{"id":"01170616-0ef7-4022-af16-99fdcb7e2f9e","arxiv_id":"2504.16577","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Optimizing pinching-antenna locations and user powers via fractional programming raises simulated uplink sum-rate in multiuser MISO systems beyond what a fixed antenna array achieves.","lead":"This paper develops an algorithm that simultaneously selects pinching-antenna positions and user transmit powers to maximize uplink data rates in a multiuser MISO wireless system. The simulations show gains over fixed antenna arrays, which matters for the emerging 6G flexible-antenna research direction.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Eqs. (11)–(13) do not maximize the stated F2: the position-update gradient omits waveguide/path-phase derivatives and is not a real-valued ascent direction, so the simulated PASS gains in Figs. 2–4 are not yet verified.","rationale":"The reader's weakest_assumption is the physical pinching-antenna channel model, which is a legitimate external-model risk. I focus instead on an internal correctness issue that is more decisive because it fails even under the paper's own model. The FP transforms in Lemmas 1–2 are standard and likely correct, and the BCD structure is recognizable; the soft spot is the nonconvex Psi_p block, where no global method is claimed but a local stationary point is required. The printed gradient appears to have missing conjugates and, more importantly, missing -j phase terms, making Eq. (12) not the derivative of any real objective and the line-search update in Algorithm 1 not an ascent step for f2. If the finite-difference check confirms the mismatch, the numerical evidence for the central claim collapses until corrected. If it does not, my objection is withdrawn and the reader's physical-model caveat becomes the dominant residual risk. Because the paper is otherwise a plausible application of FP/BCD, I would keep the reader's CONDITIONAL verdict rather than move to REJECT: the issue is substantive but addressable, provided the authors supply the corrected gradient or the simulation code and the finite-difference check passes with the corrected update.","tokens_in":7997,"tokens_out":13831,"duration_ms":143949,"concrete_test":"Take a random feasible instance with N=M=2, fix alpha, beta, p from the closed-form updates, and numerically differentiate F2 in Eq. (8) with respect to x_p^n using central finite differences of step 1e-6. Compare the resulting real derivative to the real part of Eq. (12)–(13) at the same point. If the two differ beyond numerical tolerance, Algorithm 1 does not ascend the stated objective, and the PASS curves in Figs. 2–4 must be regenerated with a corrected gradient (or with supplied code) before the headline gain claim can be assessed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that joint PA-position/power optimization, via Algorithm 1 inside BCD, produces the PASS sum-rate gains in Figs. 2–4. That claim requires the position subproblem P3 to actually maximize F2 with respect to x_p^n. As printed, it does not. Eq. (11) is meant to rewrite f1 in (10) using g_mn = sqrt(eta) C_mn, but it writes beta_mn C_mn where beta_m^H g_m involves beta_mn^* C_mn, and the quadratic interference term appears as a product of C_mn C_mn' without the required |sum_n beta_mn^* C_in|^2. Even if the missing conjugates are considered extraction artifacts, the gradient in (13) is still not the correct derivative: C_mn = exp(-j phi_n) exp(-j k r_mn)/r_mn has an x_n-derivative containing -j(2 pi/lambda_g + k (x_n - x_m)/r_mn) C_mn, while D_m in (13a) contains only the real 1/r term (x_m - x_n)/r^2, and F_m in (13c) contains a spurious exponential. Eq. (12) is complex-valued while x_p^n is real, so the update x <- x + l*gradient in Algorithm 1 is not a well-defined ascent step for f2. Consequently the monotonicity argument in (16) does not establish convergence of Algorithm 2 to a stationary point of the stated problem. With no code or corrected derivation provided, the numerical comparisons in Figs. 2–4 are not reproducible. This is an internal correctness objection that applies before any debate about the pinching-antenna physical model.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies uplink sum-rate maximization in a pinching-antenna system (PASS) with multiple dielectric waveguides, assuming MMSE-SIC and MMSE-nSIC receivers. The authors formulate two non-convex problems in user transmit powers and pinching-antenna positions, propose a fractional-programming (FP) reformulation with auxiliary variables, and solve the resulting problem by block coordinate descent (BCD), including a gradient-descent update for PA positions with backtracking line search. Numerical results claim that the proposed joint optimization significantly improves the sum-rate over a conventional fixed ULA, and that MMSE-nSIC nearly matches MMSE-SIC when positions are optimized.","tokens_in":8376,"tokens_out":7819,"duration_ms":70147,"significance":"The topic is timely, and the paper is among the first to treat uplink multi-waveguide PASS with joint position and power optimization. The FP-BCD architecture is standard, and the observation that the low-complexity nSIC receiver can approach SIC performance after position optimization