{"id":"16e2e8a8-eb67-4779-9610-976483e7494a","arxiv_id":"2502.05711","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"RIS-assisted OFDM physical-layer network coding with UE power control enables 4-QAM and 16-QAM two-way relay communication, with BER gains that grow strongly with the number of reflecting elements.","lead":"This paper uses smart reflecting surfaces (RISs) to align two users' signals at a relay, enabling faster network-coded data exchange with higher-order QAM. It reports that more surface elements sharply reduce required transmit power, though the abstract's specific '200% SNR improvement' claim conflicts with the paper's own curves.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The RIS phase shifts in Eq. (5) are frequency-flat, but OFDM subcarriers see frequency-selective channel phases; a single phase cannot make α_A and α_B real on all subcarriers, so the multipath claim is unsupported.","rationale":"The reader's weakest assumption focused on CSI accuracy and reciprocity, which are practical limitations that the paper itself partially acknowledges. My concern is more fundamental: even under perfect CSI with no reciprocity issues, the proposed RIS phase alignment in Eq. (5) is inherently frequency-flat, while the OFDM subcarriers experience frequency-selective phases in any true multipath channel. The paper claims to address a 'multipath fading scenario' and uses OFDM for that purpose, yet the equations treat α_A and α_B as subcarrier-independent and real. If the simulation used a flat Rayleigh channel, then the OFDM setup is superfluous and the multipath claim is unsubstantiated; if it used a frequency-selective channel, the core alignment mechanism fails at all but one subcarrier. Neither case supports the advertised generality. The reader's concerns about the 200% SNR claim and missing simulation details are valid, but they are secondary to this modeling gap. A simple per-subcarrier BER test would settle the issue. Until then, the central quantitative results cannot be evaluated, so the verification status should move to UNVERDICTED rather than CONDITIONAL.","tokens_in":9447,"tokens_out":21033,"duration_ms":216313,"concrete_test":"Run the same simulation with a frequency-selective two-tap channel (e.g., 50–100 ns delay spread) or with the original simRIS channel impulse response. Set the RIS phase shifts using Eq. (5) with the channel estimate at one reference subcarrier (say the center, k=32), and keep the same θ for all subcarriers. Compute per-subcarrier BER for 4-QAM and 16-QAM. If the edge-subcarrier BER degrades by more than an order of magnitude compared with the center subcarrier, the reported average BER and the 'multipath robustness' claim are not supported. Also report the per-subcarrier values of ∠α_A[k]−∠α_B[k] to verify whether the power-and-phase equalization condition holds.","verdict_should_be":"UNVERDICTED","load_bearing_attack":"The central mechanism requires α_A and α_B in Eq. (10) to be real, equal, subcarrier-independent coefficients. Equations (8)–(9) achieve this by setting θ_{m,i} = −(∠h_{m,i} + ∠g_{m,i}) per Eq. (5). This is valid only if the cascaded channel h_{m,i} g_{m,i} has the same phase across all OFDM subcarriers. In a frequency-selective multipath channel, h_{m,i}[k] and g_{m,i}[k] vary with k because of path delays; each RIS element can impose only one phase shift θ_{m,i} for the entire band. Consequently, at subcarriers other than the reference, each term keeps a residual phase e^{j(φ_i[k]−φ_i[k0])}, so α_A[k] and α_B[k] become complex and frequency-dependent. The power control in Eqs. (13)–(15), which solves for a single scalar power ratio from |α_A| and |α_B|, cannot equalize amplitudes on every subcarrier. The paper's text claims a 'multipath fading scenario' and uses OFDM (64 subcarriers, 10 MHz) but never specifies subcarrier-dependent RIS phases. If the simRIS channel model includes delay spread, the reported BER curves are likely optimistic; if the simulation used flat Rayleigh channels, the OFDM/multipath claim is not demonstrated. This is independent of the CSI error issue: even with perfect CSI, the frequency-flat limitation remains.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a double-RIS-assisted OFDM physical-layer network coding (PNC) system for M-QAM signals. The key idea is to use RIS phase shifts, configured according to Eq. (5), and a UE power-control mechanism (Eqs. 11-15) so that the two users' signals arrive at the relay with equal power and phase rotation, enabling unambiguous PNC mapping via modular addition. The system is evaluated by simulation for 4-QAM and 16-QAM, RIS sizes from 1 to 256, and under channel estimation error (CEE). The headline claim is that increasing the RIS size from 1 to 256 at BER=10^-3 triples the SNR (a 200% improvement).","tokens_in":9738,"tokens_out":7869,"duration_ms":79299,"significance":"If validated, the paper offers a practical way