{"id":"a355c751-ef0f-4007-b7c4-34a9b1f43433","arxiv_id":"2411.10524","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"A mixed-criticality superposition coding scheme for RIS-assisted THz links that decodes high-priority data whenever either the direct or the reflected path is available, reducing critical-data queueing delay versus time-sharing.","lead":"Terahertz links are fast but fragile, so the authors propose sending high-priority data over a reliable reconfigurable-surface path and low-priority data over the fast direct path, using superposition coding. Simulations with queuing and beam-misalignment models suggest this mixed-criticality scheme cuts delays for critical data while keeping overall throughput higher than time-sharing.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"HC SINR (18) adds powers of the two HC paths as scalars, but the same symbol is sent over both paths; without modeling phase/combining, the 'either path available' guarantee and the queue-stability gains are not established.","rationale":"Read in good faith: the paper builds an internally consistent model, and the optimization and queueing framework are plausible. The half-power outage threshold is explicitly acknowledged and is conservative in the direction of individual-path failure, so it does not by itself threaten the qualitative claim that MC-SC improves HC reliability and reduces critical-data delay. The unstated assumption that HC signal powers from the two paths add as scalars in (18) is more load-bearing, because it underlies both the rate constraints (31d) and the outage probability (30), and hence the feasibility region and queuing results. If the two paths combine coherently and the receiver implements the appropriate combining, (18) may be conservative and the claim is safe; if phases are random or antagonistic, the HC reliability guarantee can fail exactly in the state β=(1,1) where both paths are available. The paper does not specify a receiver architecture or phase model that would make (18) the correct instantaneous SINR, so this gap must be checked before the quantitative claims can be considered established. Because this is an addressable modeling gap rather than a demonstrated falsehood, the reader's CONDITIONAL verdict remains appropriate; I would not move the verdict, hence UNCHANGED. Agreement with the reader is partial: the soft spot is in the outage/decoding model, but the specific mechanism I find load-bearing is the power-combining model in (17)-(18), not the half-power threshold itself.","tokens_in":18525,"tokens_out":20417,"duration_ms":211994,"concrete_test":"Recompute the physical layer with complex channel coefficients h = β_d η_d sqrt(ρ_d) e^{jφ_d} and g = β_r η_r sqrt(ρ_r) e^{jφ_r}, with phases either i.i.d. uniform per slot (no instantaneous CSI) or perfectly aligned (best case), then re-solve (31) and re-run the queue simulations of Figs. 9-10. Compare the stable α boundary and Pout,h to the paper's values; if α_max moves by more than about 10% or Pout,h exceeds the claimed 0.05 under strict requirements, the central comparison is not robust to the unmodeled combining mechanism.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim (Section I-B) is that HC data is reliably delivered whenever at least one of the two paths is available, and the queuing gains in Figs. 9-10 follow from this. The load-bearing physical-layer step is the HC SINR in (18): Γh = (|h|^2 p_h^(d)+|g|^2 p_h^(r))/( |h|^2 p_l^(d)+|g|^2 p_l^(r)+σ_n^2 ). In the receive model (17), the same HC symbol s_h is sent on both beams, so the HC component before noise is (h√p_h^(d)+g√p_h^(r)) s_h; its power is |h√p_h^(d)+g√p_h^(r)|^2, which generally differs from the sum-of-powers expression by the cross term 2Re(hg^*)√(p_h^(d)p_h^(r)). Equations (1)-(2) define h and g as real positive scalars, and the BS has only statistical knowledge of β and ϵ; no phase model or receiver combining rule is specified. If the two path phases are not guaranteed equal (or if the receiver does not have two independent receive branches for MRC), the combined HC signal can be smaller than (18), including near-cancellation, so rate constraints (31d) and outage probability (30) can overstate HC reliability when both paths are simultaneously present. This is distinct from the acknowledged half-power threshold approximation: that threshold makes individual-path failure events conservative, but the missing phase/combining model can go in the opposite direction and directly affects the claimed guarantee.