{"id":"19a7e32a-9626-43f7-b3db-3cc6a0ecd81e","arxiv_id":"2412.13892","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"For uplink pinching-antenna systems, the authors propose a successive-convex-approximation placement algorithm plus a Lambert-W closed form for resource allocation, maximizing the minimum user rate.","lead":"Engineers at Aristotle University of Thessaloniki, Manchester, and Khalifa University built an algorithm that positions sliding pinching antennas along a waveguide and splits uplink airtime so the slowest user's data rate is maximized. It is one of the first studies to optimize uplink connections for this next-generation antenna technology, which NTT DOCOMO introduced in 2022.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The convex-positioning claim breaks at constraint C'8: f^2 <= z is nonconvex because the phase residual f in (6) is sign-changing, so P1.1.4 is not a convex program and the SCA convergence guarantee is unsupported.","rationale":"We agree with the reader's weakest_assumption: constraint C'8 is the load-bearing point. The rest of the paper's analytic contribution--the SOCP form, the SCA convergence assurance, and the 'convex reformulation' headline--all depend on C'8 being convex. It is not, and the provided counterexample is not contrived: it uses the actual f from (6) with parameters within the simulation range (d=3 m, n_e=1.4). The resource-allocation closed form and the numerical comparisons may survive as a heuristic, and the final evaluation against the true rate expression (5) via the epsilon sweep is a sensible safeguard, so the flaws are correctable rather than destructive of the engineering results. Additional issues noted by the reader--the misprinted min t in P1.1.4, the surrogate objective, and unspecified Lagrange multipliers--reinforce the conditional verdict but are secondary. Since our read does not move the verdict, we leave it at CONDITIONAL (UNCHANGED).","tokens_in":8583,"tokens_out":15562,"duration_ms":149177,"concrete_test":"Run the following convexity check for C'8. Use the parameters x_m=15, y_m=0, d=3, n_e=1.4, theta_m=0, and let h(x)=sqrt((x-15)^2+9)-1.4x. Compute h(14)=-16.438 and h(16)=-19.238; set z1=h(14)^2 and z2=h(16)^2. Both (x=14,z1) and (x=16,z2) satisfy h(x)^2<=z. The convex combination with weight 1/2 is (x=15, z=(z1+z2)/2 about 320.14). Evaluate h(15)^2=324. Since 324>320.14, the midpoint violates C'8, so the feasible set is not convex. This single numerical check settles whether the reformulation is convex as claimed; if the authors instead impose a nonnegativity constraint on f or replace C'8 with a genuinely convex surrogate, the claim should be restated accordingly.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central methodological promise is that the PA-placement subproblem is 'reformulated into a convex one' (Abstract; Section III-A). The load-bearing step is constraint C'8 in P1.1.3, written as ||[f(xP_n, theta_m), sqrt(z_mn)]|| <= sqrt(2 z_mn), which is algebraically equivalent to f(xP_n, theta_m)^2 <= z_mn. The text justifies convexity by saying the norm on the left is convex and the square root on the right is concave, but this does not make the constraint set convex: f in (6) is a convex, nonlinear, sign-changing function of xP_n and theta_m, and {f^2 <= z} is not generally the epigraph of a convex function. Concretely, for x_m=15, y_m=0, d=3, n_e=1.4, theta_m=0, define h(x)=sqrt((x-15)^2+9)-1.4x (the phase residual up to a positive constant). The points (x=14, z=h(14)^2) and (x=16, z=h(16)^2) both satisfy C'8, but their midpoint (x=15, z=(h(14)^2+h(16)^2)/2) has h(15)^2=324 while z is about 320.14, violating the constraint. Hence C'8 is not a convex constraint, P1.1.3/P1.1.4 is not a convex program, and 'Algorithm 1 always converges, since the nature of SCA ensures it' does not follow from the stated reformulation. The final epsilon-sweep evaluation against the true rate (5) means the method can still function as a heuristic, but the claimed convex reformulation and convergence guarantee must be either proved under additional assumptions or withdrawn.