{"id":"cd04a09b-c870-4914-96b0-b5c46a5de069","arxiv_id":"2606.09753","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":6.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"A jamming-resilient sparse delay-Doppler NOMA scheme with randomized active sets, unitary precoding justified by Marchenko-Pastur conditioning, and Merkle-Hellman superincreasing power allocation achieves no jammer-induced BER floor and exact ML-equivalent SIC performance.","lead":"The paper proposes a sparse delay-Doppler NOMA scheme using random active bin subsets, unitary precoding, and superincreasing power allocation to eliminate jammer-induced error floors. A smart generalist might read it for techniques that could improve reliability of multi-user wireless links under intentional interference.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"Marchenko-Pastur asymptotic may not control conditioning of small sparse unitary submatrices when seed is compromised","rationale":"The reader's weakest_assumption directly identifies the same Marchenko-Pastur step. Because the full text is now available, the concrete_test above can be run against the paper's own parameters and precoders; if the small-matrix conditioning holds, the central no-floor claim survives and the verdict can be raised; otherwise it must remain conditional or be lowered.","tokens_in":1856,"tokens_out":375,"duration_ms":16859,"concrete_test":"Fix the system parameters (e.g., 64 total bins, sparsity 4–8), generate 10^5 independent random unitary submatrices of the corresponding size for each precoder (Hadamard/DFT/Haar), compute their condition numbers, and report the fraction exceeding 20; if this fraction is non-negligible, recompute the closed-form BER under the worst-case conditioning and check whether a floor appears.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The claim that seed compromise leaves BER inside the unjammed envelope rests on a Marchenko-Pastur conditioning argument that is asserted to control any random unitary submatrix. Marchenko-Pastur governs the bulk eigenvalue distribution of Wishart matrices in the large-dimension limit (rows, columns → ∞ with fixed ratio). The scheme, however, activates only a small random subset of delay-Doppler bins per frame; the resulting submatrix therefore has small row dimension equal to the sparsity level. For such finite small matrices the extreme singular values are not guaranteed to remain bounded away from zero and infinity with high probability, so an adversary who knows the seed can in principle select a jammed pattern that produces an ill-conditioned effective channel and induces a BER floor.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The paper proposes a sparse delay-Doppler NOMA scheme for jamming resilience: user data is placed on a small random subset of delay-Doppler bins drawn from a shared pseudo-random seed, spread via a unitary precoder (Hadamard/DFT/Haar), recovered by least-squares after discarding jammed bins, and decoded by SIC. It asserts a closed-form BER with no jammer-induced floor (unlike conventional OTFS-NOMA), that Marchenko-Pastur conditioning ensures any random unitary submatrix remains well-conditioned even under seed compromise, exact equivalence of low-complexity SIC to ML detection for >2 users via superincreasing (Merkle-Hellman) power allocation, an OMA-friendly partitioning rule for >4 users, and extension to Rician fading. Monte Carlo results are stated to track the analysis within 3 dB with large gains versus pattern-aware jammers.","tokens_in":2038,"tokens_out":680,"duration_ms":11651,"significance":"If the central claims hold, the work would offer a concrete, parameter-free approach to eliminating the partial-band jammer floor in OTFS-NOMA while preserving low-complexity decoding, with the exact SIC-ML equivalence and seed-compromise resilience as notable technical strengths. The Rician extension and multi-user partitioning broaden applicability, though the magnitude of the reported 40 dB BER-ratio improvement would need to be weighed against the sparsity overhead.","major_comments":[{"comment":"Abstract (Marchenko-Pastur conditioning argument): the claim that this law controls the conditioning of any random unitary submatrix (including small sparse ones when the shared seed is compromised) is load-bearing for both the no-jammer-floor BER and the seed-compromise resilience statements. Marchenko-Pastur governs bulk eigenvalue distributions in the large-dimension limit; for finite row dimension equal to the sparsity level the extreme singular values are not guaranteed to remain bounded away from zero and infinity with high probability, so an adversary knowing the seed could in principle select a jammed pattern that produces an ill-conditioned effective channel and induces a BER floor. A concrete finite-dimensional bound or counter-example check is required.","section":"Abstract (precoder performance and seed compromise paragraph)"},{"comment":"Abstract (partitioning rule for >4 users): the OMA-friendly partitioning into pairs on disjoint bin subsets is presented as reaching floor BER at eight users by ~20 dB, but the rule appears post-hoc and its effect on the overall cross-user BER and the claimed jammer-independence must be derived explicitly; without this the multi-user scaling claim is not fully supported.","section":"Abstract (multi-user extension paragraph)"}],"minor_comments":[{"comment":"The abstract states that simulations track analytical predictions within 3 dB but provides no error bars, number of Monte Carlo trials, or data-exclusion criteria; these details should be added to the