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies a downlink RIS-assisted THz system with an intermittent direct line-of-sight path and a more reliable but weaker RIS path. It proposes a mixed-criticality superposition coding (MC-SC) scheme in which high-criticality (HC) and low-criticality (LC) packets are superimposed, with HC data always carried over both paths and LC data over the direct path only. A power allocation problem is formulated to maximize the minimum queue-stability gap subject to rate constraints evaluated at a half-power misalignment threshold and outage constraints, and is solved by an iterative SCA/fractional-programming algorithm in the appendix. Numerical simulations compare the feasible rate region, outage probabilities, and queuing delays against a time-sharing baseline, and show that MC-SC supports a substantially larger fraction of HC data with lower HC queueing delay.","tokens_in":18794,"tokens_out":15315,"duration_ms":142775,"significance":"If the physical-layer model is sound, the paper offers a useful way to trade rate for reliability in THz links by exploiting RIS path diversity for critical data. The paper is clearly written, the optimization is carefully formulated, the rate constraints are conservative with respect to misalignment, and the simulation study is extensive, including blockage, misalignment, and beamwidth adaptation. The main contribution is the combination of mixed-criticality superposition coding with a queue-stability formulation in a RIS-aided THz channel, which is not present in prior work. The central claims are falsifiable and the numerical setup is reproducible.","major_comments":[{"comment":"The HC SINR in (18) does not follow from the received signal model in (17). The same symbol s_h is sent over both beams, so the HC component before noise is (h√p_h^d + g√p_h^r)s_h; its power is |h|^2 p_h^d + |g|^2 p_h^r + 2Re{h g^*}√(p_h^d p_h^r). The cross term is omitted in (18). If h and g are understood as the real positive scalars defined in (1)–(2), the omitted cross term is positive and (18) is a lower bound; if the usual complex baseband model with a relative phase is intended, the cross term can be negative and (18) can be optimistic. The paper does not specify a phase or combining model, and this issue affects the rate constraint (31d) in the state β=(1,1) and the outage approximation (30). Consequently, the claim in Section I-B that HC data is reliably delivered whenever either the direct or the RIS path is available is not established for the both-paths-present state. Please state the assumed phase/combining model (e.g., coherent combining with known phase, or worst-case phase) and replace (18) accordingly, or argue explicitly why (18) is a valid bound.","section":"Section II-B, Eqs. (17)–(18)"},{"comment":"The HC outage probability is introduced as an approximation, and the text notes that the combined signal could support decoding even when both individual paths fail the half-power threshold. The paper does not quantify the error of this approximation or demonstrate by simulation that it is accurate or a guaranteed bound. Since (1 − Pout,h) enters the stability constraint (31a), the feasible regions in Fig. 5 and the queueing results in Figs. 9–10 inherit this uncertainty. Please provide a numerical or analytical comparison of (30) with the empirical outage probability of the proposed scheme, and state whether (30) is an upper or lower bound.","section":"Section III, Eq. (30)"},{"comment":"The appendix reformulates (31) into a sequence of convex problems and alternates between solving (38) and updating µ via (36)–(37), but no convergence proof, monotonicity argument, or stopping criterion is given. All numerical results in Section IV are generated by this algorithm. Please provide a convergence analysis (e.g., convergence to a stationary point of (31) under the standard assumptions of the FP framework in [38]) or an empirical convergence study, and specify the termination condition used in the simulations.","section":"Appendix, Algorithm 1"}],"minor_comments":[{"comment":"In Eq. (36), the first terms in the numerator and denominator use η_r^2 where the direct-path term should be η_d^2, as in Eq. (34).","section":"Appendix, Eq. (36)"},{"comment":"The phrase 'has beed studied' should read 'has been studied'.","section":"Section I-A"},{"comment":"Replace 'until Convergence' with an explicit stopping criterion, such as a tolerance on the relative change of the objective.","section":"Appendix, Algorithm 1"},{"comment":"The definition of 'normalized queue peak' appears only in the text; please add it to the figure captions.","section":"Figures 9 and 10"},{"comment":"The sentence 'with more power allocated to the HC stream and by leveraging path diversity, critical data experiences fewer outages caused by beam misalignment' is a claim that can be verified from the model; please clarify whether it is an observation from the simulations or a property of the constraints.","section":"Section II-B"},{"comment":"The positive-part operator is defined as [x]^+ in the Notation paragraph, but Eqs. (24)–(25) use parentheses; please use one notation consistently.","section":"Notation, Eqs. (24)–(25)"}],"recommendation":"major_revision","confidential_remarks":"The main risk is the phase/combining issue in Eq. (18). If the authors can either justify the real-coefficient assumption with a phase-alignment argument or revise (18) to a worst-case expression, the paper could be suitable for publication. The paper is a substantial extension of the authors' ICC 2024 paper; the added value from the queuing model, optimization, and extensive simulations is sufficient for a journal publication. The self-citation rate is noticeable but not extreme."