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the uplink of a multi-user pinching-antenna (PA) system under orthogonal multiple access. The authors formulate a max-min achievable-rate problem over PA positions and resource allocation, and propose to solve it by decoupling the two. For the PA-position subproblem, they introduce a phase-alignment surrogate and claim a convex reformulation, solving it with successive convex approximation (SCA) over a grid of phase-error tolerances. For resource allocation, they claim a closed-form expression based on a Lagrangian/Karush-Kuhn-Tucker analysis. Numerical results show gains over a conventional fixed-antenna system, with reported gains of roughly 5 dB for M=N=2 and 3 dB for M=N=3.","tokens_in":8856,"tokens_out":10938,"duration_ms":101502,"significance":"If the technical claims were correct, the paper would be a useful contribution: it is the first to address uplink optimization for PA systems, it uses a physically motivated waveguide channel model imported from the foundational literature, and the decoupling into position optimization and resource allocation is a sensible design principle. The epsilon-sweep with final evaluation against the true rate is a pragmatic safeguard. However, the central methodological promise—that the antenna-positioning problem is exactly reformulated into a convex program with guaranteed SCA convergence—is not supported by the derivation as written, and the closed-form resource allocation solution suffers from an algebraic sign problem. These issues must be resolved before the contribution can be considered reliable.","major_comments":[{"comment":"The claim that C'8 is convex is false. The inequality is written as ||[f(xP_n, theta_m), sqrt(z_mn)]|| <= sqrt(2 z_mn), which is algebraically equivalent to f(xP_n, theta_m)^2 <= z_mn. The set {(xP_n, theta_m, z_mn) : f^2 <= z_mn} is not generally convex because f in (6) is a sign-changing convex function of xP_n and theta_m. A concrete counterexample is obtained with x_m=15, y_m=0, d=3, n_e=1.4, theta_m=0, where the one-dimensional function h(x)=sqrt((x-15)^2+9)-1.4x satisfies h(14)^2 <= z and h(16)^2 <= z for suitable z, but at the midpoint x=15 the required z would be h(15)^2=324 while the interpolated z is about 320.1, violating the constraint. Thus P1.1.3/P1.1.4 is not a convex program, and the statement in Algorithm 1 that convergence is guaranteed 'since the nature of SCA ensures it' is unsupported. The authors should either provide a valid convexification (for instance by linearizing f itself within an SCA loop, or by introducing a transformation that guarantees convexity) or recast the method as a heuristic and remove the convergence and convexity claims.","section":"Section III-A, P1.1.3/P1.1.4, constraint C'8"},{"comment":"The closed-form resource allocation in (10) is algebraically wrong for the stated definition of the Lambert W function. The expression is q*_m = eta E_m \\bar{W} / (N sigma^2 (1+\\bar{W}) |sum|^2) with \\bar{W} = W(-e^{-1 - lambda_1 ln 2 / lambda_2}). For lambda_1, lambda_2 >= 0, the argument of W lies in [-1/e, 0), so the principal branch satisfies \\bar{W} in [-1,0) and 1+\\bar{W} >= 0. Hence q*_m as written is non-positive for every feasible multiplier, contradicting constraint C4 (q_m >= delta > 0). The correct expression should involve -\\bar{W} or an equivalent sign correction; the Lagrangian derivation leading to (10) should be reworked and the resulting formula verified numerically.","section":"Section III-B, Eq. (10)"},{"comment":"The surrogate used for PA placement is not connected to the true objective by any approximation guarantee. The objective in P1.1.1 replaces the coherent sum magnitude in (5) with the sum of inverse distances, while phase alignment is enforced only through the weighted residual constraint C6: sum_n w_mn f(xP_n, theta_m)^2 <= epsilon. The paper provides no bound relating the magnitude of the phase residual to the difference between the surrogate and the true coherent sum. Consequently, a solution that is optimal for the convexified surrogate may not be near-optimal for the true rate, even though the final epsilon-sweep in Algorithm 1 evaluates the true rate and selects the best trial. This is a limitation of the approach as stated and should be acknowledged explicitly; if the authors wish to claim optimality or a performance guarantee, an error analysis is