{"id":"0b242b59-53e6-40a1-8653-aee6c0385439","arxiv_id":"2501.08576","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Deploying IRSs near transceivers or in distributed arrangements can improve wireless coverage and throughput, as shown by simulations and two urban field trials.","lead":"This paper surveys how to place and coordinate intelligent reflecting surfaces (IRSs) across single- and multi-reflection wireless architectures, and reports two field trials in 5G and mmWave networks. It is a practical synthesis for network engineers, though the empirical validation is preliminary and lacks statistical rigor.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Field-trial gains in Section V are not yet attributable to the IRS: no A/B protocol, scheduler/MCS controls, or variance reporting, and the paper's own massive-MIMO caveat admits the main confound.","rationale":"I agree with the reader that the weakest assumption is attribution of field-trial gains to the IRS. My read of Sections IV-V does not manufacture a new objection; it strengthens the existing one by identifying a specific internal admission (massive MIMO dynamism) and a missing resource-allocation control. The theoretical deployment comparisons in Sections II-III are plausible and mostly cite prior work, so they are not the main risk. The central claim of the abstract and conclusions is validated by the field tests, so the empirical section is load-bearing. A conditional acceptance is appropriate: the survey content is useful, but the quantitative claims should be presented as preliminary field observations until controlled measurements or raw data are supplied. I would keep the reader's CONDITIONAL verdict; hence UNCHANGED.","tokens_in":10853,"tokens_out":2504,"duration_ms":26647,"concrete_test":"Obtain or rerun the two field trials with a strict A/B protocol: cycle IRS on/off repeatedly (e.g., 30-second intervals) along the same route, log UE positions via GPS, and freeze or record the scheduler state—allocated PRBs, MCS, transport block size, HARQ retransmissions, and offered traffic load. Then compute paired IRS-on minus IRS-off differences in RSRP and throughput with confidence intervals. As an additional analytical check, convert the measured median RSRP gain into a predicted throughput gain under the same PRB/MCS and compare with the observed 246%/360% gains; if the observed gain exceeds the SNR-based prediction by a large margin, the throughput claim is not attributable to the IRS.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section V is the only direct empirical support for the paper's central claim that IRS deployment 'greatly improves' practical network performance, but the reported numbers (RSRP gains 10.27 and 15.12 dB; throughput gains 246% and 360% in Figs. 5-6) are insufficiently controlled. No details are given on whether the massive MIMO precoder, scheduler, modulation/coding scheme (MCS), allocated physical resource blocks (PRBs), traffic load, or UE positions were held fixed between IRS-on and IRS-off measurements; no error bars, confidence intervals, sample sizes, or repeated trials are reported. The authors themselves state for the 2.6 GHz trial that massive MIMO operates dynamically while the IRS is semi-static, 'potentially reducing its overall performance impact' (Section V, Fig. 5(c) discussion), which concedes a time-varying confound in the same data used to claim gains. The 26 GHz trial has two UEs on a single route, so throughput increases could reflect scheduling or load changes rather than the IRS. Because these field trials are the load-bearing empirical validation of the headline claim, the attribution is not yet established.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This manuscript surveys deployment architectures for intelligent reflecting surfaces (IRSs) in wireless networks. It reviews single-reflection deployment for point-to-point, point-to-multipoint, and point-to-area setups, discussing trade-offs between passive and active IRS designs and between centralized and distributed IRS placements. It then covers double- and multi-reflection cases, highlights practical constraints such as hardware, maintenance, and system integration, and reports two field trials of IRS deployment in commercial 5G and mmWave networks. The paper concludes by outlining several future research directions, including movable IRSs, near-field deployments, and network-level optimization.","tokens_in":11209,"tokens_out":4066,"duration_ms":40635,"significance":"If its field-trial evidence were methodologically robust, the paper would offer a valuable practical demonstration of IRS gains in operational networks. The survey component usefully synthesizes a body of recent work, including several results from the authors' own group, and the paper candidly acknowledges a key limitation of semi-static IRS operation in massive MIMO systems. However, the empirical core is not yet convincing: the reported RSRP and throughput improvements are derived from uncontrolled measurements with no statistical support, so the central claim that IRS deployment 'greatly improves' practical network performance is not yet established.","major_comments":[{"comment":"The RSRP and throughput improvements in the 