{"id":"da2c0f9b-643b-40bb-9c81-190a3a189367","arxiv_id":"2607.24609","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"ONOMA uses overhearing plus ACK timing to order users and run symbol-reconstruction NOMA, cutting completion time up to ~34% versus TDMA and other baselines in asymmetric two-user low-feedback links.","lead":"A new wireless scheme called ONOMA cuts multi-user delivery latency when feedback is rare and channels are unequal, without needing channel knowledge at the sender. It matters for satellite links and massive IoT, where one weak device otherwise stalls everyone.","discovery_kind":"new_method","skeptic_critique":{"model":"moonshotai/kimi-k3","headline":"The abstract's \"up to 50% in larger asymmetric networks\" half of the central claim has no supporting analysis or simulation anywhere in the paper — all evidence is U=2, and even there ONOMA loses to SR-MC/SR-NOMA at R=0.5.","rationale":"The reader identified the single-shot ACK→Ω inversion (Eqs. 48–49) and perfect reconstruction as the weakest assumption, and mentioned the missing U>2 evidence and crippled λ=0.5 NOMA baseline only in passing within the rationale. I partially agree: the estimator is genuinely fragile, and I sharpen it with a concrete failure mode (negative Ω̂ when t<F(N), which occurs with order-10% probability exactly in the strong-overhearer regime). But I judge the most load-bearing issue for the central claim as stated to be the unsupported \"50% in larger asymmetric networks\" pillar: it is half of the headline quantitative claim, and the paper contains zero analysis or simulation for U>2. The correct fix is not rejection — the two-user contribution (protocol, E[T] bounds, Appendices A–D, regime-dependent 27–34% gains vs TDMA/SR-MC) is real and internally consistent, and the GF(2) rank analysis is standard and checkable. The appropriate posture is exactly the reader's CONDITIONAL: qualify the abstract to what is shown (two-user, R≥1, asymmetric regimes), state how degenerate Ω̂ realizations are handled, and either supply U>2 evidence or drop the 50% figure. Hence UNCHANGED with high confidence.","tokens_in":22436,"tokens_out":5320,"duration_ms":202409,"concrete_test":"Extend the §V Monte Carlo (300 runs, same PHY parameters, K̄=100) to U=4 and U=8 with one weak user (Ω≈−95 dBm) and the rest at −80 dBm, R∈{1,2}, implementing §IV.E scheduling with Eq. 48–49 estimation (f=1.5). Report mean completion time vs TDMA and SR-MC, and log Ω̂(2) from Eq. 48, counting runs where t_Ph-1<F(N) yields Ω̂<0. If the reduction vs the best baseline is materially below 50%, varies strongly with which user is weak, or degenerate estimates are frequent and silently clamped, the abstract must be qualified to the two-user setting with a stated estimator fallback.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The strongest claim has two quantitative pillars: \"up to 34% in two-user\" and \"up to 50% in larger asymmetric networks.\" The first is backed by Fig. 6c/6f Monte Carlo (vs TDMA/SR-MC, specific regimes). The second appears only in the Abstract and Conclusion. Section IV.E gives a qualitative U>2 scheduling sketch (TDMA-to-weak-user with overhearing, then pairwise NOMA with the strongest remaining overhearer), but there is no completion-time expression, no E[T] bound, and no simulation for any U>2 anywhere in §IV–§V. Every equation from (40)–(51) and every figure is two-user. So the 50% number is asserted, not derived or measured. This matters because the U>2 dynamics are not a trivial extension: Phase 2 pairs only two users, so per-pairing the weak user still needs its full rank before TDMA resumes; the benefit depends on the distribution of the max channel among remaining overhearers, and the ACK-timing estimator (Eq. 48) must now work repeatedly. A secondary, sharper crack in the same machinery: Eq. 48 inverts E[T]=exp(τσ²/(ΩP))·F(N) on a single realized t, but when user 2 is strong (the regime where ONOMA helps most), t<F(N)≈N+1.6 occurs with substantial probability (e.g., P(T=50)≈0.17 for N=50, P_succ=0.99), making ln(t/F(N))<0 and Ω̂(2) negative — Algorithm I's