{"id":"13749d7c-6be7-4138-8f6a-06b48f4e271c","arxiv_id":"2607.27181","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":6,"one_line_summary":"A multimode-fiber wavefront-shaping platform experimentally implements β-Partial Matching and, in simulation, supports broader one-way quantum communication-complexity decoding.","lead":"Researchers built a reconfigurable multimode-fiber and wavefront-shaping setup that runs one-way quantum communication-complexity tasks, and experimentally solved β-Partial Matching for small input sizes. The platform is flexible enough that simulations suggest it could host harder tasks and larger dimensions without new interferometer hardware.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.5","headline":"Roadmap-to-advantage claim rests on hybrid βPM-vs-VS benchmarking plus idealised (lossless, noiseless) crossings that realistic noise/loss models do not reproduce.","rationale":"Reader correctly flags |α|² log n and efficiency/noise as the weak joint assumption and already assigns CONDITIONAL with high confidence—appropriate for a sound small-n platform paper with aspirational scaling. I agree the metric and end-to-end model are the right neighborhood, but the sharper load-bearing point is not that the metric is simply 'unfair' in the abstract (the Arrazola–Lütkenhaus multi-copy reading is standard) but that the advantage roadmap in Fig. 4 is a hybrid: experimental/realistic βPM cost vs VS classical bits, with crossings only in idealised simulations that zero the limitations the realistic model shows dominate. That does not overturn soundness of the βPM demo or reconfigurability thesis; it keeps the verdict CONDITIONAL rather than moving to REJECT or ACCEPT. No internal contradiction found in the VS classical derivation or visibility comparison; no need to escalate beyond the reader's call.","tokens_in":19805,"tokens_out":760,"duration_ms":114245,"concrete_test":"Replot Fig. 4 under the realistic noise/loss model while sweeping N_modes and η_ports, using cost |α|²⌈log₂ n⌉ needed for ε≤0.2 on both βPM and simulated VS; mark the (N_modes, η_ports, n) where this cost falls below Cor. 1's VS classical d(ε=0.2). If no crossing exists for plausible parameters (e.g. N_modes≤10^4, η_ports≤0.3, n≤10^3), the roadmap claim must be tempered to 'platform demo only.'","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central platform claim is fine as a methods result; the load-bearing soft spot is the 'concrete roadmap' to quantum advantage (abstract, Fig. 4, Discussion). Experiment and realistic simulations implement βPM only, at n≤8, with transmitted cost |α|² log n ~ 5e3–2.5e4 for ε=0.2—orders of magnitude above both βPM and VS classical one-way needs at those n. Fig. 4 then benchmarks those βPM optical costs against the authors' VS classical upper bound (Thm. 2 / Cor. 1), while VS is not run experimentally (phase-only SLM; amplitude encoding missing) and is supported only by visibility simulations (Fig. 7, ≤4% gap). The only curves that approach or cross the VS classical line are ideal simulations (η=1, noiseless camera, perfect wavefront control) with N_modes up to 5000. Realistic runs (η≈54%, EMCCD noise, η_ports≈8%) stay well above and hit barriers; the text notes experiment is worse still due to finite SLM fill/pixels. So the roadmap requires simultaneous jumps in mode count, port efficiency, detection SNR, and Alice amplitude control that are illustrated qualitatively, not shown to be sufficient under the same noise model that matches the data. Separately, motivating |α|² log n via Holevo for an 'n-dimensional' state is imprecise for bright multi-mode coherent states (deep multi-photon per mode in the experiment); the multi-copy single-photon analogy is the real justification and should be stress-tested as the cost metric.","agreement_with_reader":"partial"},"referee_report":{"model":"grok-4.5","summary":"The manuscript introduces a reconfigurable optical platform for one-way quantum communication complexity that uses multimode-fiber mode mixing and SLM wavefront shaping to implement programmable high-dimensional linear optical networks. It experimentally implements the genuine β-Partial Matching (βPM) problem for n=4,6,8 input ports at error ε=0.2, reporting transmitted information |α|² log n ≈ (5×10³, 2×10⁴, 2.5×10⁴), and transmits a binary fingerprint image as a communication demonstration. Complementary simulations with measured transmission matrices compare ideal vs realistic (loss + EMCCD noise) performance, argue that the same decoding stage can support Vector-in-a-Subspace (VS) measurements with similar visibility, and present an explicit classical one-way VS error bound (Theorem 2 and closed-form Corollary 1) used as a stringent benchmark. The authors conclude that MMF wavefront shaping is a versatile hardware platform and sketch a roadmap toward regimes where one-way quantum protocols could beat known classical bounds.","tokens_in":20186,"tokens_out":1464,"duration_ms":36566,"significance":"If the platform claims hold, this is a useful methods contribution: prior photonic communication-complexity experiments used fixed interferometers tailored to narrow operator families, whereas a programmable MMF decoder can in principle address problems (notably VS) that require many distinct high-dimensional linear measurements. The experimental βPM implementation is a genuine one-way task (unlike Sampling Matching), the Supplement supplies a full operator construction, a careful EMCCD noise model, and a nontrivial expansion of Raz’s sketched classical VS protocol into an explicit error bound with prefactors—valuable for future benchmarking. The work is therefore significant as a flexible hardware architecture and as a carefully documented proof-of-principle, even if a practical quantum–classical separation is not yet demonstrated.","major_comments":[{"comment":"Abstract, Fig. 4, and Discussion frame a “concrete roadmap” to quantum advantage, but under the noise/loss model that matches the data the curves remain far above the classical VS benchmark and hit barriers; only ideal simulations (η=1, noiseless camera, perfect wavefront control) with N_modes up to 5000 approach or cross that line. Experiment is limited to βPM at n≤8 with |α|² log n ~ 10³–10⁴, orders of magnitude above both βPM and VS classical one-way needs at those n. The text correctly notes missing Alice amplitude control, finite SLM fill, η_ports≈8%, and detection SNR as gaps, but does not show that simultaneous improvements suffice under the same realistic model. Please either (i) add realistic-parameter projections that close the gap with explicit mode-count/loss/SNR targets, or (ii) substantially tone down “concrete roadmap / practical advantage” language to match what is demons","section":"Abstract; Results Fig. 4; Discussion"},{"comment":"Benchmarking experimental βPM optical cost against the classical one-way VS upper bound (Theorem 2 / Corollary 1) mixes two different problems. VS is the harder classical target and is not run experimentally (phase-only SLM); support for VS is only the ≤4% visibility gap in Fig. 7 under lossless single-photon numerics. Fig. 1 already compares the two classical bounds fairly; Fig. 4 should either (a) also plot the βPM classical bound of Theorem 1 at the same ε, and/or (b) clearly label the VS curve as an aspirational cross-problem benchmark, not as the classical cost of the implemented task. Otherwise readers can misread the experiment as nearly competitive with the relevant classical protocol.","section":"Results “Quantum βPM and classical benchmark with VS”; Fig. 4; Theorem 1–2"},{"comment":"The cost metric |α|² log(n) is motivated by the Holevo bound for an n-dimensional quantum state (Results, Alice’s encoding), but the implemented resource is a product of n coherent states, often bright, whose total extractable information is not simply |α|² log n. The multi-copy single-photon / Poisson-copy analogy (Arrazola–Lütkenhaus) is the real justification and should be stated as primary; Holevo language should be qualified or removed. Please also discuss how end-to-end loss (shared SLM counted as information-carrying, η≈50%, η_ports≈8%) and multi-photon occupancy affect fairness versus classical bit cost, including whether lost photons are charged consistently with a one-way communication-complexity accounting.","section":"Results “Alice’s encoding”; Setup efficiency; Supplement “Modeling losses”"}],"minor_comments":[{"comment":"Fig. 2 caption and main text say n=4 ports and k=4 detection modes in the illustration; ensure consistent use of n (Alice modes) vs k (Bob detection ports) throughout, including when β=1/2 makes k=n for βPM.","section":"Fig. 2; Bob’s decoding"},{"comment":"Table I lists component efficiencies whose product does not obviously equal the stated η≈50% once η_ports≈8% is separated; a one-line breakdown of which factors enter η vs η_ports would help.","section":"Table I; Setup efficiency"},{"comment":"The shared single-SLM architecture (Alice phases added on Bob’s ports) is acknowledged; a short paragraph on what changes in a two-SLM separated implementation (channel loss, independent calibration, timing) would clarify the path from proof-of-principle to a true communication experiment.","section":"Reconfigurable spatial-mode platform"},{"comment":"Typographical/spacing issues