex rather than to the input network's true topology, the representativeness assumption fails.","tokens_in":16183,"feed_emoji":"⚛️","tokens_out":10620,"duration_ms":107827,"temperature":0.7,"pith_summary":"This paper claims that Gaussian boson sampling on a programmable photonic processor can serve as a practical route into topological data analysis for complex-weighted networks. Because the sampling distribution preferentially produces high-weight, densely connected subgraphs, the photon patterns surface the k-cliques most relevant to the network's structure before a classical post-processing step. From those cliques the authors build boundary matrices, estimate Betti numbers, track the Euler characteristic through a two-dimensional filtering process, and locate topological phase transitions. They further find that the normalized Rényi entropy of the sampling patterns closely tracks the clique-percolation order parameter, offering an entropy-based percolation indicator. The central contribution is a new application: photonic quantum sampling as a heuristic guide for the computationally hard step of finding the simplicial complex underlying a network.","feed_headline":"Photonic processor finds heavy cliques and Betti numbers","feed_subtitle":"Sampling bias toward dense subgraphs exposes Betti numbers, transitions, and percolation without exhaustive clique search.","key_machinery":"The central machinery is Gaussian boson sampling (GBS), a photonic scheme in which squeezed light passes through a reconfigurable interferometer and photon-number measurements sample subgraphs with probability biased toward high edge weights and dense connectivity. The network's adjacency matrix enters through the decomposition $A' = U \\oplus_i \\tanh(r_i) U^T$, so an arbitrary complex-weighted graph can be programmed into the device. That bias is what turns sampling patterns into a guide to weighted k-cliques; the boundary matrices built from those cliques carry the Betti-number computation, and the Rényi entropy of the sampling patterns carries the percolation signal.","core_discovery":"On a reconfigurable photonic processor that encodes complex-valued adjacency matrices into squeezed-light interference, the paper claims that the output statistics of Gaussian boson sampling are biased toward high-weight dense subgraphs. The sampled k-cliques, filtered by a sampling-probability threshold and refined by a classical greedy-shrinking search, are used to reconstruct the network's clique complex; Betti numbers follow from the boundary-matrix formula $\\beta_k = m_k - r_k - r_{k+1}$, and the Euler characteristic $\\chi = \\sum_k (-1)^k m_k$ locates topological phase transitions where $\\chi$ vanishes. The paper also reports that the normalized Rényi entropy $\\tilde{H}_2$ of k-photon sampling patterns closely mirrors the clique-percolation order parameter $\\Phi$, and that shifts in this entropy track deliberately engineered topological damage. The stated objective is not a quantum speedup for Betti-number estimation but a sampling-based method that returns the identities and weights of k-cliques at once, which the authors describe as an avenue not previously explored.","pith_inferences":["The authors leave implicit that the sampling-probability threshold defines a measurement-selected subcomplex, so the reported Betti numbers are best read as homology estimates of that subcomplex; a direct comparison with exact Betti numbers on small synthetic networks would make this explicit.","A testable extension is to run the Rényi-entropy percolation indicator online while a network's weights drift, using it as a dynamic order parameter in streaming or evolving networks; the paper hints at dynamic application but does not demonstrate it.","One could ablate the imaginary part of the complex edge weights and rerun the experiment to see how much of the topological signal comes from the phase layer rather than the real-valued magnitudes, directly testing the claimed advantage of complex-weight encoding over non-negative graph encodings."],"forward_implications":["For weighted networks with uneven weight distributions, the GBS bias narrows the search to high-weight regions first, so the method can find high-weight k-cliques faster than uniform or squashed-state classical sampling baselines.","The same sampled cliques feed a boundary-matrix construction that yields Betti numbers and Euler characteristics, giving a clique-density-guided filtering process for weighted networks.","The normalized Rényi entropy of k-photon sampling patterns can serve as a percolation indicator, identifying percolation thresholds and topological damage without enumerating all k-cliques.","The processor is programmable and modular, and the paper presents it as scalable toward hundreds of modes, with applications envisaged in brain-network analysis, protein engineering, and connectivity in materials or porous media.","The paper explicitly stops short of claiming quantum computational advantage for Betti-number estimation; the claimed contribution is the sampling-based construction of the simplicial complex, not a