{"id":"c0a7e319-ab7b-40bf-b848-164292d718bc","arxiv_id":"2502.06239","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A beacon-antenna pre-equalization scheme shifts grant-free massive MIMO random access to a semi-blind iterative detection problem with coarse data detection, data-aided channel estimation, and fine data detection.","lead":"The paper proposes a grant-free massive MIMO uplink access scheme in which users pre-equalize their signals against one base-station beacon antenna, so the base station can detect activity and data without first estimating the full channel. A smart generalist might read this because it targets lower access latency for massive machine-type communication by replacing pilot-based channel acquisition with a semi-blind iterative detector.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 'same access latency' claim is unsupported: Section V counts only pilot symbols for Baselines 1/2, while the proposed scheme's entire data frame and beacon overhead are not charged to the latency budget.","rationale":"The reader's verdict is CONDITIONAL, with the primary weak assumption being perfect user-side channel estimation and reciprocity. I agree that this is a serious idealization, but I find an even more direct problem in the comparison methodology: the paper's own definitions in Section V imply that Baselines 1 and 2 are charged only for their pilot symbols, while the proposed scheme is charged for its entire data frame, and the beacon period is not counted at all. Thus the plotted 'same access latency' comparisons do not actually equalize latency, which undermines the headline claim even under perfect CSI. This is a distinct issue from the reader's perfect-CSI concern, although the reader did list the comparison restriction as a secondary fragile premise; hence partial agreement. The verdict should remain CONDITIONAL because the comparison can be fixed and the algorithmic idea is plausible, but the numerical evidence as presented is not conclusive.","tokens_in":12573,"tokens_out":9839,"duration_ms":97397,"concrete_test":"Run the provided repository and re-generate Figs. 3 and 5 with a single end-to-end latency budget: for each scheme, set total access latency L = beacon_period + T_uplink for the proposed scheme, and L = T_pilot + T_data for Baselines 1 and 2, with T_data equal to the number of data symbols used in the baseline BER computation. Then, for each x-axis value, allocate the same total number of OFDM symbols to all schemes and recompute ADEP and BER. If the proposed scheme still outperforms the baselines under this accounting, the central claim stands; if the margins shrink or reverse, the reported gains are not at equal access latency.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim of outperformance 'at the same access latency' is not actually demonstrated by the simulations. In Section V, the paper states that for Baselines 1 and 2, which use a 'pilot+data' frame, the time-slot overhead T refers to the pilot signal only, whereas for the proposed scheme T is the number of OFDM symbols in the entire uplink data frame. The proposed scheme also requires a beacon broadcast period before each uplink frame (Section II), which is omitted from the latency comparison. Therefore, in Figs. 3 and 5 the proposed scheme's full T-symbol data frame is compared against baselines that spend T symbols on pilots and then need additional data symbols for the data detection whose BER is plotted. If the baselines' data slots are counted, their total access latency exceeds T; if the beacon period is counted, the proposed scheme's latency exceeds T. Either way, the plotted comparison does not hold the total time-frequency resource or end-to-end latency fixed. The reported BER/ADEP gains may thus be an artifact of unequal latency budgets, independent of the algorithmic contribution. The perfect user-side channel estimation and reciprocity assumption in Section III is also load-bearing, but the comparison mismatch alone invalidates the numerical support for the stated central claim.