appear in several places (e.g., “communicationcomplexityoffersapromisingroute”, “β-Partial Matching”, figure labels). A full copy-edit pass is needed.","section":"Throughout"},{"comment":"Fig. 1 and Fig. 4 would benefit from explicit statement of β and of whether classical curves are worst-case or average-case error, matching the theorems.","section":"Fig. 1; Fig. 4"}],"recommendation":"major_revision","confidential_remarks":"Solid platform/methods paper with careful supplementary theory (especially the VS classical bound). The overclaim is concentrated in abstract/roadmap framing rather than in a broken experiment. Suitable for a quantum-optics or quantum-information journal after the authors separate demonstrated platform performance from aspirational advantage projections; I would not reject on novelty grounds relative to fixed-interferometer fingerprinting/Hidden-Matching experiments."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The useful core is a working reconfigurable decoder: MMF plus wavefront shaping used as a programmable linear network for one-way quantum communication complexity, validated on genuine βPM (not Sampling Matching) at n=4,6,8. That is new relative to fixed interferometer fingerprinting and delay-line HM setups. They measure transmitted optical information at ε=0.2, match realistic simulations that include measured TMs, losses, and EMCCD noise, and show in numerics that the same Bob-side architecture handles VS-type projectors with visibility within a few percent of βPM. The Supplement is careful: full βPM operator, pixel-level noise model, and an original prefactor analysis of Raz’s sketched classical VS protocol plus a closed-form corollary. Circularity is low; they benchmark against external classical bounds.\n\nWhat they do not have is near-term advantage. Optical cost is thousands to tens of thousands of “bits” of |α|² log n while classical one-way needs at these n are tiny. VS is not run in the lab (phase-only SLM), only visibility-simulated. Fig. 4’s curves that approach the VS classical line are ideal (η=1, noiseless, perfect control) at high mode count; realistic runs stay well above and hit barriers, and the experiment is worse still from SLM fill and pixel limits. Calling that a “concrete roadmap” oversells simultaneous jumps in modes, port efficiency, SNR, and Alice amplitude control that are sketched, not shown under the noise model that fits the data. The Holevo-motivated |α|² log n metric is also a stretch for bright multi-mode coherent states; the multi-copy single-photon analogy is the real justification and should be stated more tightly.\n\nNone of that sinks the platform result. For people building photonic linear networks or exploring QCC/hybrid crypto hardware, this is worth reading. Math and citations look solid; data are small-n but honestly reported. I would send it to referees as an experimental methods paper, with pressure to separate demonstrated βPM performance from projected VS/advantage language.","headline":"Solid methods demo of programmable MMF decoding for one-way QCC; the βPM experiment is real, the “roadmap to advantage” is still aspirational.","tokens_in":20856,"tokens_out":532,"would_cite":true,"duration_ms":13840,"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":"Multimode-fiber wavefront shaping is a reconfigurable optical platform that can run genuine one-way quantum communication complexity tasks on present-day photonics.","keywords":["quantum communication complexity","multimode fiber","wavefront shaping","β-partial matching","vector in a subspace","coherent states","photonic quantum advantage","one-way communication"],"falsifier":"Repeat β-Partial Matching on a fiber with thousands of modes and substantially lower detection noise; if the measured information cost at 20 percent error does not move toward the classical one-way bounds the way the ideal simulations predict, the platform roadmap fails.","tokens_in":20611,"feed_emoji":"📡","tokens_out":932,"duration_ms":35819,"temperature":0.7,"pith_summary":"This paper sets out to show that a multimode fiber plus wavefront shaping can act as a programmable decoder for one-way quantum communication complexity, replacing the fixed interferometers used in earlier demos. The authors run the genuine β-Partial Matching problem—an established task with an exponential quantum–classical one-way gap—for small input sizes and measure how much optical information must be sent to keep the error under 20 percent. Numerical work with the same decoding architecture shows it can also realize the harder Vector-in-a-Subspace measurement at comparable visibility, and that more fiber modes with quieter