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThis is an extension of Karacora et al.'s ICC 2024 paper on mixed-criticality superposition coding for RIS-aided THz. The genuinely new material is the two-queue mixed-criticality model, the joint blockage-misalignment analysis, the beamwidth-adaptation study, and the SCA-based power allocation. The central idea is simple and sensible: send HC data over both the intermittent direct LoS and the weaker but more reliable RIS path, decode HC first, and decode LC only when the direct path is available. The optimization (31) and the simulations support the claimed gains: a stable region up to α≈0.63 versus ≈0.18 for time sharing, and about 10× lower HC queueing delay at small α. I found the system model internally consistent, and the paper is honest about the two approximations it makes: the half-power threshold for link availability and the treatment of outages as independent per path.\n\nThe soft spots are real but not fatal. First, the half-power threshold (ρ ≥ A/2) is heuristic; the paper does not quantify how much the outage probabilities (28)-(30) change if the true decoding condition differs. Second, the outage probability (30) is an acknowledged approximation, and its accuracy against the combined-signal case is not checked. Third, the time-sharing baseline is only sketched, so the comparison in Figs. 9-10 is not fully reproducible. Finally, no code or data is shipped, and the simulation curves lack confidence intervals. These are addressable in revision.\n\nOne thing that surprised me in the stress test was the concern about phase/combining in the HC SINR (18). That concern does not land as written. In this model, h and g are real positive scalars (βη√ρ), so the same HC symbol sent over both beams combines coherently; the sum-of-powers expression is a lower bound, not an optimistic one. The approximation is conservative, which is the direction the authors claim. If they later generalize to complex channel coefficients with random phases, they would need to specify a combining rule, but for the current scalar model the math is consistent.\n\nOverall: a solid system-level study, not a breakthrough. The queueing and reliability gains are plausible under the model, and the paper is worth a serious referee. I would send it to review, asking for a check of the outage approximation, a fuller baseline description, and ideally code or at least detailed simulation methodology. I would cite it if I worked on criticality-aware THz/RIS.","headline":"Solid extension of the authors' SC-based criticality work; queueing/misalignment studies are new, claims hold up under the model, but the outage approximation needs scrutiny.","tokens_in":19373,"tokens_out":4307,"would_cite":true,"duration_ms":40839,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["94A40","60K25","90B22"],"pacs":[],"model":"deepseek-v4-flash","headline":"A THz downlink can keep critical data flowing whenever either the direct or the RIS-reflected path is up.","keywords":["terahertz communication","reconfigurable intelligent surface","mixed-criticality","superposition coding","beam misalignment","queue stability","outage probability","successive interference cancellation"],"falsifier":"Recompute the outage and queue-stability results using the exact joint decoding condition: declare HC success whenever $\\Gamma_h(\\beta,\\epsilon)\\geq 2^{R_h/B}-1$ with continuous misalignment variables $\\rho_d,\\rho_r$, instead of the per-link half-power threshold and the independent-path product in (30). If the stability boundary and delay numbers barely move, the paper's conclusion stands; if the boundary moves substantially, the reported $\\alpha\\approx0.63$ is an artifact of the approximation.","tokens_in":18274,"feed_emoji":"📡","tokens_out":7011,"duration_ms":65623,"temperature":0.7,"pith_summary":"This paper tries to establish that the classic THz tradeoff between rate and reliability can be broken by treating data by criticality instead of treating all packets equally. It proposes a mix of superposition coding and a RIS-reflected path so that high-criticality packets survive whenever either link is usable, while low-criticality packets ride along opportunistically on the direct path. The reported payoff is concrete: queue-stable operation for up to about 63% critical data (versus 18% for time sharing), roughly tenfold lower HC queueing delay for small critical fractions, and a throughput-versus-reliability frontier that dominates time sharing. If