required.","section":"Section III-A, P1.1.1-C6"}],"minor_comments":[{"comment":"The objective of P1.1.4 is printed as 'min t', but the preceding reformulation P1.1.3 maximizes t, and with constraint C'7 (sum_n v_mn >= t) a minimization would be unbounded below. This should be 'max t'.","section":"Section III-A, P1.1.4"},{"comment":"The Lagrangian in (9) contains the symbol t, but the resource allocation subproblem (P1.2.1) uses the auxiliary variable R2; the notation should be made consistent.","section":"Section III-B, Eq. (9)"},{"comment":"In the paragraph following (P1), the text says that 'C2 limits the resource blocks allocated to the M devices', but C2 is the antenna-position bound |xP_n - xP_{n'}| >= Delta, and the resource constraint is C3. This mislabeling should be corrected.","section":"Section III, P1"},{"comment":"The definition of the weights w_mn in (7) is only explicit for the case M=N, with the remark that 'if M != N, (7) must be modified appropriately'. Since the numerical section considers M=N=2,3,4, the general case is never specified; a precise construction for arbitrary M and N would be needed for the framework to be complete.","section":"Section III, Eq. (7)"}],"recommendation":"major_revision","confidential_remarks":"The paper comes from an experienced group and the system model is reasonably motivated, but the two main theoretical contributions—a convex PA-position reformulation and a closed-form resource allocation—are both flawed as written. The nonconvexity of C'8 is a fundamental issue for the stated SCA convergence claim, and the sign error in (10) is a concrete algebraic mistake. I believe the authors are capable of fixing these problems, but the revision will require more than minor editing: the optimization framework needs to be reworked or the claims need to be substantially downgraded to a heuristic with numerical validation. I would not recommend acceptance in the current form."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The genuinely new thing here is the problem: first max-min uplink rate formulation for pinching-antenna systems, with OMA resource allocation and antenna position optimization. That's worth someone's time. The channel model is imported from Ding-Schober-Poor and Pozar; the numerical gains (3–5 dB) are plausible and internally consistent, and Remark 1 is honest about the fixed-position limitation.\n\nThe soft spot is the center, though. The abstract says the antenna positioning problem is 'reformulated into a convex one,' and Algorithm 1 claims SCA convergence 'since the nature of SCA ensures it.' Both rest on constraint C'8 in P1.1.3: ||[f, sqrt(z)]|| <= sqrt(2z), i.e., f^2 <= z. The paper justifies convexity by saying the left-hand norm is convex and the right-hand square root is concave. That's not sufficient. f in (6) is convex but sign-changing, so {f^2 <= z} is not generally convex. A concrete counterexample: with x_m=15, y_m=0, d=3, n_e=1.4, theta=0, the points (x=14, z=f(14)^2) and (x=16, z=f(16)^2) both satisfy C'8, but the midpoint (15, average of the z's) violates it. So P1.1.3/P1.1.4 is not a convex program, and the 'always converges' claim doesn't follow. The method may still work as a heuristic—the epsilon sweep and evaluation against the true rate (5) suggest it often does—but the main methodological claim is currently false.\n\nOther issues are smaller. P1.1.4 is printed as min t where it should be max t—likely a typo, but confusing. The closed-form allocation (10) depends on lambda_1 and lambda_2 and a Lambert W branch that are never specified, so it's not actually closed-form as written. And there's no code or data, with sparse simulation detail.