detection would push performance toward classical one-way bounds. A sympathetic reader cares because communication complexity is one of the few settings where current photonic hardware might demonstrate a practical quantum advantage without a large-scale quantum computer. The work therefore supplies both a working apparatus and a concrete hardware roadmap (mode count, loss, noise) for stronger separations.","feed_headline":"Multimode fiber runs reconfigurable quantum communication tasks","feed_subtitle":"β-partial matching experiments and simulations sketch a photonic path to quantum advantage","key_machinery":"Digital phase conjugation on a calibrated multimode-fiber transmission matrix: each SLM input port encodes one column of Bob’s linear operator so that fiber mode-mixing plus wavefront shaping realizes a programmable high-dimensional linear optical network whose two camera regions implement the boolean decision.","core_discovery":"Multimode-fiber wavefront shaping is a versatile reconfigurable platform for one-way quantum communication complexity. The authors experimentally implement the genuine β-Partial Matching protocol, for which an exponential quantum–classical one-way separation is known, at error 0.2 for n = 4, 6 and 8, and show by simulation that the same programmable decoding stage supports more general one-way tasks such as Vector in a Subspace with comparable performance and a route to higher dimension without increasing hardware complexity.","pith_inferences":["Lab demos that share one SLM between Alice and Bob understate the channel loss a true separated Alice–Bob link would face, so any field demonstration will need separate modulators and a revised loss budget.","If the conjectured Ω(√n) classical two-way lower bound for Vector in a Subspace is proved with tight constants, the same hardware targets become a direct test of one-way quantum advantage against interactive classical strategies.","Once Alice gains amplitude as well as phase control, the same decoder can host a wider family of linear-optics communication tasks beyond phase-only matching problems."],"forward_implications":["A single programmable multimode-fiber decoder can replace fixed interferometer setups for many one-way communication-complexity problems.","Increasing physical mode count and detection signal-to-noise can bring a one-way quantum implementation into a regime that beats best-known classical one-way bounds for Vector in a Subspace.","Spectral, temporal, or polarization multiplexing, or multi-photon Fock inputs, can enlarge the usable Hilbert space without changing the hardware topology.","The platform yields explicit experimental targets—mode count, end-to-end efficiency, and camera noise—for protocols with stronger quantum–classical separations."],"fun_headline_variants":["Multimode fiber wavefront shaping runs one-way quantum communication","Reconfigurable optics implement β-partial matching protocol","Fiber platform decodes one-way quantum communication tasks","Wavefront shaping enables programmable quantum communication complexity","Same multimode setup supports higher-dimensional one-way protocols"],"cache_read_input_tokens":16512,"weakest_assumption_plain":"That average transmitted information defined as mean photon number times log of the mode count is a fair comparison between the coherent-state optical protocol and classical bit communication once real loss and camera noise are included.","fun_headline_variants_meta":{"raw":{"variants":["Multimode fiber wavefront shaping runs one-way quantum communication","Reconfigurable optics implement β-partial matching protocol","Fiber platform decodes one-way quantum communication tasks","Wavefront shaping enables programmable quantum communication complexity","Same multimode setup supports higher-dimensional one-way protocols"]},"model":"grok-4.5","effort":"low","cost_usd":0.004957,"raw_usage":{"total_tokens":1380,"prompt_tokens":722,"num_sources_used":0,"completion_tokens":61,"cost_in_usd_ticks":49568000,"prompt_tokens_details":{"text_tokens":722,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":597,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":722,"tokens_out":61,"duration_ms":10383,"temperature":1.0,"reasoning_tokens":597,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-30T11:08:58.525753+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Repeat β-Partial Matching on a fiber with thousands of modes and substantially lower detection noise; if the measured information cost at 20 percent error does not move toward the classical one-way bounds the way the ideal simulations predict, the platform roadmap fails.","supporting_citations":[],"review_version":2}