handling of this is unstated. Finally, \"outperforms ... classical NOMA\" is vacuous at R≥1, where the paper sets λ=0.5, violating its own feasibility condition (6) (infinite E[T], admitted in §V), and at R=0.5 ONOMA is beaten by SR-MC and SR-NOMA (Fig. 6g–i) — so the claim's breadth exceeds the evidence in both directions.","agreement_with_reader":"partial"},"referee_report":{"model":"moonshotai/kimi-k3","summary":"The paper proposes ONOMA, a cross-layer transmission scheme for low-feedback multi-user networks combining RLNC with symbol-aware NOMA. In Phase 1, the transmitter sends RLNC-coded packets to one user while others overhear; the relative ACK timing implicitly reveals channel-strength ordering without CSI. In Phase 2, the strong user, having already decoded the weak user's packets, regenerates the weak user's symbols and cancels them, achieving interference-free decoding and \"decoupling\" user latencies. An adaptive power-allocation policy (Algorithm I) is derived from channel estimates obtained by inverting the expected-completion-time formula on a single observed ACK time. The authors derive expected completion times for ONOMA and five baselines (TDMA, multicast, FDMA, inter-session coding, classical NOMA) in a two-user Rayleigh-fading setting and report Monte Carlo results (300 runs) showing up to 34% completion-time reduction in the high-rate asymmetric regime; the abstract additionally claims up to 50% in larger asymmetric networks.","tokens_in":22973,"tokens_out":2975,"duration_ms":103711,"significance":"If the results hold, the paper offers a practical, feedback-light mechanism for mitigating weakest-user latency domination in asymmetric multi-user networks, a relevant problem for NTN and massive-IoT settings. Notable strengths: the two-user expected-time derivations for both baselines and ONOMA are explicit and reproducible (Eqs. 15–47, Appendices A–D); the ACK-timing channel inference idea is simple and implementable; and the Monte Carlo study (300 runs, sweep over R and channel asymmetry) is consistent with the analytics. However, the significance is currently overstated at the abstract level: the '50% in larger networks' figure is unsupported, the classical-NOMA comparison is degenerate at R≥1, and the power-allocation policy rests on an unvalidated single-sample estimator. With those gaps addressed, the two-user core would be a solid contribution.","major_comments":[{"comment":"The abstract and §VI claim completion-time reduction of 'up to 50% in larger asymmetric networks,' but the paper contains no analysis and no simulation for any U>2. Every equation from (40)–(51) and every figure in §V is two-user, and §IV.E is a qualitative scheduling sketch only. The U>2 dynamics are not a trivial extension: Phase 2 pairs only two users, the weak user must still accumulate full rank before TDMA resumes, the benefit depends on the channel distribution of the strongest remaining overhearer, and the ACK-timing estimator (48) must be applied repeatedly. Either provide U>2 analysis/simulations supporting the 50% figure, or remove/scope the claim to what is shown.","section":"Abstract; §IV.E; §VI"},{"comment":"Eq. (48) inverts the expected-value formula E[T]=exp(τσ²/(Ω̂P))·F(N) on a single realized ACK time t. When user 2 is strong — precisely the regime where ONOMA helps most — t<F(N)≈N+1.6 occurs with substantial probability (e.g., P(T=50)≈0.17 for N=50, P_succ=0.99), making ln(t/F(N))<0 and the estimate Ω̂(2) negative (or undefined). Algorithm I's behavior in this event is unstated. More generally, a one-shot inversion of a mean-value identity is a high-variance estimator, and no sensitivity analysis of the λ policy to estimation error is given. Since λ directly controls both users' Phase-2 success probabilities (38)–(39), the robustness of this estimator is load-bearing for the claimed gains. At minimum, state the estimator's failure handling and show performance under estimator noise or a simple multi-sample variant.","section":"§IV.D, Eq. (48)–(49), Algorithm I"},{"comment":"The JR-ONOMA completion time (44) is written as max{E[T_Ph-2^(1)], E[T_Ph-2^(2)]}, and