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a pre-equalization aided grant-free massive access scheme for massive MIMO systems. In the proposed scheme, the base station activates one beacon antenna to broadcast a beacon; each UE estimates the downlink channel from that antenna, pre-equalizes its uplink signal on the corresponding subcarriers, and transmits a spread-spectrum data frame without any uplink pilot. The BS then performs joint activity and data detection through three iterative modules: coarse data detection based on the beacon-antenna signal, data-aided channel estimation using the detected data as pilots, and fine data detection using all BS antennas. The authors claim via simulations that the scheme outperforms state-of-the-art massive MIMO grant-free NOMA schemes at the same access latency, and they provide simulation code for reproducibility.","tokens_in":12752,"tokens_out":6780,"duration_ms":61000,"significance":"If the claims were fully validated, the scheme would be a meaningful contribution to low-latency grant-free massive access: it removes the uplink pilot overhead, exploits mMIMO diversity in a semi-blind setting, and breaks the joint activity/data/channel estimation problem into three standard linear estimation sub-problems solvable by AMP-type algorithms. The paper also provides reproducible simulation code and compares against several baselines. However, the numerical support for the central \"same access latency\" claim is undermined by an inequitable comparison setup, and a few load-bearing modeling assumptions are either internally inconsistent or left untested. The algorithmic idea is still interesting and potentially correct, but the manuscript needs substantial revision before publication.","major_comments":[{"comment":"The claimed comparison at \"the same access latency\" is not supported as stated. The text explicitly says that for Baselines 1 and 2, which use a \"pilot+data\" frame, the time slot overhead T refers to the pilot signal only, whereas for the proposed scheme T is the total number of OFDM symbols in the entire uplink data frame. The baselines then require additional data symbols for the BER that is plotted, so the total latency of Baselines 1 and 2 exceeds T; conversely, the proposed scheme requires a beacon broadcast period before each frame (Section II) that is not charged to its latency. Thus the comparisons in Figs. 3 and 5 do not hold either the end-to-end latency or the total time-frequency resources fixed. The reported ADEP/BER gains could therefore be artifacts of unequal latency budgets rather than of the algorithmic design. Please redo the comparison under a common latency definition, for example by counting pilot-plus-data symbols for Baselines 1/2 and beacon-plus-data symbols for the proposed scheme, or by comparing against baselines at their total end-to-end frame length.","section":"Section V, Figs. 3–5"},{"comment":"Equation (5) assumes that the pre-equalized beacon-antenna channel is exactly H_η,:,: ◦ Θ = 1, so that Y_η,:,: = S X + W. However, the pre-equalization defined in Eq. (2) includes a nulling threshold h0: for subcarriers with |H_η,m,k| < h0, the pre-equalization factor θ_m,k is set to 0, and the product H_η,m,k θ_m,k is 0 rather than 1. Consequently, the received signal at the beacon antenna is not exactly S X + W on those subcarriers, and the coarse data detection model is invalid as stated. The paper needs to either incorporate the nulled subcarriers into the signal model (e.g., via a mask matrix) or justify that the fraction of nulled subcarriers is negligible under the simulation parameters, ideally by quantifying its effect on ADEP and BER.","section":"Section IV-A, Eq. (5)"},{"comment":"The load-bearing assumption that UEs have perfect downlink channel estimates and perfect uplink/downlink reciprocity is stated only as \"we suppose the CE at UEs' side is sufficiently accurate, i.e., Ĥ = H.\" This assumption is essential for Eq. (5) and for the entire coarse detection stage. The paper provides no sensitivity analysis with respect to channel estimation error, reciprocity calibration mismatch, or beacon SNR. Since user-side channel estimation is performed on a single beacon signal and the users are low-complexity IoT devices, this assumption is not automatically satisfied. Please add simulations that perturb the user-side channel estimate (e.g., with Gaussian estimation error or a bounded reciprocity mismatch) and show how ADEP/BER degrade, or discuss the beacon SNR and channel coherence conditions under which the assumption is reasonable.","section":"Section III"},{"comment":"The AMP update for Z in Eq. (7) appears to contain an error. As written, it is Z_i_m,t = ∑_k s_m,k x̂_i_k,t − V_i_m,t / [σ^2 + V^{i-1}_m,t (Y_η,m,t − Z^{i-1}_m,t)]. In standard AMP, the residual term is V_i_m,t (Y_η,m,t − Z^{i-1}_m,t)/(σ^2 + V^{i-1}_m,t), i.e., the denominator is a scalar variance term and the residual is a