{"id":"a4251eb7-876d-408d-a27d-07f37caf012c","arxiv_id":"2411.16083","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":3.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"DAUR, an alternating optimization algorithm for maximizing processed bits per delay-plus-energy in blockchain-Metaverse wireless networks, relies on an invalid fractional-programming transform.","lead":"The paper proposes a data processing efficiency metric for blockchain-Metaverse wireless systems and an algorithm, DAUR, that alternately optimizes user association and resource allocation. The algorithm's core fractional-programming transformation contains an algebraic error, so the claimed optimality is not established.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The real weak point is Lemma 5.3, not Appendix C: P3's objective is unbounded and cannot be equivalent to P2; the Reader's Eq.-63 algebra actually checks out.","rationale":"The Reader's verdict (REJECT) is correct, but the specific weakest assumption identified by the Reader is not. Appendix C's fractional-programming equality is algebraically valid: substituting v = 1/(2χr) into the transformed transmission term reproduces χ/r exactly, and the derivative equalities in Equations (58) and (61) follow. The load-bearing flaw is earlier, in Lemma 5.3. As written, P3 removes the two constraints that define θ, leaving θ and α free in a bilinear objective. The resulting problem is unbounded and its KKT stationarity condition in θ contradicts the claimed optimality relation α = 1/cost. The Appendix B proof computes KKT conditions for P2, not for P3, and then asserts equivalence without a derivation. Since Theorem 5.1 depends on Lemma 5.3, the claimed reduction of the sum-of-ratios DPE problem to a solvable convex problem is unsupported. The numerical results therefore cannot be interpreted as solving P1. This is an internal inconsistency in the argument, not a disagreement with consensus. The scalar test in concrete_test isolates the structure of P3 and demonstrates the non-equivalence directly. The paper's other limitations, such as the heuristic treatment of the binary association and the weak baselines, are secondary once the main transformation is invalid. I recommend keeping the Reader's REJECT verdict, but for the Lemma 5.3 reason rather than the Appendix C algebra.","tokens_in":26127,"tokens_out":11175,"duration_ms":109570,"concrete_test":"Instantiate the exact P2→P3 structure on a scalar analogue: A(x)=x, B(x)=1+x, x∈[0,1]. P2 is max θ s.t. θ(1+x) ≤ x, whose optimum is max_x x/(1+x)=1/2. Lemma 5.3's transformation would give max_{θ,α,x} α(x − θ(1+x)) with no constraint linking α to 1/B. Taking x=1, θ=1/2, α→∞ yields +∞; taking x=0, θ=1, α→−∞ yields −∞, so the transformed problem is unbounded and plainly not equivalent. Repeating this one-line check on P3 (with A = c(1−φ)d, B = cost_u, and similarly for server terms) settles the non-equivalence. If needed, also verify Eq. 63 numerically with sample values to confirm the Reader's Appendix-C objection is not the issue.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that DAUR provably maximizes DPE by a chain of equivalent transformations. The Reader's flagged flaw in Lemma 5.5 / Appendix C does not land: with χ = xρpφd and v = 1/(2χr), χ²v + 1/(4r²v) = χ/(2r) + χ/(2r) = χ/r, so F = G and the derivative equalities (58) and (61) hold. The load-bearing gap is Lemma 5.3. P3 is stated as a maximization over x, φ, γ, φ, ρ, ζ, ψ, θ, α, T of Σ α_u[c_u(1−φ)d − θ_u cost_u] + Σ α_s[c_s xφd − θ_s cost_s], with constraints (15a)–(15b) dropped. Even with α fixed positive, the term −αθ cost is unbounded above as θ → −∞ because cost > 0 and nothing bounds θ; if θ is restricted to be nonnegative, α → ∞ gives unboundedness whenever the bracketed term is positive. More fundamentally, stationarity of P3 with respect to θ gives −α cost = 0, contradicting the claimed KKT value α = 1/cost > 0. The Appendix B proof derives KKT conditions of P2, not of P3, then asserts the transformation. Thus P3 is not equivalent to P2. Every later step (P4, P5, P6–P10) optimizes an ill-posed surrogate, so Theorem 5.1's reduction is unproven and the reported DPE gains do not establish optimality for P1.