is practically interesting. However, the central algorithmic derivation contains load-bearing errors in the position-update gradient and in the reformulated objective, so the numerical claims in Figs. 2-4 are not yet supported. The physical model also assumes ideal isotropic pinching elements without mutual coupling or waveguide loss, perfect instantaneous CSI, and exact continuous positioning; these idealizations are common in first studies of this kind but should be stated as limitations.","major_comments":[{"comment":"Eq. (11) is not an equivalent reformulation of f1 in Eq. (10). The first term should read Re(β_{m,n}^* C_{m,n}) (with the complex conjugate on β), and the second term should be p_i |Σ_n β_{m,n}^* C_{i,n}|^2 for each interfering user i, not the product Σ_n Σ_{n'} β_{m,n} β_{m,n'} C_{m,n} C_{m,n'} using C_{m,n} for all users. As printed, the expression changes the objective and therefore the subsequent minimization is not maximizing the sum-rate of the original problem.","section":"Section III.A, Eq. (11)"},{"comment":"The gradient formulas in Eqs. (12)-(13c) do not differentiate f2 with respect to the real variable x_p^n. The derivative of C_{m,n} = e^{-jφ_n} e^{-j k r_{mn}}/r_{mn} contains the phase terms -j(2π/λ_g) and -j k (x_n-x_m)/r_{mn} multiplying C_{m,n}, plus the magnitude derivative; none of these appear correctly in (13a)-(13c). Moreover, the right-hand side of (12) is complex-valued while x_p^n is real, so the update x_p^n^{(t+1)} = x_p^n^{(t)} + l ∇_{x_p^n} F2 in Algorithm 1 is not a real-valued ascent step. The convergence argument in Eq. (16) therefore does not apply.","section":"Section III.A, Eqs. (12)-(13c)"},{"comment":"The closed-form power update in Eq. (15) does not respect the non-negativity constraint on p_m. Minimizing f3(P) in (14) over p_m ≥ 0 yields p_m^* = 0 whenever Re(a_m) ≤ 0, whereas the expression min{Pmax, Re(a_m)^2/B_m^2} is positive in that case. The paper does not state the domain assumption or handle this case, so the BCD update for P is not valid for all channel realizations.","section":"Section III.A, Eqs. (14)-(15)"},{"comment":"The monotonicity proof in Eq. (16) is not complete. Even if the equality marked with (*) follows from Lemma 2 after updating α and β, the inequality F2(α^{(j+1)},β^{(j+1)},P^{(j+1)},Ψ_p^{(j+1)}) ≥ F2(α^{(j)},β^{(j)},P^{(j)},Ψ_p^{(j)}) requires that Algorithm 1 and the power update each increase F2. Since the gradient used in Algorithm 1 is not the correct derivative (see major comment above) and the power update has the issue described in the previous comment, the monotonic increase and the claimed convergence to a stationary point are not established.","section":"Section III.A, Eq. (16)"}],"minor_comments":[{"comment":"The identity matrix is written as I_L in both equations, but the dimension should be I_N (N antennas); the symbol L is never defined.","section":"Eqs. (3) and (4)"},{"comment":"The complex conjugates on β_{m,n} and on one of the C factors are missing in the printed expressions; these are not mere notation issues because they change the real part and the quadratic term that are being optimized.","section":"Eq. (11) and Eq. (13b)"},{"comment":"'gradient decent method' should be 'gradient descent method'.","section":"Section III.A, before Eq. (10)"},{"comment":"The complexity expression has an unmatched parenthesis and should clarify whether log3 denotes the logarithm base 3 (which is consistent with the l ← l/3 backtracking) or a logarithm of a variable named '3'.","section":"Section III.A, complexity paragraph"},{"comment":"The publisher location is printed as 'Cambridge U.K,'; it should read 'Cambridge, U.K.'.","section":"Reference [16]"}],"recommendation":"major_revision","confidential_remarks":"The reported numerical results cannot be accepted as evidence until the position-update gradient is corrected and the algorithm is re-run. The FP transformations themselves are standard and the framework is salvageable, so a major revision is appropriate rather than rejection. The authors should also be asked to clarify the physical modeling assumptions (e.g., no mutual coupling, ideal isotropic elements) and to state the domain issues in the power update."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nQuick read on arXiv:2504.16577. The paper is the first to formulate uplink sum-rate maximization for a multi-waveguide pinching-antenna system (PASS) with both MMSE-SIC and MMSE-nSIC receivers, jointly optimizing pinching-antenna positions and user powers. That problem statement is new and worth having in the literature. The high-level architecture—fractional programming, block coordinate descent, backtracking line search—is standard, but applying it to this model is a sensible thing to do. The main numerical observation, that optimized PA placement lets the lower-complexity nSIC receiver nearly match SIC, is interesting if it holds up.