correct, this gives latency- and safety-sensitive 6G services a way to obtain both reliability and high throughput in channels that are frequently blocked or misaligned.","feed_headline":"Superposition coding keeps critical THz data flowing via either link","feed_subtitle":"Mixed-criticality superposition keeps queues stable for up to 63% critical data, versus 18% for time sharing.","key_machinery":"The load-bearing object is mixed-criticality superposition coding (MC-SC): the base station transmits $x_d=\\sqrt{p_h^{(d)}}s_h+\\sqrt{p_l^{(d)}}s_l$ toward the user and $x_r=\\sqrt{p_h^{(r)}}s_h+\\sqrt{p_l^{(r)}}s_l$ toward the RIS, with the HC stream carrying more power. The receiver applies successive decoding: HC first, treating LC as noise; after cancellation, LC is decoded from the residual. The optimization in (31) allocates the four powers so that HC rate constraints hold for every blockage state with at least one available path and LC rate constraints hold only for the unblocked direct path. Rates are evaluated at the half-power misalignment threshold $\\rho=A/2$, which turns continuous pointing errors into per-link outage probabilities through the parameters $\\gamma_d$ and $\\gamma_r$. The non-convex problem is solved by successive convex approximation with a fractional-programming quadratic transform.","core_discovery":"The central discovery is that mixed-criticality superposition coding turns THz link intermittency into a power-allocation problem with a much larger feasible region than time sharing. In the proposed scheme, the high-criticality message is superimposed with the low-criticality message at different powers on both the direct and RIS beams, and the user decodes the HC message first and cancels it. Because HC can be decoded from either link, its outage probability is the product of two per-path failure probabilities, and its rate constraints are enforced for every blockage state except the one where both links are blocked. The resulting feasible region supports HC fractions up to about α=0.63 with stable queues (versus 0.18 for time sharing), and at the recommended tradeoff point α≈0.62 the HC throughput nearly doubles while total throughput drops only about 12%.","pith_inferences":["Beyond the paper: because the per-link half-power threshold treatment ignores that two partially misaligned beams add coherently at the receiver, the reported HC outage probability is likely an upper bound, so the stable-HC boundary could be above α≈0.63 under the paper's own channel model.","Beyond the paper: the MC-SC structure transfers to other paired links with a strong-but-fragile and weak-but-stable profile, such as mmWave with a reflective surface or a satellite link with a terrestrial relay, wherever data can be split by criticality.","Beyond the paper: the tradeoff parameter α is chosen offline by a one-dimensional search; an online estimator that tracks blockage and misalignment statistics from acknowledgments could adapt α per coherence block and approach the reported Pareto front without knowing those statistics in advance."],"forward_implications":["The HC stream is decodable whenever at least one of the two links is available, while LC is decodable only when the direct line-of-sight is up, so the optimal solution sets LC power on the RIS beam to zero.","Queue stability holds for HC fractions up to about α=0.63, versus about 0.18 for time sharing, and HC average delay is roughly ten times lower for small α.","Total throughput peaks at α=0.28, and the recommended tradeoff point α=0.62 nearly doubles HC throughput while losing only about 12% of total throughput.","As direct-path blockage rises, total throughput drops from about 5 to 3.5 bit/s/Hz while HC throughput stays near 2.5 bit/s/Hz, and beam misalignment degrades both streams but HC remains relatively protected.","Under strict HC reliability requirements, MC-SC outperforms time sharing by about 35% in throughput and nearly triples the throughput of treating all data as HC at high misalignment.","HC outage probability is reduced by path diversity: the HC stream is disrupted only when both the direct and RIS paths fail, whereas LC is disrupted whenever the direct path fails."],"supporting_citations":[{"why":"Supplies the stochastic pointing-error model (Rayleigh displacement and the PDF/CDF of the misalignment fading coefficient) that defines per-link misalignment outage.","marker":"[30]"},{"why":"Provides the THz path-loss and molecular-absorption gain expressions for the direct link and the RIS-reflected link.","marker":"[8]"},{"why":"Supplies the quadratic transform for fractional SINR expressions that makes the non-convex power allocation tractable and drives the iterative algorithm.","marker":"[38]"},{"why":"Gives the mean-rate stability criterion used to impose the queue-stability constraints in (31a) and (31b).","marker":"[36]"},{"why":"Gives Little's law, the relation used to convert average