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper defines data processing efficiency (DPE) as processed data bits divided by delay plus energy consumption, formulates a sum-of-ratios maximization problem P1 for user association and resource allocation in a blockchain-enabled Metaverse wireless system, and proposes the DAUR algorithm. DAUR is claimed to transform P1 into a sequence of tractable problems: an equivalent summation problem P2, a 'splitting' problem P3 with auxiliary variables, a fractional-programming subproblem P5 for communication and computing resources, and an SDR-based QCQP subproblem P10 for user association and offloading. Numerical experiments compare DAUR with four heuristic baselines and report DPE gains.","tokens_in":26497,"tokens_out":6503,"duration_ms":63983,"significance":"The DPE metric is a reasonable system-level efficiency objective, and the system model usefully combines wireless offloading, server processing, and blockchain-related delays. If the claimed transformation chain were correct, DAUR would be a valuable contribution to resource allocation in similar MEC/blockchain settings. The paper also provides a clearly described simulation setup and reports reproducible-looking convergence behavior. However, the mathematical core of the paper is not established: the equivalence of P2 and P3 is invalid, and the proof of Lemma 5.3 derives KKT conditions for a different problem. Since the correctness of DAUR, the stationarity claim, and the interpretation of the simulation results all rest on this chain, the central contribution is currently unsupported. I note that the suspected algebra in Eq. (63) is actually correct; the load-bearing flaw is in Lemma 5.3, not in the fractional-programming transformation.","major_comments":[{"comment":"Problem P3 is not equivalent to P2 and is generally unbounded. In P3 the variables alpha and theta appear only in the objective and are not constrained by the cost expressions. For fixed feasible (x, phi, gamma, phi, rho, zeta, psi), the term alpha_u (c_n(1-phi_n)d_n - theta_u cost_u) can be made arbitrarily large by taking theta_u to -infinity, or, if theta is required nonnegative, by taking alpha_u to infinity with theta_u = 0 whenever c_n(1-phi_n)d_n > 0. The proof in Appendix B forms the Lagrangian of P2, derives stationarity and complementary slackness for P2 (Eqs. (38)-(45)), and then asserts the transformation to P3. But stationarity of the P3 objective with respect to theta_u gives -alpha_u cost_u = 0, which contradicts the claimed value alpha_u = 1/cost_u in Eq. (17). Since Lemma 5.3 is the basis for the alternating update of alpha and theta in Algorithm 1, the claimed reduction of P1 to a solvable problem is unproven.","section":"Lemma 5.3, Eq. (16)"},{"comment":"Lemma 5.7 restricts the system parameter omega_b to 1 without justification. The proof shows two different minimizers: gamma = omega_b/(1+omega_b) minimizes the energy term in Eq. (64), and gamma = 1/(1+omega_b) minimizes T_sp + T_sg in Eq. (66). These values coincide only when omega_b = 1. The paper then substitutes gamma = 1/2 into Problem P7 and uses that substitution throughout P8-P10. Since omega_b is a data-size changing ratio specified in the general problem and is only set to 1 in the default simulation, the SDR formulation solves a different problem whenever omega_b != 1. This is a load-bearing step for Theorem 5.6.","section":"Lemma 5.7, Eq. (66)"},{"comment":"Because Lemma 5.3 and Lemma 5.7 are load-bearing, the convergence claims in Algorithm 1 and the statement that the returned point is a stationary point of P3 do not follow from the provided proofs. The numerical DPE values in Figs. 3 and 4 therefore do not establish that DAUR maximizes DPE for P1; they only show that the proposed heuristic outperforms four hand-crafted baselines. Without a valid proof of equivalence, the reported gains cannot be attributed to the claimed optimality properties.","section":"Theorem 5.1 and Algorithm 1"}],"minor_comments":[{"comment":"The text describing Fig. 4(e) says the emphasis shifts from 'delay-centric (0.1, 0.9)' to 'energy-centric (0.9, 0.1)', but with weights (omega_t, omega_e), the pair (0.1, 0.9) gives low weight to delay and high weight to energy. The labels appear reversed and should be