\n\nThe problem is that, as printed, the mathematics behind the position update does not check out. Equation (11) rewrites the objective without the required conjugates: the term β_m^H g_m should involve β_{m,n}^* C_{m,n}, and the interference term should be |Σ_n β_{m,n}^* C_{m,n}|², not a product of unconjugated C’s. More seriously, the gradient in (12)–(13) is not the derivative of F2 with respect to x_p^n. C_{m,n} = exp(-j φ_n) exp(-j k r)/r has an x-derivative containing -j(2π/λ_g) plus -j k (x_n - x_m)/r contributions; (13a) keeps only the real 1/r² term, and (13c) has a spurious exponential. Since the gradient is complex-valued, updating the real position as x ← x + l ∇ is not even a valid ascent step. That breaks the monotonicity argument in (16) and means the simulation gains in Figs. 2–4 are not yet supported by the paper’s own math. The nSIC case is dismissed with “derivations omitted” rather than derived, and no code or data is provided, so nothing is independently checkable.\n\nThe baseline is also weak: a fixed ULA at the center of the area mostly shows that moving antennas helps, which is not surprising. And I would not even raise the physical channel-model assumptions here—the internal correctness issue is the first barrier.\n\nIf the gradient is corrected, a proper real-valued ascent direction is derived, and the simulations are re-run, the paper could be a solid contribution to the PASS literature. As written, it is not acceptable. I would send it to peer review because the problem formulation is timely and the errors are likely repairable, but a careful referee should demand the corrections before any positive recommendation. If you work on pinching antennas, cite it for the problem setup, not for the algorithm.","headline":"First uplink multi-waveguide PASS sum-rate formulation, but the position-update gradient is wrong and the simulation gains are not yet supported by the paper's own math.","tokens_in":8899,"tokens_out":3628,"would_cite":false,"duration_ms":33757,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper claims that jointly optimizing pinching-antenna positions and user powers via an FP-BCD algorithm raises uplink sum rate over fixed arrays, and that optimized positions let MMSE without SIC nearly match MMSE with SIC.","keywords":["pinching-antenna systems","uplink sum-rate maximization","fractional programming","block coordinate descent","MMSE-SIC","MMSE-nSIC","multiuser MISO","antenna position optimization"],"falsifier":"A full-wave electromagnetic simulation or a 28 GHz prototype measurement of a dielectric-waveguide pinching array, comparing the measured channel vectors against Eq. (1) for the same positions, would settle the central claim: if the measured channels do not follow the assumed phase-and-distance law, or if the optimized-position sum-rate gain over a fixed ULA does not reproduce, the claimed gains would not hold.","tokens_in":7817,"feed_emoji":"📡","tokens_out":7951,"duration_ms":74997,"temperature":0.7,"pith_summary":"The paper tries to establish that a pinching-antenna system, a set of small dielectric particles that can be slid along dielectric waveguides, can increase the uplink sum rate of a multiuser MISO link when antenna positions and user powers are optimized together. The authors formulate two sum-rate maximization problems, one for MMSE reception with successive interference cancellation and one without, and solve both with a fractional-programming block-coordinate-descent algorithm. Numerical simulations at 28 GHz show that the optimized pinching system outperforms a conventional fixed uniform linear array under both receivers. The results also suggest that, once positions are optimized, the simpler parallel MMSE receiver without SIC achieves nearly the same sum rate as MMSE with SIC. The significance, if the model is right, is that antenna mobility can raise spectral efficiency without adding RF chains.","feed_headline":"Pinching antennas beat fixed arrays on uplink sum rate","feed_subtitle":"Optimizing antenna positions and user powers lifts sum rate; simple MMSE decoding nearly matches SIC.","key_machinery":"The load-bearing mechanism is the position-dependent phase-and-distance coupling in the effective channel $g_m = \\varphi \\circ h_m$. Sliding the $n$-th pinching element changes both the physical distance $\\|u_m-\\psi_n^p\\|$ in the spherical-wave term and the accumulated waveguide phase $\\varphi_n$, so one scalar position coordinate $x_n^p$ controls two channel effects at once. On the algorithmic side, the fractional-programming transformation in Lemmas 1 and 2 replaces the sum of logarithm-of-quadratic ratios with an equivalent objective linearized by auxiliary variables $\\alpha_m$ and $\\beta_m$; each block update is then either closed-form for $\\alpha$, $\\beta$, and the powers $p_m$, or a one-dimensional gradient ascent for $x_n^p$ with backtracking line search. This block-coordinate descent is what makes the otherwise nonconvex, phase-sensitive joint optimization computationally feasible.","core_discovery":"On the paper's own terms, the central discovery is that moving pinching antennas along waveguides is a new spatial degree of freedom that pays off in the uplink. With the effective channel $g_m = \\varphi \\circ h_m$, the spherical-wave free-space response from user $m$ to each pinching element modulated by the waveguide phase $\\varphi_n = e^{-j2\\pi(x_n^p-x_0)/\\lambda_g}$, jointly choosing the positions $x_n^p$ and powers $p_m$ raises the sum rate under both MMSE-SIC and MMSE-nSIC decoding. The proposed algorithm converts the nonconvex sum-log objective