queue occupancy into the reported mean queueing delays.","marker":"[37]"},{"why":"Supplies the prior idea of superposition coding for mixed-criticality delivery that the proposed MC-SC scheme adapts to a single-user RIS-assisted THz downlink.","marker":"[25]"},{"why":"Provides the line-of-sight channel model for 100-450 GHz, including the frequency-dependent molecular absorption coefficient used in the gain coefficients.","marker":"[29]"},{"why":"Gives the effective-area radius relation and misalignment fading parameterization used for the user equipment aperture in equations (9)-(12).","marker":"[33]"},{"why":"Preliminary version of this work that introduced the data-significance approach to THz intermittency, which this paper extends with misalignment, queueing, and beamwidth adaptation.","marker":"[1]"}],"fun_headline_variants":["Superposition coding doubles critical THz throughput with 12% loss","Mixed-criticality superposition codes stabilize THz queues","Dual-link superposition coding boosts critical THz data rates","THz reliability via mixed-criticality coding on dual beams"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The results rest on treating a link as usable only when misalignment fading is at least half the perfectly aligned power (the half-power beamwidth heuristic), and on approximating HC outage as the product of two independent per-link failure events, even though the receiver could in principle combine two partially misaligned signals.","fun_headline_variants_meta":{"raw":{"variants":["Superposition coding doubles critical THz throughput with 12% loss","Mixed-criticality superposition codes stabilize THz queues","Dual-link superposition coding boosts critical THz data rates","THz reliability via mixed-criticality coding on dual beams"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000438,"raw_usage":{"total_tokens":2226,"prompt_tokens":948,"completion_tokens":1278,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":564,"completion_tokens_details":{"reasoning_tokens":1211}},"tokens_in":564,"tokens_out":1278,"duration_ms":10711,"temperature":1.0,"reasoning_tokens":1211,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T19:37:08.236093+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the outage and queue-stability results using the exact joint decoding condition: declare HC success whenever $\\Gamma_h(\\beta,\\epsilon)\\geq 2^{R_h/B}-1$ with continuous misalignment variables $\\rho_d,\\rho_r$, instead of the per-link half-power threshold and the independent-path product in (30). If the stability boundary and delay numbers barely move, the paper's conclusion stands; if the boundary moves substantially, the reported $\\alpha\\approx0.63$ is an artifact of the approximation.","supporting_citations":[{"cited_title":"Outage Capacity Optimization for Free-Space Optical Links With Pointing Errors,","cited_arxiv_id":null,"evidence_quote":"Supplies the stochastic pointing-error model (Rayleigh displacement and the PDF/CDF of the misalignment fading coefficient) that defines per-link misalignment outage."},{"cited_title":"On the Downlink Coverage Performance of RIS-Assisted THz Networks,","cited_arxiv_id":null,"evidence_quote":"Provides the THz path-loss and molecular-absorption gain expressions for the direct link and the RIS-reflected link."},{"cited_title":"Introduction to Queues,","cited_arxiv_id":null,"evidence_quote":"Gives the mean-rate stability criterion used to impose the queue-stability constraints in (31a) and (31b)."},{"cited_title":"A Proof for the Queuing Formula: L = λW ,","cited_arxiv_id":null,"evidence_quote":"Gives Little's law, the relation used to convert average queue occupancy into the reported mean queueing delays."},{"cited_title":"Rate-splitting enabled multi-connectivity in mixed-criticality systems,","cited_arxiv_id":null,"evidence_quote":"Supplies the prior idea of superposition coding for mixed-criticality delivery that the proposed MC-SC scheme adapts to a single-user RIS-assisted THz downlink."},{"cited_title":"A line-of-sight channel model for the 100–450 gigahertz frequency band,","cited_arxiv_id":null,"evidence_quote":"Provides the line-of-sight channel model for 100-450 GHz, including the frequency-dependent molecular absorption coefficient used in the gain coefficients."},{"cited_title":"On the Joint Effect of Rain and Beam Misalignment in Terahertz Wireless Systems,","cited_arxiv_id":null,"evidence_quote":"Gives the effective-area radius relation and misalignment fading parameterization used for the user equipment aperture in equations (9)-(12)."},{"cited_title":"Intermittency Versus Path Loss in RIS-aided THz Communication: A Data Significance Approach,","cited_arxiv_id":null,"evidence_quote":"Preliminary version of this work that introduced the data-significance approach to THz intermittency, which this paper extends with misalignment, queueing, and beamwidth adaptation."}],"review_version":1}