{"id":"b8245825-cb30-4beb-bc2c-39ff0a02f2e9","arxiv_id":"2506.12668","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Under optimized precoders, SIC-free RSMA achieves nearly the same weighted sum-rate and max-min fairness as SIC-based RSMA with finite constellations, across small and large-scale systems.","lead":"This paper derives finite-constellation rate formulas for rate-splitting multiple access with and without SIC receivers, and optimizes precoders for both. It finds that SIC-free receivers suffer only minor performance losses, making them a practical low-complexity option for future wireless systems.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The headline SIC-free-vs-SIC comparison is computed with the approximate rates of Eq. (14)-(16); the paper never states that final figures use exact rates, and the approximation error can differ between receivers.","rationale":"Reading in good faith: the rate derivations, Proposition 1, Proposition 2/Algorithm 3, and the null-space dimensionality reduction are internally coherent; the contribution is a plausible algorithmic extension. However, the paper's own strongest claim is the numerical comparison, and that comparison has not been shown to use true achievable rates. The reader's weakest assumption identifies the same point: (6) is a surrogate, and no exact-rate verification is reported. I agree with that assessment. The concern is not a disagreement with consensus; it is a correctness risk in the reported numbers. The appropriate response is to keep the CONDITIONAL verdict, adding a requirement that the authors report exact-rate verification at the operating points of Figures 3 and 7-9, and clarify which formula generated each curve. If exact rates confirm the small SIC-free loss, the claim is substantially strengthened; if not, the conclusion may need to be softened to 'minor losses in the approximate rate'.","tokens_in":21038,"tokens_out":7683,"duration_ms":93346,"concrete_test":"Take the K=2, NT=2, QPSK/8QAM setup of Figure 3(a). At SNR = 10 dB and 20 dB, run Algorithm 1 to convergence, freeze the returned precoders, and compute the exact achievable rates using (7)-(9) with the noise expectation in (5) estimated by Monte Carlo over at least 10^5 noise realizations (or by numerical integration). Compare SIC-based and SIC-free exact rates with the approximate values (14)-(16). If the SIC-based minus SIC-free gap, or the common/private split, changes by more than 0.1 bits/channel-use at either SNR, the headline claim is not supported unless the figures are re-plotted with exact rates. Also state explicitly in the paper which formula generated each figure.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central numerical claim, that optimized SIC-free RSMA sacrifices little compared with SIC-based RSMA, rests on rates computed from the Jensen approximation (6)/(14)-(16). Section IV states that the approximation is used 'during the optimization process,' but Section VI never states whether Figures 3, 6-9 report exact rates (5)/(7)-(9) or the same approximations. Because (6) is obtained by moving the noise expectation inside the logarithm and rescaling the squared distance by 1/2, it is not exact, and its error is SNR-, constellation-, and interference-dependent. Nothing in the paper quantifies this error at the operating points of the figures, and no code is provided to reproduce the exact-rate evaluation. If the reported curves use (14)-(16), then the optimized precoders are only optimal for the surrogate, and the 'minor loss' of SIC-free RSMA could be an artifact of a receiver-dependent approximation error rather than a property of true achievable rates. This is the load-bearing assumption for the main claim and it is currently unchecked.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies downlink rate-splitting multiple access (RSMA) under finite-alphabet inputs and compares two receiver architectures: SIC-based and SIC-free. It derives constellation-constrained rate expressions for both architectures (Eqs. (7)-(9)), introduces a Jensen-based approximation (Eqs. (14)-(16)), and proposes projected subgradient ascent (PSGA) algorithms for weighted sum-rate and max-min fairness optimization, with closed-form common-rate allocations in Propositions 1 and 2. To handle large-scale systems, it adds user grouping, null-space projection, and dimensionality reduction. Numerical results claim that SIC-free RSMA incurs only minor losses relative to SIC-based RSMA under optimized precoders, extending prior conclusions from low-complexity designs to fully optimized