the text states an upper bound follows by substituting (42)–(43). The true Phase-2 completion time is E[max(T^(1),T^(2))], which is lower-bounded (not upper-bounded) by max of the expectations, and substituting an upper bound on one branch into a max-of-means does not yield an upper bound on the quantity of interest. As written, (44) is not a valid characterization of JR-ONOMA latency. Please either derive E[max] (the two ranks evolve in the same slots under independent channels, so a union-bound or exact expression should be feasible) or clearly label (44) as a heuristic approximation and validate it against the Monte Carlo results.","section":"§IV.C, Eq. (44)"},{"comment":"The claim 'outperforms ... classical NOMA' is vacuous at R≥1: the paper sets λ=0.5 for SR-NOMA, which violates its own feasibility condition (6) (τ(R)≥1), yielding infinite expected transmissions — admitted in §V — so NOMA appears in Figs. 6a–f only as a degenerate baseline. Conversely, at R=0.5 (where (6) holds), SR-ONOMA is beaten by SR-MC and SR-NOMA (Fig. 6g–i). The headline comparison to five baselines therefore holds only in the high-rate asymmetric regime; the abstract and §V conclusions should be scoped accordingly, and ideally a feasible NOMA baseline (rate R<1 with λ satisfying (6), or SIC-feasible power split) should be included for a fair comparison.","section":"§III.E, §V, Figs. 6–7"},{"comment":"Interference-free decoding at user 2 (Eqs. 36–37) assumes perfect regeneration of user 1's symbols: user 2 must have decoded user 1's entire Phase-1 block, know M, R_c, and the shared coding vector, and cancel with zero residual. No sensitivity to imperfect cancellation (residual interference fraction, header decoding failure at rate R′, or channel-estimation error in the analog subtraction) is analyzed or simulated. Since this is the mechanism that 'decouples' the latencies, even a brief robustness study (e.g., residual interference at −20 dB) would substantially strengthen the central claim.","section":"§IV.B, Eqs. (36)–(39)"}],"minor_comments":[{"comment":"Fig. 6 caption states 'N=5 packets per data block' for R=2 bpcu, but §V text gives N=K̄/R=50. Please correct.","section":"Fig. 6 caption"},{"comment":"Ω is a channel-power (variance) parameter; quoting it in dBm (e.g., Ω(2)=−90 dBm) is dimensionally unusual for a Rayleigh variance. Clarify the normalization to noise power σ², since only the ratio τσ²/(ΩP) enters the formulas.","section":"§V"},{"comment":"Terminology alternates between 'SR-MC'/'SC-MC' and 'SR-NOMA'/'SC-NOMA' (e.g., after (20), (26), (35)). Please unify.","section":"§III.B–E"},{"comment":"Typos/formatting: 'the the strong user' (§I), 'The, using the tail-sum formula' (§II.B), 'UA V' (§I), 'up to34%' and '50%in' in the abstract, 'R= 2,bpcu' (§V), 'whose proof' punctuation around (19)–(20).","section":"Throughout"},{"comment":"Footnote 1 on security ('Security is preserved under standard layered encryption assumptions') appears unmotivated — no security threat model is discussed anywhere in the paper. Either motivate or remove.","section":"§IV.B, footnote 1"},{"comment":"The design constants f=1.5 and t_max are free parameters; a short sensitivity sweep of f (and statement of how t_max is chosen in the simulations) would help reproducibility.","section":"§IV.D, §V"},{"comment":"In §III.C the FDMA description says rates are set 'as , i.e., R(1)=R(2)=R' — missing text before the comma.","section":"§III.C"}],"recommendation":"major_revision","confidential_remarks":"The two-user technical core is competent and the protocol idea is a reasonable fit for the journal. The main editorial concern is the gap between the abstract-level claims (U>2 gains of 50%, blanket outperformance of five baselines) and the evidence actually in the paper (two-user, high-rate, asymmetric regimes only, with some baselines infeasible by construction in the compared regime). I would recommend revision rather than rejection: the required changes are partly presentational (claim scoping) and partly additional experiments, but no fundamental flaw in the two-user analysis was found."