2.6 GHz field trial are not attributable to the IRS because the paper does not report whether the massive MIMO precoder, scheduler, MCS, PRB allocation, traffic load, or UE positions were held fixed between IRS-on and IRS-off measurements. The paper's own discussion of Fig. 5(c) states that massive MIMO operates dynamically while the IRS is semi-static, which is a time-varying confound in the same data. No error bars, confidence intervals, sample sizes, or repeated-trial descriptions are provided, so the observed gains cannot be separated from other network dynamics.","section":"Section V, Figs. 5(b)-5(c)"},{"comment":"The 26 GHz trial reports results from only two UEs on a single route, and the throughput increases of 246% and 360% could reflect changes in scheduling, traffic load, or UE position rather than the effect of the two IRSs. The manuscript does not describe a controlled A/B protocol, variance reporting, or any statistical measure of uncertainty, making the numerical gains non-reproducible and insufficient to support the claimed causal effect.","section":"Section V, Fig. 6"},{"comment":"The sentence 'These results demonstrate that deploying IRSs can effectively boost signal strength and data transmission efficiency, thereby greatly improving the system's performance' overstates the evidentiary value of the uncontrolled trials. The paper should either add rigorous statistical controls and repeated trials, or explicitly rephrase this conclusion as an illustrative case study whose quantitative results are limited by the uncontrolled field environment.","section":"Section V, final paragraph"}],"minor_comments":[{"comment":"The heading 'CONCUSIONS' is a typo and should read 'CONCLUSIONS'.","section":"Section VI"},{"comment":"The word 'commerical' in the caption should be spelled 'commercial'.","section":"Fig. 5 caption"},{"comment":"The word 'pasive' in the sentence about the hybrid IRS architecture is a typo and should be 'passive'.","section":"Section II-A-1"},{"comment":"The numerical examples in Figs. 2-4 are presented without simulation parameters such as channel models, geometry, noise figures, or CSI assumptions; either provide this information or state that the figures are reproduced from the cited works.","section":"Figs. 2-4"},{"comment":"Two references are arXiv preprints; if the manuscript is intended as a journal review, consider citing the published versions if they have appeared.","section":"References [15], [22]"}],"recommendation":"major_revision","confidential_remarks":"The field trials are self-reported by the same group that developed the IRS technology, and several of the cited theoretical results also come from the authors' own group. An independent evaluation or a clear statement of competing interests would strengthen the credibility of the empirical claims. The editor should assess whether the paper's conclusions are sufficiently supported given the uncontrolled nature of the field measurements."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThe useful part of this paper is the survey of IRS deployment architectures; the new part is the two field trials, and those are the weak spot. The survey systematically organizes single- and multi-reflection deployment guidelines — near-BS versus near-user, centralized versus distributed, passive versus active — and it accurately reflects what the literature says. The numerical examples are illustrative and consistent with cited results. As a summary of the design space, it is readable and likely useful for someone new to IRS deployment.\n\nThe field trials, though, are not controlled enough to support the paper’s claims. The Nanjing 2.6 GHz trial reports CDFs of RSRP and throughput, but the authors themselves note that massive MIMO precoding updates dynamically while the IRS is semi-static, which is a time-varying confound. The Shanghai 26 GHz trial reports two UEs on one route, with RSRP gains of 10.27 and 15.12 dB and throughput gains of 246% and 360%, but there are no details on scheduler, MCS, PRB allocation, traffic load, or repeated trials. No error bars, no A/B protocol. So the attribution of these gains to the IRS is not established. That is a load-bearing flaw because the paper’s headline claim is that field tests validate IRS effectiveness.\n\nI don’t think the rest of the paper is problematic. The self-citation pattern is heavy but mostly to peer-reviewed results, and the paper is transparent about the massive MIMO limitation. The qualitative deployment rules are consistent with earlier work and the paper doesn’t overclaim there.