separate factor. The current expression places the residual inside the denominator, leading to a complex, dimensionally inconsistent denominator. This error propagates to Algorithm 1 and the subsequent derivations in Eqs. (8)–(12). Please correct the formula or clarify the intended notation; as stated, the algorithm is not well-defined.","section":"Section IV-A, Eq. (7)"},{"comment":"The large-scale fading parameter g_k is used inconsistently. In Eq. (1) and Section II, g_k is a large-scale channel fading factor that multiplies the small-scale fading in the form √g_k H:,m,k, which implies g_k is a linear (power) gain. In Section V, the authors write \"the large scale fading follows the Log-distance path loss model as g_k = 128.1 + 37.6 log10(d_k)\", which is a formula for path loss in dB, not a linear gain factor. If interpreted literally, g_k is on the order of 100 or more, and √g_k is not a valid multiplicative channel amplitude; the power control factor p_k = 1/g_k would also be ill-defined. This needs to be corrected, likely by defining a dB path-loss PL_k and then setting g_k = 10^{-PL_k/10} (for power) or 10^{-PL_k/20} (for amplitude), with a matching power-control formula. As written, the simulation parameter setup is not reproducible.","section":"Section V, simulation parameters"}],"minor_comments":[{"comment":"The simulation results do not report error bars, confidence intervals, or the number of Monte Carlo realizations over which the ADEP/BER/NMSE curves are averaged. Since the channel and data realizations are random, please add this information to the figures or text.","section":"General"},{"comment":"There is a baseline-numbering inconsistency between the table and the text. In Table I, Baseline 3 is the pre-equalization aided JADD scheme with a single antenna and OAMP, and Baseline 4 is the SOMP variant. In the text, however, \"Baseline 3\" is first described as \"Same as Baseline 1 except that JADCE is performed with SOMP\", which appears to be a typo for Baseline 2. Please align the text with Table I.","section":"Section V, Table I and text"},{"comment":"The definition of the average pre-equalization power p_e is not used consistently. The paper says E[|θ_m,k|^2] = p_e and sets the UE transmit power to ρ = 7 dBm, but it does not explain how p_e and the total transmit power constraint are related, nor whether the nulling operation changes the average power normalization. Please clarify the power model.","section":"Section III, power control"},{"comment":"The beacon broadcast period is mentioned but its duration, periodicity, and resource overhead are not quantified. This is relevant to the latency comparison in major comment 1, but also to the practical framing of the scheme.","section":"Section II, beacon overhead"},{"comment":"The text notes that the proposed scheme's performance is not shown for T ≥ 50 because GMMV-AMP does not converge when T ≥ K̂_a, and that LMMSE could be used instead. Since T = 50 is within the plotted range in Fig. 5a (up to 90), please either include the LMMSE-based results for large T, or clearly indicate the range of T for which the proposed receiver is applicable.","section":"Section V, Fig. 5a"}],"recommendation":"major_revision","confidential_remarks":"The paper addresses a relevant problem and has a plausible algorithmic framework, but the current numerical comparison does not support the headline claim of \"same access latency\". The issues with Eq. (7), the nulling inconsistency in Eq. (5), and the simulation parameter handling of g_k are all fixable, but they require a careful revision and rerun of the simulations. I would not recommend rejection, provided the authors can demonstrate the claimed gains under a fair latency comparison and validate the robustness of the pre-equalization assumption."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read this one for the architecture, not for the headline comparison. The genuinely new piece is taking the single-antenna beacon pre-equalization idea from [13] into mMIMO, adding a nulling threshold on pre-equalization factors and an iterative receiver that does coarse DD on the beacon antenna, then data-aided CE in the angular domain, then LMMSE fine DD. That modular design is sensible: each subproblem is a standard linear model, and reusing AMP, GMMV-AMP, and LMMSE makes the algorithm understandable and reproducible. The provided code is a real plus, and the baselines are the right ones.