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The useful core is a concrete cross-layer recipe for low-feedback multi-user delivery when the transmitter has no CSI: RLNC to the weak user while others overhear, ACK timing to order channels and set a heuristic λ, then symbol-aware NOMA where the strong user regenerates and subtracts the weak user’s symbols instead of doing same-slot decoding SIC. That assembly is new enough, and the two-user analysis is done properly.\n\nThey derive explicit E[T] expressions for TDMA, multicast, FDMA, inter-session, classical NOMA, and both JR/SR-ONOMA (Eqs. 15–47), with the usual RLNC rank sums and Rayleigh outage, plus appendices that bound residual ranks with the familiar 3/4 factor. Simulations (300 runs, R in {0.5,1,2}, asymmetric Ω) show clear wins versus TDMA and the better baselines in the high-rate asymmetric cases—up to the stated ~34% in Fig. 6c/f. When both channels are strong or R is low, SR-MC is competitive or better; they mostly show that. Classical NOMA with λ=0.5 is correctly noted as infeasible for R≥1, so that comparison is weak but not hidden.\n\nSoft spots are real but bounded. The abstract and conclusion claim “50% in larger asymmetric networks.” Section IV.E is only a qualitative scheduling sketch; every equation and every figure is U=2. That number is asserted, not measured. The ACK-timing Ω̂ inversion (Eqs. 48–49) plus hand-tuned f=1.5 and t_max is a heuristic; when the strong user finishes early the log can go negative and the paper does not say what Algorithm I does. Perfect symbol regeneration (known M, Rc, identical coding vector) is load-bearing and untested under header errors or imperfect sync. None of this sinks the two-user contribution.\n\nThis is for people working low-feedback NTN/IoT scheduling and RLNC-NOMA hybrids. The math and citation pattern look solid; circularity is low. I would send it to referees after the multi-user claim is qualified or backed, the power-policy edge cases are handled, and error bars/code appear. Worth engaging for the two-user protocol; do not cite the 50% number as is.","headline":"Clean two-user protocol with real gains in the asymmetric high-rate regime; the abstract’s 50% multi-user claim is unsupported and should be cut or qualified.","tokens_in":24074,"tokens_out":594,"would_cite":false,"duration_ms":20972,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.5","headline":"Overhearing and ACK timing let a transmitter finish multi-user deliveries faster without channel knowledge, by decoupling strong users from the weakest link.","keywords":["low-feedback networks","multi-user latency","RLNC","symbol-aware NOMA","overhearing","ACK timing","completion time","power allocation"],"falsifier":"In a controlled two-user Rayleigh setup with known asymmetric gains, measure whether SR-ONOMA’s measured completion time stays near the paper’s bounds and beats TDMA/SR-MC by the claimed margins when power is set only from the phase-1 ACK time (and t_max, f), versus when the true gains are given to the allocator; a large gap or loss of the 34% gain would refute the claim.","tokens_in":23664,"feed_emoji":"📡","tokens_out":912,"duration_ms":18501,"temperature":0.7,"pith_summary":"In low-feedback wireless settings such as non-terrestrial links and massive IoT, the transmitter often cannot get frequent ACKs or channel state, so a single weak user can force everyone to wait. This paper proposes ONOMA, a two-phase scheme that pairs random linear network coding with symbol-aware non-orthogonal multiple access. In phase one the transmitter sends coded packets aimed at one user while others overhear; the order and timing of block ACKs reveal who is strong without any CSI. In phase two it superimposes packets so the strong user, already holding the weak user’s data, regenerates and subtracts those symbols and decodes interference-free. Adaptive power is set from the observed ACK time. Analysis and simulations claim this cuts completion time by up to about a third in two-user cases and half in larger asymmetric networks versus TDMA, multicast, FDMA, inter-session