corrected.","section":"Fig. 4(e), Section 7"},{"comment":"The assumption that the server validation delay T_sv is negligible relative to processing and block generation delays is used implicitly in Lemma 5.7 but is not stated in the system model. This assumption should be made explicit, or the proof should handle T_sv.","section":"Section 3.1.5 and Lemma 5.7"},{"comment":"Several proofs say 'Refer to Appendix A/B/C in the full version paper [14]' even though the corresponding appendices appear in this manuscript. These references should be updated to the local appendices.","section":"References to full version"}],"recommendation":"reject","confidential_remarks":"The suspected algebra in Eq. (63) is actually correct; the fatal issue is Lemma 5.3, where P3 is unbounded and the proof derives KKT conditions for P2 rather than P3. This is not a local fix: the equivalence chain from P2 to P3 is the foundation of the alternating algorithm, so the paper's central claim is unproven. Lemma 5.7 adds a second unrestricted specialization (omega_b = 1). In my view the manuscript cannot be accepted without a substantially new correctness argument."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the paper defines DPE (processed bits divided by delay plus energy) and a DAUR algorithm to maximize it in a blockchain-Metaverse MEC system. The system model is detailed and P1 is a reasonable sum-of-ratios problem. Lemma 5.2's epigraph reformulation is fine. But the chain breaks at Lemma 5.3, not at Lemma 5.5. The Reader's Eq.(63) algebra actually checks out: with chi = x rho p phi d and v = 1/(2 chi r), chi^2 v + 1/(4 r^2 v) = chi/(2r) + chi/(2r) = chi/r, so the FP transform in Appendix C preserves F = G. That part is standard and correct.\n\nLemma 5.3 is the real problem. P3 drops constraints (15a)-(15b) and instead puts alpha times (c d - theta cost) in the objective, with theta and alpha both free variables. If theta is unrestricted, the objective is unbounded above because negative theta with positive alpha and positive cost gives arbitrarily large values. If theta is restricted to be nonnegative, then alpha can go to infinity whenever the bracketed term is positive. Either way P3 has no finite optimum. The claimed KKT relations (17)-(20) are not KKT conditions of P3: stationarity with respect to theta gives -alpha cost = 0, contradicting alpha = 1/cost. Appendix B actually derives KKT conditions of P2 and then asserts P3 follows; it does not. Since P3 is not equivalent to P2, the later problems P4-P10 do not inherit any claim to solve P1, and Theorem 5.1 is unproven.\n\nWhat is legitimately new: the DPE metric is a simple but reasonable composite, and the paper formulates a concrete joint user-association and resource-allocation problem. The simulations are clearly described and show consistent gains over four heuristics. But those baselines are all home-made (random or greedy connection plus average allocation), with no comparison to an existing FP-based method or any bound on suboptimality. The reported convergence in one outer iteration across all tested sizes also makes the alternating-optimization claims hard to trust. No code or data is provided; the appendices point to a OneDrive full version.\n\nBottom line: this paper is not ready for serious peer review as a correctness contribution. The DPE metric and the system model might be useful to someone working in this niche, but the algorithm's guarantees do not survive contact with Lemma 5.3. I would desk-reject rather than send it to referees in its current form.","headline":"The DPE idea is fine and the system model is detailed, but Lemma 5.3 breaks the equivalence chain: P3 is unbounded, so DAUR's optimality claims do not hold.","tokens_in":26975,"tokens_out":6976,"would_cite":false,"duration_ms":67090,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":false},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proposes the DAUR algorithm, which maximizes data processing efficiency (processed bits per delay-plus-energy) in blockchain-enabled Metaverse wireless systems by jointly optimizing user association, work offloading, and…","keywords":["data processing efficiency","blockchain","Metaverse","user association","resource allocation","fractional