\n\nThe soft spot is where the stress-test lands, and it lands hard. In Section V the paper states that for Baselines 1 and 2, T is the pilot overhead, while for the proposed scheme T is the whole uplink data frame. The beacon broadcast period is not charged to the proposed scheme's latency. So Figs. 3 and 5 do not compare at the same access latency. The baselines need T pilot symbols plus additional data symbols to produce the BER that is plotted; the proposed scheme needs T data symbols plus the beacon period. The reported gains are therefore not a fair reading of the central claim in the abstract and conclusion. This is not a small omission because the claim is explicitly 'same access latency.'\n\nThe other load-bearing assumption is perfect user-side channel estimation and perfect reciprocity, stated in Section III. Equation (5) only gives the clean SX model if the beacon-antenna channel is exactly equalized. The paper gives no sensitivity analysis for channel estimation error, reciprocity mismatch, or the nulling threshold h0. That is a missing robustness study, and it matters more than the missing error bars. The non-convergence of GMMV-AMP when T >= K_a is at least acknowledged in the text, with a reasonable fallback to LMMSE, so I count that as a minor caveat rather than a flaw.\n\nNet: the algorithmic idea is coherent and probably worth a serious referee, but the numerical evidence for the main claim is currently not valid as presented. The authors need to redo the comparison with total latency or total resource blocks held fixed, add a beacon-period cost, and show the scheme survives imperfect reciprocity. If those come back clean, this is a solid subfield contribution. I would send it to review, with the expectation of major revision.","headline":"A coherent mMIMO extension of beacon pre-equalization with a real comparison flaw: the 'same latency' plots charge the baselines only for pilots, not their full frames.","tokens_in":13393,"tokens_out":2583,"would_cite":true,"duration_ms":23756,"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":"Pre-equalizing a single beacon antenna lets a massive MIMO base station detect user activity and data without any CSI, and the paper shows this beats pilot-based grant-free NOMA at equal access latency.","keywords":["grant-free massive access","massive MIMO","massive non-orthogonal multiple access","pre-equalization","approximate message passing","activity detection","channel estimation","data detection"],"falsifier":"Run the Section V simulation with a controlled mismatch between the channel the users estimate from the beacon and the channel they actually transmit over--for example, adding Gaussian estimation error of increasing variance or a fixed reciprocity calibration offset--and record ADEP and BER. If the coarse detection performance degrades sharply at small mismatches, the scheme's reliance on exact pre-equalization is the deciding factor; if it degrades gracefully, the method is robust.","tokens_in":12303,"feed_emoji":"📡","tokens_out":9075,"duration_ms":68711,"temperature":0.7,"pith_summary":"Massive machine-type communication needs to identify which of many users are active and decode their data with minimal latency, but standard grant-free schemes first estimate a high-dimensional channel from pilot symbols, which costs time and spectrum. This paper argues that the pilot stage can be largely bypassed: the base station activates one beacon antenna, users estimate only that antenna's downlink channel and pre-equalize their uplink signals against it, and the receiver then treats the beacon antenna's received signal as a sparse linear model with known spreading codes and unknown data. From there, an iterative detector alternates data-aided channel estimation and fine data detection across all antennas, exploiting massive MIMO diversity. Simulation results show lower activity-detection error and bit error rate than state-of-the-art pilot-based mMIMO grant-free NOMA baselines at the same access latency, and also gains over the single-antenna beacon baseline. The reason a reader should care is that, if the scheme holds up, grant-free access can move a substantial part of the channel-estimation burden from the base station to the users and cut the latency floor for massive IoT.","feed_headline":"Pre-equalized beacon antenna skips CSI pilots in grant-free mMIMO","feed_subtitle":"Users flatten one base-station antenna's channel, letting the receiver detect activity and data without CSI.","key_machinery":"The machinery is the pre-equalized beacon antenna with