\n\nWho should read this? Anyone wanting a concise orientation on IRS deployment trade-offs. The field trial data should be cited with caution, if at all, until the methodology is expanded or the claims are softened. I would send it to peer review, but with major-revision expectations on Section V. A serious referee could help the authors either add the missing control information or narrow the empirical claims to what the data can support.","headline":"A solid survey of IRS deployment guidelines, but the field trials are the only new data and they are not yet controlled enough to support the empirical claims.","tokens_in":11599,"tokens_out":2847,"would_cite":true,"duration_ms":27161,"reading_group":"yes","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 argues that intelligent reflecting surfaces (IRSs) deployed near the base station or near the user — not in between — can sharply improve coverage and throughput, and it backs the claim with two operator-style field trials.","keywords":["intelligent reflecting surfaces","IRS deployment","single-reflection","multi-reflection","field trials","wireless coverage","capacity enhancement","passive and active IRS"],"falsifier":"Repeat the Nanjing and Shanghai field tests with IRS panels physically covered or set to a fixed random phase configuration while keeping the same route, same UEs, and same network load; if the RSRP and throughput distributions are statistically indistinguishable from the IRS-on case, the central empirical claim fails.","tokens_in":10694,"feed_emoji":"📡","tokens_out":3942,"duration_ms":36521,"temperature":0.7,"pith_summary":"This review-style paper aims to establish practical deployment principles for intelligent reflecting surfaces (IRSs) in wireless networks and to show, with field evidence, that properly placed IRSs improve real network performance. It argues that placement is governed by a trade-off between two-hop path loss and beamforming gain, so a single IRS should sit either close to the base station or close to the user. For multi-reflection links, cooperative gain scales as $O(N^4)$ but requires enough elements to overcome extra path loss. The paper also compares centralized versus distributed IRS architectures and reports field trials in which IRSs raised RSRP by 10.27 and 15.12 dB and downlink throughput by 246 and 360 percent.","feed_headline":"IRS field trials: 10–15 dB signal gain, 246–360% faster downlink","feed_subtitle":"Placement rules from two commercial 5G and mmWave tests show where intelligent reflecting surfaces deliver.","key_machinery":"The central object is the two-hop IRS link and its path-loss/beamforming-gain trade-off, combined with the distinction between single- and multi-reflection architectures. A single IRS minimizes product distance by sitting near transmitter or receiver; an active IRS, whose amplification depends on the first-hop channel, prefers the receiver side. In the multi-antenna case, multiple small IRSs can create orthogonal sub-channels and raise channel rank, enabling spatial multiplexing. For double reflection, cooperative phase alignment gives a passive beamforming gain of $O(N^4)$ with $N$ total elements. The field trials are the validation instrument: CDFs of RSRP and downlink throughput with and without IRS.","core_discovery":"On its own terms, the paper's central claim is that IRS deployment is a first-order design variable: choosing where surfaces go, how many reflections are used, and how elements are split between passive and active operation determines whether the technology delivers coverage and capacity gains in practice. The strongest evidence is the two field trials: a single IRS in a 5G commercial network improved RSRP distributions across a test route, and two IRSs in a 26 GHz mmWave network raised two UEs' RSRP from -84.98 to -74.71 dB and from -88.39 to -73.27 dB, with downlink throughput rising from 785.94 to 2720.72 Mbps and from 566.34 to 2607.29 Mbps. These gains are presented as validating the deployment principles derived from the analytic results earlier in the paper.","pith_inferences":["The paper's placement rules are derived from idealized LoS models; in dense urban environments, site-specific effects like blockage and scatterer dynamics may shift the optimum, so ray-tracing-based pre-screening is a natural next step.","The field-trial methodology (two UEs, one route, no error bars) suggests that a larger randomized trial with multiple routes and controlled traffic would be needed before operators can rely on the quantitative gains.","If the $O(N^4)$ double-reflection gain holds under realistic channel estimation overhead, then pairing a few large IRSs may be cheaper than many single-reflection links for coverage of blind spots.","The paper's near-field and movable-IRS agendas imply that the current placement rules may change as arrays grow and surfaces acquire mechanical reconfiguration, an implicit admission that today's guidelines are a baseline."],"forward_implications":["An operator can improve coverage by placing an IRS close to the base station or close to the user rather than at mid-distance, where two-hop path loss dominates.","Active IRSs should generally be deployed near the receiver to exploit amplification, while passive IRSs gain from either end position; hybrid active-passive designs can balance cost and SNR.","In the multi-antenna case, distributed small IRSs outperform a centralized large IRS when the total element count is large, because they unlock spatial multiplexing.","Double-reflection IRS links can yield $O(N^4)$ cooperative beamforming gain, but require a sufficiently large element count to overcome the higher cascaded path loss.","Field trials show that per-UE RSRP gains of about 10–15 dB and throughput gains of 246–360 percent are achievable in commercial 5G and mmWave deployments."],"supporting_citations":[{"why":"Supplies the proof that a single IRS should be deployed