simulation section for reproducibility.","section":"Abstract (simulation paragraph)"},{"comment":"Notation for the active-set selection and the precise definition of the superincreasing sequence (Merkle-Hellman construction) should be introduced with an equation number in the main text rather than left implicit.","section":null}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the careful reading and constructive comments. We address the two major comments point by point below.","responses":[{"response":"We acknowledge that the Marchenko-Pastur law is an asymptotic result and that a finite-dimensional guarantee on the extreme singular values of random unitary submatrices would strengthen the argument. Our current justification relies on the fact that the three specific unitary precoders (Hadamard, DFT, Haar) produce submatrices whose empirical condition numbers remain bounded in all simulated regimes, including small sparsity levels, with no observed BER floor even under seed-compromise scenarios. In the revision we will add either a finite-dimensional bound (leveraging known results on the singular-value distribution of submatrices of these structured unitaries) or an expanded set of Monte-Carlo checks that explicitly test adversarial bin selection when the seed is known. This addresses the referee's request for a concrete check.","revision_made":"yes","referee_comment":"[Abstract (precoder performance and seed compromise paragraph)] Abstract (Marchenko-Pastur conditioning argument): the claim that this law controls the conditioning of any random unitary submatrix (including small sparse ones when the shared seed is compromised) is load-bearing for both the no-jammer-floor BER and the seed-compromise resilience statements. Marchenko-Pastur governs bulk eigenvalue distributions in the large-dimension limit; for finite row dimension equal to the sparsity level the extreme singular values are not guaranteed to remain bounded away from zero and infinity with high probability, so an adversary knowing the seed could in principle select a jammed pattern that produces an ill-conditioned effective channel and induces a BER floor. A concrete finite-dimensional bound or counter-example check is required."},{"response":"The partitioning assigns users to disjoint delay-Doppler bin subsets so that each pair operates exactly as the two-user case with its own superincreasing power allocation; because the supports are disjoint, cross-pair interference is zero and the jammer acts independently on each pair. Consequently the no-floor property and the SIC-ML equivalence carry over directly. We agree that an explicit derivation of the aggregate cross-user BER under this rule is needed to support the scaling claim. In the revision we will insert a short derivation showing that the overall BER is the average of the per-pair BERs and remains independent of the jammer pattern.","revision_made":"yes","referee_comment":"[Abstract (multi-user extension paragraph)] Abstract (partitioning rule for >4 users): the OMA-friendly partitioning into pairs on disjoint bin subsets is presented as reaching floor BER at eight users by ~20 dB, but the rule appears post-hoc and its effect on the overall cross-user BER and the claimed jammer-independence must be derived explicitly; without this the multi-user scaling claim is not fully supported."}],"tokens_in":1681,"tokens_out":595,"duration_ms":16755,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The paper puts forward a delay-Doppler NOMA method that activates only a small random subset of bins each frame, spreads via a unitary precoder drawn from a shared seed, and applies superincreasing power levels so that simple SIC equals maximum-likelihood detection exactly.\n\nWhat is new is the specific mix: per-frame re-randomization of the active set, the claim that any unitary submatrix stays well-conditioned under the Marchenko-Pastur law even if the seed leaks, and the Merkle-Hellman-style powers that remove the usual SIC error floor for more than two users. The extension to Rician fading while keeping jammer independence is also a clean addition.\n\nSimulations are said to stay within 3 dB of the closed-form BER and to deliver large gains over standard OTFS-NOMA under oracle jamming. That level of reported agreement is useful.\n\nThe soft spot is the conditioning claim itself. Marchenko-Pastur is an asymptotic result for large matrices; the active sets here are small and sparse, so the extreme singular values of the effective channel are not automatically bounded away from zero or infinity. An adversary who knows the seed could in principle choose a jammed pattern that produces a poorly conditioned submatrix and creates a BER floor. The paper would need explicit finite-size bounds or exhaustive checks on small matrices to close this gap.\n\nThe partitioning rule for more than four users also reads as a practical patch rather than a fundamental part of the design.