a nulling threshold. Each user divides by its estimated beacon-antenna channel on subcarriers where the channel magnitude exceeds $h_0$ and sends zero where it is below, so the beacon antenna's channel is flattened and the received tensor there becomes $SX$ plus noise. This one identity converts a semi-blind problem--both channel and data unknown--into three standard linear models: coarse data detection via approximate message passing (AMP) with a spike-and-slab prior plus EM parameter learning and nearest-neighbor sparsity-pattern updates; data-aided channel estimation via GMMV-AMP in the virtual angular domain; and fine data detection via an LMMSE estimator using all antennas. The iteration between the second and third modules is what extracts the massive MIMO diversity gain.","core_discovery":"The paper's central claim is that pre-equalization, not pilot-based channel estimation, can carry the first step of grant-free massive access in massive MIMO. By having each user multiply its uplink signal by the reciprocal of its estimated channel to the beacon antenna, the beacon antenna's received signal over one frame reduces to $Y_{\\eta,:,:}=SX+W_{\\eta,:,:}$, where $S$ is the known spreading-code matrix and $X$ contains the unknown activity-and-data matrix. This makes joint activity and data detection a compressed-sensing problem solvable without any channel state information at the base station. The paper further claims that after this coarse detection, treating the estimated data as pilots lets the receiver estimate the equivalent channel of all antennas in the virtual angular domain and refine the data with LMMSE combining, and that iterating these last two steps harvests multi-antenna diversity. On its simulation settings, the scheme beats both the single-antenna beacon baseline and the pilot-based JADCE baselines under the same time-frequency overhead.","pith_inferences":["The authors assume users know the beacon-antenna channel exactly; an obvious extension is to quantify ADEP and BER as functions of user-side estimation error or reciprocity calibration mismatch, which will likely show that the coarse detection step is the first place errors appear.","The fairness of the headline comparison rests on giving pilot-based baselines the same time-frequency resources for their pilots as the entire data frame of the proposed scheme; a different pilot-allocation policy for baselines could shift the crossover points.","The same beacon-antenna idea could be extended to several beacon antennas or to cell-free massive MIMO, trading a small pilot overhead for better immunity to deep fades on a single antenna.","Because the coarse detector uses only one antenna, activity detection is diversity-starved; a plausible modification is to add a second beacon antenna or to use the channels estimated by the data-aided CE module to re-check the activity set."],"forward_implications":["The uplink pilot phase can be removed entirely from the random-access frame: the first activity-and-data decision is made from the beacon antenna alone, so access latency is set by the data frame rather than by preamble overhead.","Because the coarse detection step does not use the other antennas, activity detection performance is essentially the same as the single-antenna beacon scheme; the multi-antenna gain appears in channel estimation and fine data detection.","The data-aided channel estimation problem is sparse in the virtual angular domain, so standard compressed-sensing solvers apply, and the scheme is most attractive at very low frame lengths where pilot-based JADCE fails to converge.","When the frame length meets or exceeds the number of detected active users, the GMMV-AMP-based channel estimation becomes overdetermined and must be replaced by an LMMSE-style estimator, a regime the paper counts as covered."],"supporting_citations":[{"why":"Supplies the GMMV-AMP algorithm used in the data-aided channel estimation module and as the JADCE solver in Baseline 1.","marker":"[4]"},{"why":"Provides the AMP-based joint activity and data detection formulation and the spike-and-slab prior used in the coarse detection module.","marker":"[12]"},{"why":"The single-antenna beacon-aided pre-equalization scheme that the paper extends to mMIMO and uses as Baselines 3 and 4.","marker":"[13]"},{"why":"Provides the expectation-maximization Gaussian-mixture AMP framework used to initialize noise variance and sparsity ratio in coarse DD.","marker":"[15]"},{"why":"Supplies the nearest-neighbor