complexity-theoretic speedup."],"supporting_citations":[{"why":"supplies the temporal-polarization hybrid loop architecture that the reconfigurable processor is built from","marker":"[34]"},{"why":"establishes the dense-subgraph search capability of GBS that underlies the clique-finding bias","marker":"[35]"},{"why":"provides evidence that GBS sampling correlates with high-density cliques for complex-valued weights and introduces the clique-density measure used for filtering","marker":"[37]"},{"why":"supplies the uniform-sampling and quantum-inspired classical baselines against which the improvement ratios are measured","marker":"[38]"},{"why":"provides the graph-to-squeezing decomposition and the greedy-shrinking and local-search post-processing that turn raw samples into k-cliques","marker":"[60]"},{"why":"defines k-clique percolation clusters and the order parameter Phi used as ground truth for the entropy-based indicator","marker":"[42]"},{"why":"supplies the two-dimensional filtering framework and Euler entropy used to locate topological phase transitions","marker":"[41]"},{"why":"gives the formula beta_k = m_k - r_k - r_{k+1} connecting clique counts and boundary-matrix ranks to Betti numbers","marker":"[64]"},{"why":"states the #P-hardness of clique-complex construction and Betti-number computation that motivates the heuristic sampling approach","marker":"[22]"}],"fun_headline_variants":["Photonic processor maps heavy cliques and Betti numbers","Boson sampling reveals network topology via heavy subgraphs","Programmable photonic chip detects topological transitions","Squeezed-light sampling finds cliques and phase changes","Quantum photonics decodes network topology from cliques"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument depends on the assumption that the high-weight cliques GBS preferentially samples are the ones that determine a network's topological structure, so that cutting the clique complex at a sampling-probability threshold yields Betti numbers and percolation signals that describe the original network rather than a sampling-biased subcomplex.","fun_headline_variants_meta":{"raw":{"variants":["Photonic processor maps heavy cliques and Betti numbers","Boson sampling reveals network topology via heavy subgraphs","Programmable photonic chip detects topological transitions","Squeezed-light sampling finds cliques and phase changes","Quantum photonics decodes network topology from cliques"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000201,"raw_usage":{"total_tokens":1362,"prompt_tokens":914,"completion_tokens":448,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":530,"completion_tokens_details":{"reasoning_tokens":372}},"tokens_in":530,"tokens_out":448,"duration_ms":19971,"temperature":1.0,"reasoning_tokens":372,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T18:25:49.366835+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"On a small weighted network whose exact Betti numbers and percolation threshold are known from exhaustive clique enumeration, one could encode the network in the photonic processor and compare the GBS-derived $\\beta_k$ and $\\tilde{H}_2$-based threshold with the exact values while varying the sampling-probability threshold and the number of samples. If the estimates converge to the topology of a sampling-selected subcomplex rather than to the input network's true topology, the representativeness assumption fails.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"supplies the temporal-polarization hybrid loop architecture that the reconfigurable processor is built from"},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"establishes the dense-subgraph search capability of GBS that underlies the clique-finding bias"},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"provides evidence that GBS sampling correlates with high-density cliques for complex-valued weights and introduces the clique-density measure used for filtering"},{"cited_title":"Jiang, L","cited_arxiv_id":null,"evidence_quote":"supplies the uniform-sampling and quantum-inspired classical baselines against which the improvement ratios are measured"},{"cited_title":"& Arra- zola, J","cited_arxiv_id":null,"evidence_quote":"provides the graph-to-squeezing decomposition and the greedy-shrinking and local-search post-processing that turn raw samples into k-cliques"},{"cited_title":"& Vicsek, T","cited_arxiv_id":null,"evidence_quote":"defines k-clique percolation clusters and the order parameter Phi used as ground truth for the entropy-based indicator"},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"supplies the two-dimensional filtering framework and Euler entropy used to locate topological phase transitions"},{"cited_title":"& Chen, G","cited_arxiv_id":null,"evidence_quote":"gives the formula beta_k = m_k - r_k - r_{k+1} connecting clique counts and boundary-matrix ranks to Betti numbers"},{"cited_title":"& Lloyd, S","cited_arxiv_id":null,"evidence_quote":"states the #P-hardness of clique-complex construction and Betti-number computation that motivates the heuristic sampling approach"}],"review_version":1}