\n\nNet: the paper is a reasonable first cut at an important new system model, and the flaws are correctable. It shouldn't be published as is. Send it to referees who can force the convexity claim to be either proved under real assumptions or downgraded to a heuristic, and who will fix the objective direction and the allocation details. If the authors do that, it could be a useful contribution to the flexible-antenna literature.","headline":"First uplink pinching-antenna max-min rate formulation with plausible gains, but the advertised convex reformulation breaks at C'8, leaving the main algorithmic guarantee unsupported.","tokens_in":9568,"tokens_out":3403,"would_cite":true,"duration_ms":32197,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["94A05","90C25"],"pacs":[],"model":"deepseek-v4-flash","headline":"A pinching-antenna uplink design that decouples antenna placement from resource allocation maximizes the worst user's rate, reporting more than 5 dB gain over a fixed-antenna baseline.","keywords":["pinching antennas","leaky-wave antennas","uplink","max-min fairness","antenna positioning","resource allocation","convex optimization","line-of-sight communications"],"falsifier":"Fix the paper's parameters ($f_c=28$ GHz, $n_e=1.4$, $d=3$ m) and a device at $(15,0,0)$; for a fixed $\\theta_m$ and $z_{mn}$, search for two antenna positions $x_a<x_b$ along the waveguide that both satisfy $f(x,\\theta_m)^2\\le z_{mn}$ while their midpoint violates it. Since the phase residual $f$ is a convex function that can change sign, such a pair should exist, and its existence would prove the feasible set of $C8$ is nonconvex and the convex-reformulation claim is wrong.","tokens_in":8225,"feed_emoji":"📡","tokens_out":10963,"duration_ms":98316,"temperature":0.7,"pith_summary":"This paper proposes an uplink design for pinching-antenna (PA) systems in which small dielectric particles activate radiation points along a waveguide. The goal is to maximize the minimum achievable data rate across devices, and the paper's central claim is that this joint problem can be solved by separating antenna placement from resource allocation: antenna positions are optimized with a convex reformulation, and the resource split is given in closed form. If the claim holds, dynamic PA placement becomes a practical way to fight path loss and line-of-sight blockage, since the same antenna positions serve all devices while the airtime allocation enforces fairness. The reported gains over a fixed-antenna baseline are more than 5 dB for two devices and two antennas and about 3 dB for three devices and three antennas.","feed_headline":"Placing pinching antennas lifts worst-user uplink rate by 5 dB","feed_subtitle":"With two antennas it gains more than 5 dB; with three around 3 dB, versus a fixed antenna array.","key_machinery":"The load-bearing object is the phase-alignment constraint $C8$/$C'8$: for every device $m$ and antenna $n$, the residual $$f(x_n^P,\\theta_m)=\\tfrac{2\\pi}{\\$\\lambda$}\\sqrt{(x_m-x_n^P)^2+$y_m^{2}$+$d^{2}$}-\\tfrac{2\\pi}{\\lambda_g}x_n^P-\\theta_m$$ is squared and bounded by a slack variable $z_{mn}$, with the weighted sum $\\sum_n w_{mn} z_{mn}\\le\\epsilon$ controlling phase error while the weights $w_{mn}$ favor small distances. The paper converts the nonconvex rate objective into an auxiliary bound on the sum of reciprocal distances, rewrites the phase bound as a second-order cone constraint, linearizes the reciprocal-distance constraint by a first-order Taylor expansion, and solves the resulting convex program (P1.1.4) with SCA, scanning $\\epsilon$. Resource allocation is then solved in closed form from the Lagrangian, giving $q_m^*$ through the Lambert W function.","core_discovery":"On its own terms, the paper establishes a max-min fairness formulation for the uplink of a multi-user PA system, where the achievable rate of device $m$ is $R^P_m = q_m \\log_2(1+(\\eta E_m/(N q_m \\sigma^2))|\\sum_{n=1}^N e^{j\\frac{2\\pi}{\\lambda}\\|\\psi_m-\\psi_n^P\\|-j\\frac{2\\pi}{\\lambda_g}\\|\\psi_0^P-\\psi_n^P\\|}/\\|\\psi_m-\\psi_n^P\\||^2)$. For fixed equal resource shares, the antenna-position subproblem (P1.1.1) forces the weighted squared phase residual between the free-space path and the waveguide path to stay below $\\epsilon$, and the paper claims this subproblem is recast as the convex problem (P1.1.4) via a second-order cone constraint and successive convex approximation. For fixed positions, the resource-allocation subproblem is convex and its optimal solution is the closed-form Lambert-W expression $q_m^*$ in (10). The paper further claims Algorithm 1 always converges and, in simulation, delivers more than 5 dB min-rate gain at $M=N=2$ and about 3 dB at $M=N=3$ over the conventional fixed-antenna system.","pith_inferences":["The same position-solver structure would carry over to a downlink PA design, since