coding, and classical NOMA.","feed_headline":"ACK timing cuts multi-user wait by up to 50% without CSI","feed_subtitle":"Overhearing plus symbol reconstruction lets strong users finish without waiting on the weakest link","key_machinery":"ONOMA: a phased protocol that first builds side information via overhearing of RLNC packets, then runs symbol-aware NOMA in which the strong user regenerates the weak user’s modulated symbols from known coding vectors and cancels them, plus a power-allocation rule driven by inverting the RLNC completion-time formula on observed ACK times.","core_discovery":"The paper claims that overhearing-driven NOMA (ONOMA) minimizes multi-user completion latency without instantaneous or statistical CSI by using ACK timing to order channels and reconstruction-based symbol cancellation so strong users finish without waiting on the weakest user.","pith_inferences":["The same ACK-timing estimator could be stress-tested under mobility or non-block fading, where one shot time-to-ACK may mix path loss with short-term fades.","If header/coding-vector delivery at the lower rate R' fails often, the strong user’s regeneration step breaks; header reliability is an unemphasized single point of failure.","Extending the idea to uplink grant-free or multi-cell settings would need a different overhearing geometry than the downlink PtP model used here."],"forward_implications":["In asymmetric low-feedback downlinks, overall block delivery time need not track the weakest user’s PtP completion time.","Sparse block ACKs can replace explicit CSI for ordering users and setting NOMA power.","Reconstruction-based cancellation after overhearing can replace classical same-slot SIC when side information is built first.","Scheduling for U>2 can alternate TDMA overhearing rounds with ONOMA pairings as early ACKs identify strong users.","Reported gains are largest at higher rates and severe channel asymmetry, and shrink when channels are both strong or rates are very low."],"fun_headline_variants":["ACK timing orders channels so strong users finish early without CSI","ONOMA decouples multi-user latency via overhearing and symbol cancel","Overhearing-driven NOMA cuts completion time up to 50% in asymmetric nets","ACK-timed power and reconstruction free strong users from weak-link wait","No CSI: ONOMA uses ACK timing to slash multi-user wait up to 50%"],"cache_read_input_tokens":16512,"weakest_assumption_plain":"A single observed ACK time, inverted through the expected RLNC completion formula plus a chosen maximum time for the weak user, is treated as a good enough estimate of channel strength to set power so latency truly decouples.","fun_headline_variants_meta":{"raw":{"variants":["ACK timing orders channels so strong users finish early without CSI","ONOMA decouples multi-user latency via overhearing and symbol cancel","Overhearing-driven NOMA cuts completion time up to 50% in asymmetric nets","ACK-timed power and reconstruction free strong users from weak-link wait","No CSI: ONOMA uses ACK timing to slash multi-user wait up to 50%"]},"model":"grok-4.5","effort":"low","cost_usd":0.003776,"raw_usage":{"total_tokens":1163,"prompt_tokens":746,"num_sources_used":0,"completion_tokens":85,"cost_in_usd_ticks":37764000,"prompt_tokens_details":{"text_tokens":746,"audio_tokens":0,"image_tokens":0,"cached_tokens":128},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":332,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":746,"tokens_out":85,"duration_ms":7353,"temperature":1.0,"reasoning_tokens":332,"cache_read_input_tokens":128,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-31T10:38:50.122143+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"In a controlled two-user Rayleigh setup with known asymmetric gains, measure whether SR-ONOMA’s measured completion time stays near the paper’s bounds and beats TDMA/SR-MC by the claimed margins when power is set only from the phase-1 ACK time (and t_max, f), versus when the true gains are given to the allocator; a large gap or loss of the 34% gain would refute the claim.","supporting_citations":[],"review_version":1}