sparsity pattern learning used to update the sparsity ratio in coarse DD.","marker":"[16]"},{"why":"Gives the AMP message-passing update rules on which the coarse DD factor-node and variable-node equations are based.","marker":"[17]"},{"why":"Provides SOMP, the greedy compressed-sensing algorithm used in Baselines 2 and 4.","marker":"[18]"},{"why":"Gives the LMMSE estimator used for fine data detection.","marker":"[19]"}],"fun_headline_variants":["Pre-equalized beacon makes grant-free mMIMO pilot-free","Skip CSI pilots: pre-equalized beacon for massive MIMO","Grant-free mMIMO with pre-equalized beacon: no pilot CSI","Beacon pre-equalization sidesteps CSI estimation in grant-free mMIMO","Pilotless grant-free access in mMIMO via pre-equalized beacon"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that each user's estimate of the beacon-antenna downlink channel is perfect and that downlink/uplink reciprocity holds exactly, so the pre-equalization truly flattens that channel; if estimation noise, calibration mismatch, or transmit-power limits break equation (5), the CSI-free coarse detection loses its foundation.","fun_headline_variants_meta":{"raw":{"variants":["Pre-equalized beacon makes grant-free mMIMO pilot-free","Skip CSI pilots: pre-equalized beacon for massive MIMO","Grant-free mMIMO with pre-equalized beacon: no pilot CSI","Beacon pre-equalization sidesteps CSI estimation in grant-free mMIMO","Pilotless grant-free access in mMIMO via pre-equalized beacon"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001297,"raw_usage":{"total_tokens":5354,"prompt_tokens":1065,"completion_tokens":4289,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":681,"completion_tokens_details":{"reasoning_tokens":4193}},"tokens_in":681,"tokens_out":4289,"duration_ms":25978,"temperature":1.0,"reasoning_tokens":4193,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T16:16:00.548651+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the Section V simulation with a controlled mismatch between the channel the users estimate from the beacon and the channel they actually transmit over--for example, adding Gaussian estimation error of increasing variance or a fixed reciprocity calibration offset--and record ADEP and BER. If the coarse detection performance degrades sharply at small mismatches, the scheme's reliance on exact pre-equalization is the deciding factor; if it degrades gracefully, the method is robust.","supporting_citations":[{"cited_title":"Compressive sensing-based adaptive active UE detection and channel estimation: Massive access meets massive MIMO,","cited_arxiv_id":null,"evidence_quote":"Supplies the GMMV-AMP algorithm used in the data-aided channel estimation module and as the JADCE solver in Baseline 1."},{"cited_title":"Approximate m essage passing-based joint sser activity and data detection for NO MA,","cited_arxiv_id":null,"evidence_quote":"Provides the AMP-based joint activity and data detection formulation and the spike-and-slab prior used in the coarse detection module."},{"cited_title":"Compressive sensing-based joint activity and data detec - tion for grant-free massive IoT access,","cited_arxiv_id":null,"evidence_quote":"The single-antenna beacon-aided pre-equalization scheme that the paper extends to mMIMO and uses as Baselines 3 and 4."},{"cited_title":"Expectation-Maximization Gaussian-mixture approxi- mate message passing,","cited_arxiv_id":null,"evidence_quote":"Provides the expectation-maximization Gaussian-mixture AMP framework used to initialize noise variance and sparsity ratio in coarse DD."},{"cited_title":"Approximate Message Passing with Nearest Neighbor Sparsity Pattern Learning","cited_arxiv_id":"1601.00543","evidence_quote":"Supplies the nearest-neighbor sparsity pattern learning used to update the sparsity ratio in coarse DD."},{"cited_title":"Message passing algorithms for compressed sens- ing: I. motivation and construction,","cited_arxiv_id":null,"evidence_quote":"Gives the AMP message-passing update rules on which the coarse DD factor-node and variable-node equations are based."},{"cited_title":"On the noise robustness of simultaneous orthog- onal matching pursuit,","cited_arxiv_id":null,"evidence_quote":"Provides SOMP, the greedy compressed-sensing algorithm used in Baselines 2 and 4."},{"cited_title":"Quasi-synchronous random access for massive MIMO- based LEO satellite constellations,","cited_arxiv_id":null,"evidence_quote":"Gives the LMMSE estimator used for fine data detection."}],"review_version":1}