to offload phase synchronization in PNC from the UEs to RISs, extending prior BPSK-oriented RIS-PNC work to higher-order QAM. The system model is explicit, the power control equations are detailed, and the use of the public simRIS channel simulator and an external PNC mapping reference makes the work partially reproducible. The CEE sensitivity study is also a useful contribution. However, the central claim of a multipath/OFDM-capable system is undermined by the use of frequency-flat RIS phase shifts, and the headline quantitative result is not supported by the paper's own reported numbers. These issues need substantial revision before the contribution can be considered sound.","major_comments":[{"comment":"The RIS phase-shift design in Eq. (5) is frequency-flat: each element imposes a single phase shift θ_m,i, but the OFDM received signal is expressed in the frequency domain with subcarrier index k in Eq. (10). In a frequency-selective multipath channel, h_m,i[k] and g_m,i[k] vary with k, so a single phase shift cannot cancel the channel phase on every subcarrier. Consequently, α_A and α_B become complex and frequency-dependent in general, and the equalization conditions (11)-(15), which use scalar power values and the assumption ∠α_A = ∠α_B, cannot hold on all subcarriers. The paper's claim that the scheme 'transforms a multipath fading channel into an effective AWGN channel' is therefore not established. The authors must either specify a flat-fading channel model (which makes the OFDM framework unnecessary) or extend the model to account for per-subcarrier residual phase, which is a nontrivial change to the power-control and PNC-mapping formulation.","section":"§III, Eqs. (5)-(10)"},{"comment":"The abstract and conclusion state that at BER=10^-3, increasing the RIS size from one to 256 'triples the SNR (i.e., a 200% improvement)'. In §IV the text reports, for 16-QAM at BER=10^-4, that the required transmit power falls from 55 dBm (L=1) to -1.5 dBm (L=256), a reduction of about 56.5 dB. A factor of 3 (4.77 dB) is inconsistent with the order of magnitude of the reported power savings. The authors should either derive the 200% figure directly from the BER curves at BER=10^-3, correct the number, or remove this quantitative claim. As written, the headline result is not supported by the paper's own data.","section":"Abstract and §V (Conclusion)"},{"comment":"Equation (16) is dimensionally inconsistent. From Eq. (7) and the power control relations (11)-(15), the superposed signal should be sqrt(P_A) α_A X_A[k] + sqrt(P_B) α_B X_B[k] = sqrt(P_max)|α_B|(e^{j∠α_A} X_A[k] + e^{j∠α_B} X_B[k]) + N_R[k]. The factor γ² P_max in the printed Eq. (16) has units of power and cannot directly multiply the signal terms. The authors should correct this derivation, as it is the basis for the PNC mapping and BER analysis.","section":"§III, Eq. (16)"},{"comment":"The claim that 'our proposed PNC system attains the same BER-CEE variance performance regardless of the RIS size and modulation order' is confounded, because the four curves in Fig. 5 are obtained at different transmit powers, as stated in the legend. The curves are deliberately shifted so that all start at BER=10^-4 when CEE=-110 dBm, but this does not establish that the system's CEE sensitivity is independent of RIS size and modulation order; a properly normalized comparison (e.g., same received SNR or same transmit power) is needed before this conclusion can be drawn.","section":"§IV, Fig. 5"}],"minor_comments":[{"comment":"The phrase 'to censure power control' appears to be a typo for 'to ensure power control'.","section":"§III, before Eq. (12)"},{"comment":"The notation 'γ2Pmax' is ambiguous; it should be written as γ² P_max.","section":"§III, Eq. (16)"},{"comment":"The equivalence between the L=1 benchmark and 'a scenario where the RIS is replaced with a random scatterer and the phase synchronization is performed at UEs' is not fully justified: a single RIS element with a controllable phase is not identical to a random scatterer. Please clarify how the L=1 case is configured and why this benchmark is fair.","section":"§IV, benchmark description"},{"comment":"The simulation section does not specify how the simRIS channel model is used to generate frequency-selective channel coefficients for the 64 OFDM subcarriers. If simRIS produces single narrowband coefficients, the paper should state explicitly that the channel is flat; if not, the subcarrier-dependent channel phases should be described.","section":"§III and §IV"},{"comment":"The PNC mapping via modular addition is referenced to the authors' prior work [6], but the present paper does not derive or describe the mapping in sufficient detail for a self-contained journal article. Please include a concise description of the modular-√M addition and its decoding rule.","section":"§I and references"}],"recommendation":"major_revision","confidential_remarks":"This manuscript