programming","semidefinite relaxation","work offloading"],"falsifier":"Evaluate both sides of the claimed equality $F=G$ at a randomly chosen feasible point, substituting $\\upsilon_{n,m}^{(s)} = 1/(2 x_{n,m}\\rho_n p_n \\varphi_n d_n r_{n,m})$; if the numerical values differ, the convexified Problem P5 is not equivalent to Problem P4, and the algorithm's solution is not a DPE maximizer.","tokens_in":25911,"feed_emoji":"📡","tokens_out":7864,"duration_ms":66368,"temperature":0.7,"pith_summary":"This paper introduces data processing efficiency (DPE), defined as processed data bits divided by the sum of delay and energy consumption, as a metric for blockchain-enabled Metaverse wireless systems where users offload NFT tasks to servers. Its central claim is that the resulting sum-of-ratios optimization problem, which couples binary user-association variables with continuous offloading, bandwidth, power, and computing allocations, can be transformed into a sequence of solvable convex subproblems. The proposed DAUR algorithm alternates between optimizing user association, offloading ratios, and task-specific computing distribution on one side, and bandwidth, transmit power, and resource usage ratios on the other, using fractional programming, QCQP reformulation, semidefinite relaxation, and Hungarian rounding. The authors report that DAUR converges quickly and achieves DPE up to 86.48 Mbits/(s·J), outperforming four heuristic baselines across the simulated settings. If correct, this offers a tractable way to balance latency and energy in Metaverse task processing.","feed_headline":"86.48 reached: joint tuning lifts Metaverse data efficiency","feed_subtitle":"DAUR jointly tunes links, offloading, bandwidth, power, and computing to maximize bits per delay-plus-energy.","key_machinery":"The load-bearing mechanism is the sequence of equivalent reformulations of the sum-of-ratios objective. The named objects are the DPE ratio (processed bits over delay-plus-energy), the auxiliary variables $\\theta_n^{(u)}$, $\\theta_{n,m}^{(s)}$, $\\alpha_n^{(u)}$, $\\alpha_{n,m}^{(s)}$, and $\\upsilon_{n,m}^{(s)}$, and the alternating optimization of variable blocks $[\\mathbf{x}, \\boldsymbol{\\varphi}, \\boldsymbol{\\gamma}]$ and $[\\boldsymbol{\\phi}, \\boldsymbol{\\rho}, \\boldsymbol{\\zeta}, \\boldsymbol{\\psi}]$. The $\\upsilon$-rewrite is what turns the non-convex server cost into a convex function, and the QCQP/SDR transformation with Hungarian rounding is what handles the binary user-association constraint.","core_discovery":"The paper's discovery is that the joint DPE-maximization problem can be reduced through a chain of transformations instead of being solved directly. Auxiliary variables turn the sum of ratios into a sum of linear terms, KKT-type multipliers move the ratios back into the objective, a fractional-programming rewrite with an auxiliary variable makes the non-convex server cost convex, and the remaining mixed-integer problem becomes a QCQP that semidefinite relaxation plus Hungarian rounding solves. Alternating these blocks yields a stationary point of the transformed problem, and the paper claims this stationary point corresponds to a high-DPE operating point. Simulations support this by showing DAUR surpassing all baselines at every tested bandwidth, server capacity, user capacity, and transmit power level.","pith_inferences":["If the claimed equality $F=G$ in the fractional-programming step is checked numerically at a few feasible points, the set where it holds exactly is where the equivalence proof succeeds; outside that set, DAUR may be maximizing a surrogate objective rather than the original DPE.","The same transformation chain could be tested on smaller ratio problems with known optima to quantify the gap introduced by the SDR rounding step, which the paper does not isolate.","The centralized nature of DAUR, acknowledged in the conclusion, suggests that a distributed version would be needed for privacy-preserving Metaverse deployments; the alternating structure is a natural starting point for such an extension."],"forward_implications":["DPE gives operators a single number that trades off delay and energy; tuning the weights $\\omega_t$ and $\\omega_e$ shifts the operating point from