ones.","tokens_in":21223,"tokens_out":8030,"duration_ms":103859,"significance":"If the numerical conclusion is confirmed, the paper is valuable: it extends the SIC-free RSMA literature from heuristic low-complexity precoders to optimized precoders, covers both WSR and MMF objectives, and offers a practical complexity-reduction path for large-scale systems. The rate derivations and the closed-form common-rate allocations (Propositions 1 and 2) are sound and are useful beyond finite constellations. The large-scale complexity analysis is also a strength. However, the central performance claim currently rests on an unvalidated surrogate: the paper does not state whether the final figures use exact rates or the Jensen approximation, and no code or exact-rate validation is provided. The comparison between SIC and SIC-free receivers is therefore not yet established at the level the abstract claims.","major_comments":[{"comment":"The paper never states whether Figures 3, 6, 7, 8, and 9 plot the exact rate expressions (7)-(9) or the Jensen-approximate expressions (14)-(16). Section IV explicitly says the approximation is used 'during the optimization process,' but Section VI contains no corresponding statement for the evaluation phase. If the plotted curves use (14)-(16), then the headline SIC-free-versus-SIC comparison is between two surrogates, and the receiver-dependent approximation error could either create or mask the reported 'minor losses.' Please state explicitly which rates are plotted, and re-run the key figures with the exact rates (7)-(9) at least for the operating SNRs and constellation modes used in Figures 3, 7, 8, and 9.","section":"VI.C and VI.E, with Eqs. (14)-(16)"},{"comment":"The approximation in Eq. (6) is obtained by moving the noise expectation inside the logarithm and rescaling the squared distance by 1/2; it is not exact, and its error depends on SNR, constellation cardinality, and the interference structure. No experiment in Section VI quantifies the gap between (14)-(16) and (7)-(9) at the operating points. Because the approximation error can differ between SIC and SIC-free receivers, a validation figure or table is needed to show that the optimized precoders and the reported rate gaps reflect the true constellation-constrained mutual information rather than an artifact of the surrogate.","section":"III.A, Eq. (6), and VI"},{"comment":"Proposition 3 states that every non-trivial stationary point of P1 and P3 lies in the range of [h1,...,hK]. This claim is stronger than what the dimensionality reduction requires and is false as stated. For example, with K=1, h=[1;0]^T, and P=[0;1]^T, the objective depends only on h^H P, which vanishes, so the gradient with respect to P is zero and P is a stationary point, yet P is outside the range. The argument only needs existence of a global optimum in the range, which is true by projecting the orthogonal component to zero and renormalizing. Please restate the proposition in the existential form and provide the proof rather than omitting it.","section":"V.B, Proposition 3"}],"minor_comments":[{"comment":"The sentence describing the channel model contains a duplicated word: 'non-LoS channel component and and the Rician factor.'","section":"VI.A"},{"comment":"The mutual information expressions in Eq. (18) omit the conditioning on h_k^H P that is used consistently in Eqs. (10)-(13); adding it would improve notational consistency.","section":"III.B, Eq. (18)"},{"comment":"The proof of Proposition 2 assumes Rp,1 < Rp,2 < ... < Rp,K without discussing ties. Algorithm 3 appears to handle ties, but the proof should state how equal private rates are treated.","section":"Appendix B"},{"comment":"The transmission-mode dictionaries use constellations such as '8QAM' and '512QAM' that are not standard square QAM; please define these constellations or cite the specific signal sets used.","section":"Tables III-VI"},{"comment":"Figure 5 plots CPU time rather than iteration count; the convergence claim would be easier to interpret if iteration counts were also reported, or if the machine and implementation details were given.","section":"Figure 5"}],"recommendation":"major_revision","confidential_remarks":"The main risk is that the central comparison between SIC-based and SIC-free RSMA is evaluated with the Jensen surrogate, and no exact-rate check is reported. This is fixable within the manuscript's scope by re-evaluating the key figures with exact rates and possibly adjusting the conclusions. The derivations and closed-form allocations are solid. I would not reject on the current evidence, but the numerical claims need to be made verifiable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a solid extension of the authors' earlier SIC-free RSMA result, moving from zero-forcing precoders to arbitrary precoders and adding WSR/MMF optimization plus a large-scale grouping scheme. The rate derivations look correct, and the closed-form common-rate allocations are genuinely useful. The main thing to check before relying on the performance claims is whether the figures use the exact constellation-constrained rates or the Jensen approximation.