\n\nThis is for researchers working on secure high-mobility waveforms and physical-layer NOMA. It has enough concrete claims and technical machinery to deserve a serious referee, even if the finite-dimension analysis needs tightening.","headline":"The scheme combines sparse randomized bins, unitary precoding, and superincreasing powers to claim jammer-independent BER and exact SIC-ML match, but the Marchenko-Pastur conditioning for small submatrices looks like the main point to verify.","tokens_in":2557,"tokens_out":433,"would_cite":false,"duration_ms":14630,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"Sparse randomized delay-Doppler bins plus unitary precoding remove the partial-band jamming floor from NOMA while keeping BER inside the unjammed envelope.","keywords":["jamming resilience","NOMA","delay-Doppler","unitary precoding","superincreasing power allocation","sparse signaling","successive interference cancellation","OTFS"],"falsifier":"A Monte Carlo trial that produces a visible bit-error-rate floor under partial-band jamming whose height matches the conventional OTFS-NOMA floor would disprove the no-floor claim.","tokens_in":2761,"feed_emoji":"🛡️","tokens_out":934,"duration_ms":22222,"temperature":0.7,"pith_summary":"The paper establishes that placing each user's data on a small random subset of delay-Doppler bins, spreading the symbols through a unitary precoder, and redrawing the active subset every frame from a shared seed lets the receiver identify and discard jammed bins, recover the sparse vector by least squares, and decode by successive interference cancellation. This construction yields a closed-form bit-error-rate expression that contains no jammer-induced floor, in contrast to the well-known partial-band floor of conventional OTFS-NOMA. The same Marchenko-Pastur conditioning argument that makes any random unitary submatrix well-behaved also shows that an adversary who learns the seed still cannot force a performance loss. For more than two users a superincreasing power allocation drawn from the Merkle-Hellman knapsack ensures that the low-complexity successive cancellation decoder coincides exactly with maximum-likelihood detection, eliminating the usual error-propagation ceiling.","feed_headline":"Sparse random bins plus unitary precoding erase jamming floor in NOMA","feed_subtitle":"Closed-form BER stays inside the unjammed curve even when the shared seed is known to the jammer; superincreasing powers make simple SIC ide","key_machinery":"The combination of sparse randomized active sets drawn from a shared seed and unitary precoding, whose submatrices remain well-conditioned by the Marchenko-Pastur law, enabling jammed-bin rejection and exact successive-interference-cancellation decoding.","core_discovery":"By restricting each user to a sparse random support in the delay-Doppler domain, applying a unitary precoder, and redrawing the support per frame from a shared pseudo-random seed, the transmitter forces any jammer to hit only a fraction of the bins; the receiver discards the jammed entries, solves the resulting well-conditioned least-squares problem, and applies successive interference cancellation. The resulting bit-error-rate expression has no floor under partial-band jamming. The Marchenko-Pastur law guarantees that every random unitary submatrix remains invertible with high probability, so even if the seed is compromised the bit-error-rate curve stays inside the unjammed envelope. For mo","pith_inferences":["The same sparse-support and unitary-precoding idea could be applied to other multicarrier waveforms that admit a delay-Doppler representation.","The exact equivalence between successive cancellation and maximum-likelihood under superincreasing powers may reduce receiver complexity in power-limited jammed links.","Hardware validation would need to confirm that least-squares recovery remains accurate when channel estimation errors and synchronization offsets are present.","The 40 dB reported improvement against pattern-aware jammers suggests the scheme could be attractive for low-probability-of-intercept links."],"forward_implications":["The closed-form bit-error-rate expression contains no jammer-induced floor.","Compromising the shared seed leaves the bit-error-rate curve inside the unjammed envelope because every random unitary submatrix stays well-conditioned.","Superincreasing power allocation makes low-complexity successive interference cancellation identical to maximum-likelihood detection for more than two users.","Partitioning users into pairs on disjoint bin subsets lets the scheme reach its floor bit-error-rate at eight users near 20 dB SNR.","The jammer-independence property holds for any Rician K-factor."],"fun_headline_variants":["Sparse random bins and unitary precoding eliminate jamming floor in NOMA","Randomized active sets enable jamming-resilient sparse delay-Doppler NOMA","Marchenko-Pastur law preserves NOMA BER under known seed jamming","Superincreasing powers yield exact low-complexity SIC for jammed NOMA"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The receiver can correctly identify every jammed bin from the received energy pattern and the shared seed without misclassification.","fun_headline_variants_meta":{"raw":{"variants":["Sparse random bins and unitary precoding eliminate jamming floor in NOMA","Randomized active sets enable jamming-resilient sparse delay-Doppler NOMA","Marchenko-Pastur law preserves NOMA BER under known seed jamming","Superincreasing powers yield exact low-complexity SIC for jammed NOMA"]},"model":"grok-4.3","cost_usd":0.008582,"raw_usage":{"total_tokens":3961,"prompt_tokens":841,"num_sources_used":0,"completion_tokens":69,"cost_in_usd_ticks":85824500,"prompt_tokens_details":{"text_tokens":841,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":3051,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":841,"tokens_out":69,"duration_ms":19895,"temperature":1.0,"reasoning_tokens":3051,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-27T15:29:30.201662+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"A Monte Carlo trial that produces a visible bit-error-rate floor under partial-band jamming whose height matches the conventional OTFS-NOMA floor would disprove the no-floor claim.","supporting_citations":[],"review_version":1}