is a conference-style paper (GLOBECOM 2023) with a promising concept but two load-bearing problems: (1) the frequency-flat RIS phase configuration is inconsistent with the OFDM/multipath framework, and (2) the headline SNR improvement number is not supported by the reported data. The first issue can be addressed by reframing the work as a narrowband (single-carrier) PNC system and scaling back the OFDM claims, or by a substantial reformulation for wideband channels. The paper also leans heavily on the authors' own prior work [6] for the PNC mapping, so the incremental contribution over [6] should be clarified. If the authors correct these points and resubmit, the paper could become a useful contribution; in its current form, the main conclusions are not reliable. I recommend major revision rather than rejection, since the central idea (RIS-based phase synchronization for PNC with power control) is defensible in a narrower scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick read on arXiv:2502.05711. The paper combines two existing ideas: RIS-assisted phase synchronization from [14] and OFDM-PNC with modular-addition mapping from [6], and adds UE power control to handle different path losses. That specific combination is new, and it's a plausible way to offload phase alignment from UEs to infrastructure. The system model equations (1)-(16) are internally consistent, and the CEE sensitivity study is a useful addition. I believe the qualitative framework is sound.\n\nThe soft spots are real, though. First, the frequency-flat RIS phase assumption. Equation (5) sets a single phase per RIS element to cancel the cascade phase, and (8)-(9) then treat αA and αB as real, scalar coefficients. That only works if each h_{m,i} and g_{m,i} is frequency-flat, i.e., single-tap. But the paper claims a multipath fading scenario and uses OFDM with 64 subcarriers. In a frequency-selective channel, each subcarrier sees a different cascade phase, and one RIS phase cannot align all subcarriers. The paper never specifies whether the simRIS channel includes delay spread. If it does, the BER curves are optimistic; if it does not, the multipath claim is unsupported. Either way, the paper needs to state this clearly.\n\nSecond, the headline claim about '200% SNR improvement' from L=1 to L=256 does not match the paper's own Fig. 3. At BER=1e-4, the required transmit power drops from 55 dBm to -1.5 dBm, which is about 56 dB of power reduction, consistent with an array gain of L^2 (48 dB) plus other effects. A 200% improvement in SNR means a factor of 3 (4.8 dB). Those are different statements. The abstract's claim may refer to a different operating point, but it is not tied to a specific curve or derivation. It should be fixed.\n\nThere are also minor issues: no error bars or Monte Carlo details, the L=1 benchmark is described as equivalent to UE-side synchronization but that scenario is not actually simulated, and the PNC mapping is deferred to [6]. For a conference paper, some of this is acceptable, but the quantitative claims need more support.\n\nWho is this for? People working on RIS-assisted relay design and PNC for 6G. It is a credible incremental step, not a breakthrough. I think it deserves a serious referee: the core idea is sensible, the equations are coherent, and the CEE study is valuable. The referee should push for clarification of the channel model and reconciliation of the SNR claim. My verdict: conditional, with major revision.","headline":"A sensible incremental combination of RIS phase alignment with OFDM-PNC for M-QAM, but its headline SNR claim and channel model need a hard look before the numbers are quoted.","tokens_in":10366,"tokens_out":4072,"would_cite":false,"duration_ms":39831,"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":"This paper claims that two independently tuned reconfigurable intelligent surfaces plus a simple power-control rule can synchronize and balance the two uplink streams in physical-layer network coding, letting a relay detect superposed…","keywords":["Reconfigurable Intelligent Surface","Physical-Layer Network Coding","Quadrature Amplitude Modulation","OFDM","Phase Synchronization","Ambiguity Removal","Channel Estimation Error","28 GHz"],"falsifier":"Run the same double-RIS PNC simulation with intentional phase estimation errors drawn from a Gaussian of variance -50 dBm (above the paper's -60 dBm threshold) while all other settings match the paper: if the BER for 16-QAM no longer improves as the RIS grows from 64 to 256 elements, or the required transmit power at BER $10^{-4}$ stops falling, then the central SNR-gain claim fails at that CSI quality. A complementary over-the-air test would measure the actual BER with random phase shifts, which the paper reports as stuck near 0.5, to confirm the sensitivity.","tokens_in":9222,"feed_emoji":"📡","tokens_out":12777,"duration_ms":101079,"temperature":0.7,"pith_summary":"Physical-layer network