near the transmitter or receiver to minimize two-hop path loss.","marker":"[4]"},{"why":"Provides the result that an active IRS is best placed near the receiver to maximize amplification and minimize path loss.","marker":"[5]"},{"why":"Establishes that a centralized passive IRS outperforms distributed IRSs in the capacity region for single-antenna multi-user systems.","marker":"[13]"},{"why":"Develops the framework comparing centralized versus distributed IRS architectures for multi-antenna networks used in Fig. 3.","marker":"[15]"},{"why":"Shows BS-side IRS deployment minimizes angular deviation and two-hop path loss, giving $O(N)$ beamforming gain for point-to-area coverage.","marker":"[16]"},{"why":"Proves that two cooperative IRSs achieve $O(N^4)$ passive beamforming gain with $N$ total elements, the basis for double-reflection deployment.","marker":"[18]"},{"why":"Characterizes optimal placement of double active IRSs under total and per-element power constraints.","marker":"[19]"},{"why":"Stochastic geometry analysis showing network throughput improves when IRSs are added, motivating network-level deployment.","marker":"[24]"}],"fun_headline_variants":["IRS field trials: 10–15 dB gain, 3x faster downlink","Where to place IRS: field-tested rules for big coverage gains","Two IRS trials: RSRP up 10–15 dB, throughput triples","IRS deployment strategy dictates gains: field evidence","IRS placement: 10–15 dB signal lift and 3–4x speed in 5G tests"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The field-trial claim rests on the assumption that the measured RSRP and throughput gains are caused by the IRS rather than by unrelated factors such as dynamic massive MIMO precoding, traffic load, or the test UEs' positions, since the trials involve only two UEs on a single route with no statistical controls.","fun_headline_variants_meta":{"raw":{"variants":["IRS field trials: 10–15 dB gain, 3x faster downlink","Where to place IRS: field-tested rules for big coverage gains","Two IRS trials: RSRP up 10–15 dB, throughput triples","IRS deployment strategy dictates gains: field evidence","IRS placement: 10–15 dB signal lift and 3–4x speed in 5G tests"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000717,"raw_usage":{"total_tokens":3209,"prompt_tokens":920,"completion_tokens":2289,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":536,"completion_tokens_details":{"reasoning_tokens":2186}},"tokens_in":536,"tokens_out":2289,"duration_ms":17542,"temperature":1.0,"reasoning_tokens":2186,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T20:22:02.939425+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeat the Nanjing and Shanghai field tests with IRS panels physically covered or set to a fixed random phase configuration while keeping the same route, same UEs, and same network load; if the RSRP and throughput distributions are statistically indistinguishable from the IRS-on case, the central empirical claim fails.","supporting_citations":[{"cited_title":"Intelligent reflecting surface-aided wireless communica- tions: A tutorial,","cited_arxiv_id":null,"evidence_quote":"Supplies the proof that a single IRS should be deployed near the transmitter or receiver to minimize two-hop path loss."},{"cited_title":"Wireless communication aided by intelligent reflecting surface: Active or passive?","cited_arxiv_id":null,"evidence_quote":"Provides the result that an active IRS is best placed near the receiver to maximize amplification and minimize path loss."},{"cited_title":"Intelligent reflecting surface aided multi-user com- munication: Capacity region and deployment strategy,","cited_arxiv_id":null,"evidence_quote":"Establishes that a centralized passive IRS outperforms distributed IRSs in the capacity region for single-antenna multi-user systems."},{"cited_title":"Intelligent Reflecting Surface Aided MIMO Networks: Distributed or Centralized Architecture?","cited_arxiv_id":"2310.01742","evidence_quote":"Develops the framework comparing centralized versus distributed IRS architectures for multi-antenna networks used in Fig. 3."},{"cited_title":"Aerial intelligent reflecting surface: Joint placement and passive beamforming design with 3D beam flattening,","cited_arxiv_id":null,"evidence_quote":"Shows BS-side IRS deployment minimizes angular deviation and two-hop path loss, giving $O(N)$ beamforming gain for point-to-area coverage."},{"cited_title":"Cooperative double-IRS aided communication: Beam- forming design and power scaling,","cited_arxiv_id":null,"evidence_quote":"Proves that two cooperative IRSs achieve $O(N^4)$ passive beamforming gain with $N$ total elements, the basis for double-reflection deployment."},{"cited_title":"Double-active-IRS aided wireless communication: De- ployment optimization and capacity scaling,","cited_arxiv_id":null,"evidence_quote":"Characterizes optimal placement of double active IRSs under total and per-element power constraints."},{"cited_title":"Hybrid active/passive wireless network aided by intelligent reflecting surface: System modeling and performance analysis,","cited_arxiv_id":null,"evidence_quote":"Stochastic geometry analysis showing network throughput improves when IRSs are added, motivating network-level deployment."}],"review_version":1}