delay-optimal to energy-optimal.","The alternating FP/QCQP/SDR decomposition makes the mixed-integer problem solvable in polynomial time per iteration, with complexity of order $O((N^{3.5}+M^{3.5}+N^{3.5}M^{3.5})\\log(1/\\epsilon))$ as reported in the paper.","In the simulated settings, DAUR outperforms random and greedy user-association baselines and average-resource-allocation baselines, and its advantage persists as bandwidth, server capacity, user capacity, and transmit power vary.","The same algorithmic recipe is claimed to extend to energy-efficiency and utility-cost objectives when combined with successive convex approximation for non-concave utilities."],"supporting_citations":[{"why":"Supplies the block propagation and validation delay model used in the server cost and the reference for consensus delay components.","marker":"[5]"},{"why":"Hungarian algorithm used to round the continuous SDR solution to a binary user-association matrix.","marker":"[10]"},{"why":"Full version of the paper containing the detailed proofs of Lemmas 5.2 through 5.9 on which the transformation chain rests.","marker":"[14]"},{"why":"Shannon formula defining the transmission rate $r_{n,m}$ used throughout the wireless cost model.","marker":"[20]"},{"why":"Fractional-programming formulation for Metaverse utility-cost problems that motivates the DPE ratio transformation.","marker":"[24]"},{"why":"Provides the worst-case complexity expression used for the convex solver steps and a joint offloading/user-association baseline.","marker":"[3]"}],"fun_headline_variants":["DAUR: alternating optimization maximizes Metaverse data efficiency","Joint resource tuning boosts Metaverse data processing efficiency","Semidefinite relaxation solves Metaverse resource allocation for DPE","DPE-aware association and resource allocation lifts Metaverse efficiency"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire convexification rests on the claim that a specific algebraic rewrite of the transmission cost is exactly equal to the original transmission cost; if that equality is off, DAUR maximizes a different objective than the DPE it reports.","fun_headline_variants_meta":{"raw":{"variants":["DAUR: alternating optimization maximizes Metaverse data efficiency","Joint resource tuning boosts Metaverse data processing efficiency","Semidefinite relaxation solves Metaverse resource allocation for DPE","DPE-aware association and resource allocation lifts Metaverse efficiency"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000715,"raw_usage":{"total_tokens":3164,"prompt_tokens":843,"completion_tokens":2321,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":459,"completion_tokens_details":{"reasoning_tokens":2266}},"tokens_in":459,"tokens_out":2321,"duration_ms":14519,"temperature":1.0,"reasoning_tokens":2266,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T13:33:55.522980+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate both sides of the claimed equality $F=G$ at a randomly chosen feasible point, substituting $\\upsilon_{n,m}^{(s)} = 1/(2 x_{n,m}\\rho_n p_n \\varphi_n d_n r_{n,m})$; if the numerical values differ, the convexified Problem P5 is not equivalent to Problem P4, and the algorithm's solution is not a DPE maximizer.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the block propagation and validation delay model used in the server cost and the reference for consensus delay components."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Hungarian algorithm used to round the continuous SDR solution to a binary user-association matrix."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Full version of the paper containing the detailed proofs of Lemmas 5.2 through 5.9 on which the transformation chain rests."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shannon formula defining the transmission rate $r_{n,m}$ used throughout the wireless cost model."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Fractional-programming formulation for Metaverse utility-cost problems that motivates the DPE ratio transformation."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the worst-case complexity expression used for the convex solver steps and a joint offloading/user-association baseline."}],"review_version":1}