coding lets two users exchange a packet through a relay in two time slots, but it normally demands that the two uplink signals arrive at the relay with the same phase and power; with higher-order QAM that synchronization is especially hard. This paper claims that the job can be offloaded to two reconfigurable intelligent surfaces, one per user, whose phase shifts are set to the negative sum of the estimated channel angles, plus a transmit-power rule that makes the weaker channel transmit at maximum power and the stronger channel scale down. Under that configuration the relay sees a clean superposition of the two M-QAM constellations and can use modular $\\sqrt{M}$ addition to remove the mapping ambiguity. The payoff claimed is large: at 28 GHz and BER $10^{-3}$, increasing the RIS from 1 to 256 elements triples the SNR (a 200% gain), and for 16-QAM at BER $10^{-4}$ the required transmit power drops from 55 dBm to -1.5 dBm. The paper's own sensitivity analysis shows these gains require channel-estimation error variance below roughly -60 dBm.","feed_headline":"Two RISs triple SNR for physical-layer network coding","feed_subtitle":"At 28 GHz, growing the surface from 1 to 256 elements cuts transmit power from 55 dBm to -1.5 dBm for 16-QAM at BER 1e-4.","key_machinery":"The argument rides on three devices. (1) RIS phase configuration: each reflecting element of the two RISs is tuned to the negative sum of the estimated angles of the two channel segments it bridges, Eq. (5), turning each cascaded channel coefficient into a real positive number $\\alpha_A$ or $\\alpha_B$ when CSI is perfect. (2) Power control: with $P_B = P_{\\max}$ for the weaker link, the stronger link transmits at $P_A = \\gamma^2 P_{\\max}$, $\\gamma = |\\alpha_B|/|\\alpha_A| \\leq 1$, so the two arriving streams have equal amplitude and a common phase rotation, Eqs. (11)-(15). (3) Modular PNC mapping: the relay replaces the received symbol with $Z = (X_A + X_B) \\bmod \\sqrt{M}$, a square-root-$M$ modular addition that removes the superposition ambiguity for M-QAM and reduces to XOR for BPSK. Together these convert a random multipath channel into an effective AWGN channel at the relay, which is what lets higher-order QAM constellations stay separable.","core_discovery":"The central claim is that an OFDM-PNC system can be made to work for non-binary M-QAM by turning the two independent fading links between users and relay into two real, equal-gain links. With the phase of each RIS element set as $\\hat{\\theta}_{m,i} = -(\\theta_{\\hat{h}_{m,i}}+\\theta_{\\hat{g}_{m,i}})$ and transmit powers obeying $P_A = \\gamma^2 P_{\\max}$ with $\\gamma = |\\alpha_B|/|\\alpha_A| \\le 1$, the relay's received symbol becomes $\\sqrt{P_{\\max}}\\gamma^2 (e^{j\\angle\\alpha_A} X_A + e^{j\\angle\\alpha_B} X_B) + N_R$, where $\\alpha_A$ and $\\alpha_B$ are real positive sums in the ideal case. This removes the singular fade states that shorten inter-constellation distances, so the relay can apply a modular $\\sqrt{M}$-addition PNC mapping (reducing to XOR for BPSK) and broadcast the result; each user then recovers the peer's symbol by modular subtraction of its own. The paper further claims that the end-to-end BER is dominated by the uplink and that the scheme's sensitivity to channel estimation error is independent of RIS size and modulation order, with a sharp degradation once the error variance passes about -60 dBm.","pith_inferences":["The paper's benchmark with L=1 is essentially a random scatterer plus perfect phase-aligned reception; the 200% SNR gain therefore bundles beamforming gain from more elements with synchronization gain from phase tuning, and separating the two would clarify how much of the improvement is due to RIS size alone.","The sharp CEE threshold around -60 dBm suggests a practical adaptation rule: estimate the current CEE variance online and re-run channel estimation or fall back to lower-order modulation when it approaches the threshold; the paper does not propose such a rule.","Because the whole scheme assumes channel reciprocity within one coherence time, its viability in mobility is tied to how fast the cascaded channel changes; a Doppler-aware update-rate analysis would be a natural extension not covered here.","If the CSI burden can be met, the UEs need no precoding, which shifts complexity from lightweight user devices to the infrastructure side, an architectural consequence the paper leaves implicit."],"forward_implications":["For any given target BER, the required transmit power falls steeply as the RIS grows; 16-QAM at BER $10^{-4}$ needs 55 dBm with a single-element RIS and -1.5 dBm with 256 elements.","At BER $10^{-3}$, going from 1 to 256 RIS elements triples the SNR in the 28 GHz band (a 200% improvement) when CSI is accurate.","Random RIS phase shifts are fatal to PNC: with L=16 and random phases the BER stays around 0.5 for both 4-QAM and 16-QAM, so the phase alignment done