\n\nThe general rate expressions (7)-(9) are a real generalization of [19], and Proposition 1 is clean and holds for any input distribution. Proposition 2's closed-form solution for the common-rate allocation is practical and well proved. The complexity reduction via grouping and null-space projection is sensible, and Table II shows a real scaling improvement. The numerical comparison with GQMP suggests the proposed PSGA is competitive.\n\nThe load-bearing concern is the approximation. The paper says the approximation (14)-(16) is used 'during the optimization process,' but Section VI never states whether the reported curves evaluate the exact rates (5)/(7)-(9) or the same approximation. The Jensen bound is not exact, and its error is SNR-, constellation-, and receiver-dependent. If the figures use the approximation, the 'minor loss' of SIC-free versus SIC could be an artifact of a receiver-dependent approximation error rather than a property of true achievable rates. This is not a fatal flaw - the approach is standard in the finite-alphabet precoding literature, and prior works have used it - but it needs to be addressed: either a quantification of the gap at the operating points, or a re-evaluation of at least the main figures with exact rates, or released code. Without that, the headline conclusion is conditional.\n\nAlso minor: the abstract says 'evaluates the theoretical limits,' but the optimization is local (projected subgradient ascent), and the paper itself acknowledges convergence guarantees are non-trivial for non-convex problems. The title's 'theoretical limits' overstates what a heuristic algorithm can deliver.\n\nFor someone working on RSMA or finite-alphabet precoding, this is a useful paper that extends the design space. The core result - SIC-free RSMA can be close to SIC-based RSMA with optimized precoders - is plausible and consistent with earlier [19] findings, but the numerics as presented do not fully pin it down. I would send it to peer review, with a clear request for the authors to disclose whether final figures use exact or approximate rates and to validate the approximation error.","headline":"Competent extension of SIC-free RSMA to optimized precoders, but the headline comparison likely rides on an unvalidated Jensen approximation; worth refereeing with a demand for exact-rate checks.","tokens_in":21757,"tokens_out":1722,"would_cite":true,"duration_ms":20016,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["94A15","94A40","90C26"],"pacs":[],"model":"deepseek-v4-flash","headline":"Under optimized precoders, SIC-free RSMA nearly matches SIC-based RSMA, keeping most of the sum-rate and fairness gains with a simpler receiver.","keywords":["rate-splitting multiple access","finite-alphabet signaling","SIC-free receivers","precoder optimization","weighted sum-rate","max-min fairness","massive MIMO","constellation-constrained rates"],"falsifier":"Evaluate the exact constellation-constrained mutual information at the precoders returned by the proposed algorithms for a small case (e.g., NT = 2, K = 2, QPSK) across the SNR range 5–30 dB and compare it with the approximate rates used in optimization; a large gap would indicate that the optimized precoders and reported SIC-free gains do not reflect true rates.","tokens_in":20840,"feed_emoji":"📡","tokens_out":2821,"duration_ms":35576,"temperature":0.7,"pith_summary":"The paper asks whether rate-splitting multiple access (RSMA) can forgo successive interference cancellation (SIC) without costly performance loss when signals are drawn from finite constellations rather than Gaussian alphabets. It derives exact and approximate constellation-constrained rate expressions for both receiver types with no restriction on the precoders, then optimizes precoders under weighted sum-rate and max-min fairness objectives. Numerical results show that SIC-free RSMA preserves most of the superiority of SIC-based RSMA over SDMA and NOMA, with only minor losses, across small-scale and large-scale massive MIMO settings. If true, this makes SIC-free RSMA a practical candidate for future wireless systems where