into an equivalent fractional-programming form with auxiliary variables, then alternates closed-form updates for the auxiliary variables and powers with gradient ascent using backtracking line search for the positions; the resulting sequence is monotone and converges to a stationary point. In simulations the pinching-antenna system beats the fixed ULA baseline at all power levels and user counts, and the SIC-versus-nSIC gap nearly vanishes when antenna mobility is available, especially as the number of waveguides grows.","pith_inferences":["Editorial inference: the position-power coupling optimized here for the uplink could be transferred to downlink beamforming or to activating several pinching elements on one waveguide, scenarios the paper does not study.","Editorial inference: if the idealized spherical-wave channel is replaced by a full-wave model with mutual coupling, the gains may shrink; a simulation study with a realistic antenna model would bound the effect.","Editorial inference: the algorithm assumes perfect instantaneous channel knowledge, so a natural robustness check is to rerun the joint optimization with estimated channels and measure how fast the sum-rate gain degrades with CSI error.","Editorial inference: because the gain comes from aligning each pinching element with the users, the method should show a larger advantage over a fixed ULA in clustered or asymmetric user distributions than in the uniform random placements simulated."],"forward_implications":["With a pinching-antenna array, the same number of RF chains can be repositioned along waveguides to follow the user geometry, raising uplink sum rate over a fixed uniform linear array.","Optimized pinching positions make the MMSE receiver without SIC nearly match the SIC receiver, so a parallel decoding architecture can capture most of the gain at lower complexity.","The SIC-versus-nSIC gap shrinks as the number of waveguides grows, so the low-complexity receiver becomes more attractive in larger arrays.","The proposed FP-BCD algorithm converges monotonically to a stationary point with polynomial per-iteration complexity, making the joint position-power optimization computationally feasible for the simulated problem sizes."],"supporting_citations":[{"why":"Supplies the pinching-antenna channel model $h_m$ in Eq. (1) that the whole optimization manipulates.","marker":"[8]"},{"why":"Provides the MMSE-SIC and MMSE-nSIC sum-rate expressions that are the optimization objectives.","marker":"[16]"},{"why":"Supplies the fractional-programming framework and the equivalence lemmas used in the algorithm.","marker":"[17]"},{"why":"Gives the effective refractive index $n_{\\text{eff}}$ used in the waveguide phase shift $\\varphi_n$.","marker":"[18]"},{"why":"Provides the gradient-descent-with-backtracking method and the stationary-point convergence argument used for position updates.","marker":"[19]"}],"fun_headline_variants":["Pinching antennas and smart placement lift MISO uplink rates","Optimizing pinching antenna positions boosts sum rate","Pinching antennas close SIC gap in MISO uplink","Antenna mobility via pinching outperforms fixed arrays","With pinching antennas, simple MMSE matches SIC in uplink"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole result rests on assuming the pinching-antenna channel is exactly the spherical-wave path-loss formula multiplied by the waveguide phase shift, with no mutual coupling, no radiation pattern, no waveguide loss, perfect channel knowledge, and continuous exact positioning; if any of these fail, the optimized rate gains are not guaranteed.","fun_headline_variants_meta":{"raw":{"variants":["Pinching antennas and smart placement lift MISO uplink rates","Optimizing pinching antenna positions boosts sum rate","Pinching antennas close SIC gap in MISO uplink","Antenna mobility via pinching outperforms fixed arrays","With pinching antennas, simple MMSE matches SIC in uplink"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000356,"raw_usage":{"total_tokens":1888,"prompt_tokens":860,"completion_tokens":1028,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":476,"completion_tokens_details":{"reasoning_tokens":945}},"tokens_in":476,"tokens_out":1028,"duration_ms":9729,"temperature":1.0,"reasoning_tokens":945,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T11:00:54.765537+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A full-wave electromagnetic simulation or a 28 GHz prototype measurement of a dielectric-waveguide pinching array, comparing the measured channel vectors against Eq. (1) for the same positions, would settle the central claim: if the measured channels do not follow the assumed phase-and-distance law, or if the optimized-position sum-rate gain over a fixed ULA does not reproduce, the claimed gains would not hold.","supporting_citations":[{"cited_title":"Tse and P","cited_arxiv_id":null,"evidence_quote":"Provides the MMSE-SIC and MMSE-nSIC sum-rate expressions that are the optimization objectives."},{"cited_title":"Fractional programming for communic ation systems-Part II: Uplink scheduling via matching,","cited_arxiv_id":null,"evidence_quote":"Supplies the fractional-programming framework and the equivalence lemmas used in the algorithm."}],"review_version":1}