by the RIS is load-bearing, not optional.","The BER-versus-CEE-variance curves coincide for L=64 and L=256 and for both modulation orders, which the paper reads as evidence that the synchronous design adapts across RIS sizes and modulations without changing the algorithm.","The same modular-$\\sqrt{M}$ mapping is backward compatible with BPSK (where it is XOR), so the scheme can be introduced gradually into existing PNC deployments."],"supporting_citations":[{"why":"Introduces physical-layer network coding over two-way relay channels and the MA/BC two-phase structure that the whole system builds on.","marker":"[1]"},{"why":"The authors' earlier OFDM-PNC design with higher-order M-QAM; this paper extends it by replacing UE precoding with RIS phase synchronization and adding power control.","marker":"[6]"},{"why":"Supplies the RIS-assisted PNC phase-shift idea and the cascaded channel model that Eq. (5) is based on.","marker":"[14]"},{"why":"Channel modeling perspective for RIS that underlies the simulated cascaded channel gains.","marker":"[18]"},{"why":"The channel simulator implementation used to generate the Rayleigh fading coefficients and path-loss effects in the performance evaluation.","marker":"[19]"},{"why":"Provides the M-QAM de-noising and PNC mapping that the modular square-root-M addition at the relay is based on.","marker":"[20]"}],"fun_headline_variants":["RIS enables M-QAM network coding with 200% SNR gain","256-element RIS triples SNR for PNC at 28 GHz","RIS synchronizes phases to unlock higher-order PNC","RIS-assisted PNC: 200% SNR improvement for M-QAM","Reconfigurable surface cuts transmit power 56 dB for PNC"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire SNR gain rests on the RIS controller having accurate, near-real-time estimates of both cascaded UE-RIS-relay channels so the phase shifts of Eq. (5) truly align the two arriving signals, and on the channels staying reciprocal through one two-phase exchange (uphill and broadcast); the paper's own curves place this requirement at a channel-estimation error variance below roughly -60 dBm.","fun_headline_variants_meta":{"raw":{"variants":["RIS enables M-QAM network coding with 200% SNR gain","256-element RIS triples SNR for PNC at 28 GHz","RIS synchronizes phases to unlock higher-order PNC","RIS-assisted PNC: 200% SNR improvement for M-QAM","Reconfigurable surface cuts transmit power 56 dB for PNC"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001045,"raw_usage":{"total_tokens":4490,"prompt_tokens":1139,"completion_tokens":3351,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":755,"completion_tokens_details":{"reasoning_tokens":3262}},"tokens_in":755,"tokens_out":3351,"duration_ms":22812,"temperature":1.0,"reasoning_tokens":3262,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T18:16:21.635015+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same double-RIS PNC simulation with intentional phase estimation errors drawn from a Gaussian of variance -50 dBm (above the paper's -60 dBm threshold) while all other settings match the paper: if the BER for 16-QAM no longer improves as the RIS grows from 64 to 256 elements, or the required transmit power at BER $10^{-4}$ stops falling, then the central SNR-gain claim fails at that CSI quality. A complementary over-the-air test would measure the actual BER with random phase shifts, which the paper reports as stuck near 0.5, to confirm the sensitivity.","supporting_citations":[{"cited_title":"Hot Topic: Physical Layer Network Coding,","cited_arxiv_id":null,"evidence_quote":"Introduces physical-layer network coding over two-way relay channels and the MA/BC two-phase structure that the whole system builds on."},{"cited_title":"OFDM-based Synchronous PNC Communications Using Higher Order QAM Modulations,","cited_arxiv_id":null,"evidence_quote":"The authors' earlier OFDM-PNC design with higher-order M-QAM; this paper extends it by replacing UE precoding with RIS phase synchronization and adding power control."},{"cited_title":"RIS-Assisted Physical Layer Network Coding Over Two-Way Relay Fading Channels,","cited_arxiv_id":null,"evidence_quote":"Supplies the RIS-assisted PNC phase-shift idea and the cascaded channel model that Eq. (5) is based on."},{"cited_title":"Reconfigurable Intelligent Surfaces for Future Wireless Networks: A Channel Modeling Perspective,","cited_arxiv_id":null,"evidence_quote":"Channel modeling perspective for RIS that underlies the simulated cascaded channel gains."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The channel simulator implementation used to generate the Rayleigh fading coefficients and path-loss effects in the performance evaluation."},{"cited_title":"A new physical layer network coding de-noising mapping based on M-QAM,","cited_arxiv_id":null,"evidence_quote":"Provides the M-QAM de-noising and PNC mapping that the modular square-root-M addition at the relay is based on."}],"review_version":1}