the phase-alignment constraint depends only on geometric distances and is direction-independent.","A weighted-fairness extension (e.g., weighted max-min or sum-of-logs) would leave the position solver unchanged and alter only the resource-allocation step, which can still be handled by the same Lambert-W water-filling logic.","The $\\epsilon$ scan can be read as a robustness knob: smaller $\\epsilon$ forces tighter phase alignment at the cost of possibly worse path loss, so treating $\\epsilon$ as a tunable hyperparameter may trade worst-case rate for sensitivity to device-position uncertainty."],"forward_implications":["For fixed PA positions, the optimal orthogonal multiple access (OMA) resource allocation has a closed form $q_m^*$ via the Lambert W function, so fairness can be enforced with negligible computation.","Because the antenna-positioning subproblem is solved with SCA at claimed $O(N^3)$ complexity, the same framework can reposition antennas in near-real time for small systems.","The numerical comparison implies that antenna placement contributes more than resource allocation: removing the resource-allocation step (scheme b) costs far less than replacing PAs with fixed antennas (scheme c).","The reported gains shrink as $M=N$ grows to 4, showing that one shared antenna configuration cannot phase-align many devices simultaneously.","Rates improve with transmit power and degrade slightly when the waveguide is raised from 3 m to 4 m, consistent with the path-loss-dominated LoS model."],"supporting_citations":[{"why":"supplies the pinching-antenna system model, waveguide propagation assumptions, and the simulation parameters used in the numerical evaluation.","marker":"[9]"},{"why":"provides the waveguide propagation relation that makes each antenna's signal a phase-shifted replica, giving the phase term that the positioning constraint aligns.","marker":"[10]"},{"why":"introduces the pinching-antenna concept of activating radiation points along a dielectric waveguide.","marker":"[8]"},{"why":"supplies the spherical-wave near-field channel model used for the fixed-antenna baseline comparison.","marker":"[2]"}],"fun_headline_variants":["Pinching antennas boost worst-user uplink rate by up to 5 dB","Max-min fairness in uplink with movable pinching antennas","Optimal pinching antenna placement yields 5 dB min-rate gain","Pinching antenna design lifts uplink fairness and throughput","Algorithm places pinching antennas for max-min uplink rate"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the rewritten phase-error constraint leaves a convex set of antenna positions, a property the paper asserts on the basis of norm-versus-square-root convexity that does not hold in general.","fun_headline_variants_meta":{"raw":{"variants":["Pinching antennas boost worst-user uplink rate by up to 5 dB","Max-min fairness in uplink with movable pinching antennas","Optimal pinching antenna placement yields 5 dB min-rate gain","Pinching antenna design lifts uplink fairness and throughput","Algorithm places pinching antennas for max-min uplink rate"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000457,"raw_usage":{"total_tokens":2296,"prompt_tokens":952,"completion_tokens":1344,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":568,"completion_tokens_details":{"reasoning_tokens":1258}},"tokens_in":568,"tokens_out":1344,"duration_ms":9605,"temperature":1.0,"reasoning_tokens":1258,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T12:43:25.211522+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Fix the paper's parameters ($f_c=28$ GHz, $n_e=1.4$, $d=3$ m) and a device at $(15,0,0)$; for a fixed $\\theta_m$ and $z_{mn}$, search for two antenna positions $x_a<x_b$ along the waveguide that both satisfy $f(x,\\theta_m)^2\\le z_{mn}$ while their midpoint violates it. Since the phase residual $f$ is a convex function that can change sign, such a pair should exist, and its existence would prove the feasible set of $C8$ is nonconvex and the convex-reformulation claim is wrong.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"provides the waveguide propagation relation that makes each antenna's signal a phase-shifted replica, giving the phase term that the positioning constraint aligns."}],"review_version":1}