receiver simplicity, low latency, and small buffers matter.","feed_headline":"SIC-free RSMA nearly matches SIC receivers under optimized precoders","feed_subtitle":"Finite-constellation rate-splitting without successive interference cancellation keeps most of the sum-rate and fairness gains.","key_machinery":"The load-bearing machinery is the Jensen-based approximation of the constellation-constrained mutual information, which replaces the exact conditional-entropy expression with a tractable closed form involving distances between constellation points, together with projected subgradient ascent for the non-convex precoder optimization. Around this core sit a closed-form global optimal common-stream allocation (Algorithm 3), a log-sum-exp smoothing of the max-min objective, and, for large-scale systems, user grouping with null-space projection plus a low-dimensional subspace property that collapses the per-group precoder dimension to two. These components convert an exponentially complex finite-alphabet design problem into one whose per-iteration cost scales linearly with the number of users.","core_discovery":"The central claim is that, once precoders are optimized using the finite-constellation structure, SIC-free RSMA achieves nearly the same weighted sum-rate and max-min fairness as SIC-based RSMA, even though the two require markedly different precoders and power splits. The paper establishes this by deriving rate expressions for the common and private streams under finite-alphabet inputs without assuming zero-forcing private precoders, and by proving that for weighted sum-rate maximization only the most weighted user's message needs to be split—an allocation result that holds for any input distribution. The accompanying projected subgradient ascent algorithms, combined with a closed-form optimal common-stream allocation, show that the SIC-free loss is minor under optimized designs, extending earlier SIC-free conclusions from low-complexity precoders to fully optimized ones and from sum-rate to fairness criteria.","pith_inferences":["The paper reports approximate rates as the optimization surrogate but does not state whether final figures use the exact or approximate expressions; an editorial read is that the exact-versus-approximate gap at optimized operating points deserves a dedicated comparison before the minor-loss conclusion is taken as quantitative.","The non-monotonic effect of finite-alphabet interference cited in the paper suggests SIC-free RSMA may also be robust in other interference-dominated regimes, such as heterogeneous networks or overloaded cells, where strong structured interference can be less harmful than Gaussian models predict.","The dimensionality-reduction trick could transfer to other finite-alphabet precoding problems beyond RSMA, since the low-dimensional subspace property holds for any objective depending only on the channel-range directions.","A testable extension is to run the proposed algorithms with exact mutual information evaluations in the objective, which would reveal whether the reported SIC-free loss widens, shrinks, or stays unchanged when the Jensen surrogate is removed."],"forward_implications":["If SIC-free RSMA indeed loses little performance under optimized precoders, receiver hardware for RSMA can drop SIC blocks, reducing latency, buffer size, and implementation complexity without sacrificing most of the spectral-efficiency gains.","Dedicated precoding matters: since SIC-free and SIC-based RSMA converge to similar rates with very different precoders and common-stream power allocations, deploying SIC-free RSMA requires re-optimizing precoders rather than reusing SIC-based designs.","For massive MIMO regimes, the user-grouping and null-space-projection approach makes finite-constellation RSMA optimization computationally feasible, so the SIC-free advantage can be assessed and exploited in systems with many antennas and users.","Under max-min fairness, SIC-free RSMA also preserves most of the SIC-based gain, meaning fairness-oriented deployments need not pay a large receiver-complexity premium.","The global optimal common-stream allocation for weighted sum-rate maximization implies that only one user's message needs to be split when the goal is sum-rate, reducing control signaling in practical RSMA implementations."],"supporting_citations":[{"why":"Establishes the prior SIC-free RSMA framework under finite constellations with low-complexity precoders, which this paper extends to fully optimized precoders and WSR/MMF objectives.","marker":"[19]"},{"why":"Supplies the Jensen-based approximation of conditional entropy used throughout the optimization as the tractable surrogate for exact constellation-constrained mutual information.","marker":"[35]"},{"why":"Provides the result that strong finite-alphabet interference can be nearly harmless, motivating why SIC-free decoding may approach SIC-based rates under proper precoding.","marker":"[37]"},{"why":"Introduces the low-dimensional subspace property that the paper adapts to finite-constellation RSMA for dimensionality reduction in large-scale systems.","marker":"[43]"},{"why":"Offers the generalized quadratic matrix programming baseline used to benchmark the convergence and objective value of the proposed projected subgradient ascent.","marker":"[38]"},{"why":"Provides the mutual-information expression for finite-alphabet MIMO broadcast channels that underlies the constellation-constrained rate derivations.","marker":"[29]"}],"fun_headline_variants":["SIC-free RSMA nearly matches SIC with optimized precoders","Optimized precoders make SIC-free RSMA competitive","Rate-splitting without SIC: small gap after precoder design","Finite-constellation RSMA: SIC-free almost as good as SIC","Precoder optimization slashes SIC-free RSMA penalty"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The optimization uses the Jensen-based approximation to the constellation-constrained mutual information as a stand-in for the true rate, and the paper does not report whether final performance numbers evaluate the exact or approximate expressions.","fun_headline_variants_meta":{"raw":{"variants":["SIC-free RSMA nearly matches SIC with optimized precoders","Optimized precoders make SIC-free RSMA competitive","Rate-splitting without SIC: small gap after precoder design","Finite-constellation RSMA: SIC-free almost as good as SIC","Precoder optimization slashes SIC-free RSMA penalty"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000156,"raw_usage":{"total_tokens":1204,"prompt_tokens":920,"completion_tokens":284,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":536,"completion_tokens_details":{"reasoning_tokens":192}},"tokens_in":536,"tokens_out":284,"duration_ms":3352,"temperature":1.0,"reasoning_tokens":192,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T00:44:35.143235+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate the exact constellation-constrained mutual information at the precoders returned by the proposed algorithms for a small case (e.g., NT = 2, K = 2, QPSK) across the SNR range 5–30 dB and compare it with the approximate rates used in optimization; a large gap would indicate that the optimized precoders and reported SIC-free gains do not reflect true rates.","supporting_citations":[{"cited_title":"Rate- splitting multiple access: Finite constellations, receiver design, and SIC- free implementation,","cited_arxiv_id":null,"evidence_quote":"Establishes the prior SIC-free RSMA framework under finite constellations with low-complexity precoders, which this paper extends to fully optimized precoders and WSR/MMF objectives."},{"cited_title":"A low-complexity design of linear pre- coding for MIMO channels with finite-alphabet inputs,","cited_arxiv_id":null,"evidence_quote":"Supplies the Jensen-based approximation of conditional entropy used throughout the optimization as the tractable surrogate for exact constellation-constrained mutual information."},{"cited_title":"Optimal and suboptimal decoders under finite-alphabet interference: A mismatched decoding perspective,","cited_arxiv_id":null,"evidence_quote":"Provides the result that strong finite-alphabet interference can be nearly harmless, motivating why SIC-free decoding may approach SIC-based rates under proper precoding."},{"cited_title":"Rethinking wmmse: Can its complexity scale linearly with the number of bs antennas?","cited_arxiv_id":null,"evidence_quote":"Introduces the low-dimensional subspace property that the paper adapts to finite-constellation RSMA for dimensionality reduction in large-scale systems."},{"cited_title":"Generalized quadratic matrix programming: A unified framework for linear precoding with arbitrary input distributions,","cited_arxiv_id":null,"evidence_quote":"Offers the generalized quadratic matrix programming baseline used to benchmark the convergence and objective value of the proposed projected subgradient ascent."},{"cited_title":"Linear precoding for MIMO broadcast channels with finite-alphabet constraints,","cited_arxiv_id":null,"evidence_quote":"Provides the mutual-information expression for finite-alphabet MIMO broadcast channels that underlies the constellation-constrained rate derivations."}],"review_version":1}