{"id":"eb80ad20-28f2-4117-8c2f-feb8f0dc7560","arxiv_id":"2411.17282","paper_version":1,"verdict":"REJECT","confidence":"HIGH","novelty_score":2.0,"correctness_risk":"high","formal_verification":"none","parameter_count":9,"one_line_summary":"COVO is a COVID-19-inspired random-search optimizer whose claimed benchmark superiority is not supported by the reported experimental protocol or statistics.","lead":"This paper introduces COVO, a new optimization algorithm inspired by social distancing rules from the COVID-19 pandemic, and tests it on 13 benchmark problems plus an ECG noise removal task. The authors claim COVO beats ten existing optimization methods on all benchmarks, but the paper's own statistical tests find no significant difference.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Table V's noise-removal results contradict the paper's headline application: reconstructed MSE exceeds noisy MSE in all four sample sizes, so COVO is not demonstrated to improve ECG denoising.","rationale":"The strongest claim has two advertised parts: a new optimizer that beats ten baselines, and an ECG denoising application. The reader's weakest assumption was fairness of the benchmark comparison. I agree that is a serious issue: Section IV gives no dimension, iteration budget, number of runs, initialization scheme, or baseline tuning, and the reported Friedman/Wilcoxon p-values are all above 0.05, so 'better in all 13 cases' is not statistically supported. However, the single most load-bearing concern is the internal contradiction in the denoising evidence: Table V shows MSEo-r > MSEo-c for every row. That means under the metric chosen by the authors, the COVO-reconstructed signal is worse than the contaminated signal, undermining the practical utility claim in the title and abstract. No external comparison is needed to see this; it is visible inside the paper. The critique is on the data versus the conclusion, not on the authors. A referee should request the raw code and data and rerun the ECG experiment; if the table values are accurate, the denoising claim cannot stand. Since the reader already rejects the paper, my verdict remains unchanged.","tokens_in":19209,"tokens_out":8001,"duration_ms":79129,"concrete_test":"Reproduce the Section IV-C experiment: take the same four ECG segments, contaminate with Gaussian noise (mean=0, sd=1), optimize the ICA unmixing matrix W with COVO, back-project, and compute MSEo-r and MSEo-c for each sample size. If MSEo-r >= MSEo-c for any of the four cases, the claim that COVO removes noise is refuted by the paper's own metric. The test should also report SNR before and after reconstruction and inspect the reconstructed waveform; Table V alone is sufficient to decide the contradiction if the numbers are reproduced exactly.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper claims in the abstract and Section IV-C that COVO improves ECG noise removal and that the reconstructed signal is closer to the original. Table V reports MSEo-r (original vs reconstructed) and MSEo-c (original vs Gaussian-contaminated) for sample sizes 100-103: rows are (1.584837, 0.997821), (1.532555, 0.99999), (0.18787, 0.99889), and (0.18787, 0.998061). In every row MSEo-r > MSEo-c, so the 'denoised' signal is farther from the original ECG than the noisy input was. Section IV-C states that 'results for MSE ... demonstrate that COVO improves the demixing matrix for noise reduction,' but the only quantitative evidence in the paper shows the opposite. This is not a matter of missing baseline settings or statistical power; it is an internal contradiction in the data presented. If the noise-removal application is one of the two advertised contributions, this alone contradicts that contribution. The benchmark comparison is also unsupported by the reported significance tests (COVO's Friedman p=0.9259, Wilcoxon p=0.1277), but Table V is the cleanest load-bearing flaw because it does not depend on any external comparison protocol.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript introduces COVO, a population-based metaheuristic inspired by COVID-19 social distancing, and claims that it outperforms ten published metaheuristics on 13 standard benchmark functions and improves ECG signal noise removal by optimizing the ICA demixing matrix. The paper provides Algorithm 1, benchmark definitions, convergence plots, statistical tables, and a noise-removal experiment. The central claims are that COVO reaches a fitness error of 1.23e-18, converges better than all baselines in all 13 test cases, and produces a denoised ECG signal closer to the original than the noisy input.","tokens_in":19486,"tokens_out":5265,"duration_ms":46327,"significance":"If the claims were supported, COVO would be a competitive stochastic optimizer and a potentially useful denoising tool, adding to the family of epidemiology-inspired metaheuristics. The pseudocode and benchmark definitions are clearly stated, and the authors explicitly report p-values from Friedman and Wilcoxon tests rather than omitting them. However, the reported data contradict the central claims: the noise-removal table shows the reconstructed signal is farther from the original than the contaminated signal in every row, and the reported significance tests find no statistically significant advantage for COVO. With no experimental protocol and no reproducible numerical comparison, the contribution is not established. I credit the authors for including the full algorithm listing and for reporting the negative significance-test results, but these strengths do not overcome the internal contradictions.","major_comments":[{"comment":"The noise-removal application is contradicted by the paper's own table. For sample sizes 100-103, the reported MSEo-r values (1.584837, 1.532555, 0.18787, 0.18787) are all larger than the corresponding MSEo-c values (0.997821, 0.99999, 0.99889, 0.998061). Since MSEo-r is the error between the original and reconstructed ECG and MSEo-c is the error between the original and the contaminated input, the reconstructed signal is farther from the original than the noisy input in every row. Section IV-C's statement that the results 'demonstrate that COVO improves the demixing matrix for noise reduction' is therefore directly contradicted by Table V. This invalidates one of the two advertised contributions of the paper.","section":"IV-C, Table V"},{"comment":"The significance tests reported by the authors contradict the headline comparison. Table VII gives COVO a Friedman p-value of 0.925938, Table VIII gives a Wilcoxon p-value of 0.127665, and Table IX gives a P-test p-value of 0.936619; all exceed 0.05, and the text in Section IV-E explicitly says that no statistically significant differences were found. Section IV-A nonetheless concludes that COVO 'has performed better than the traditional models in all 13 cases' and is 'highly convergent.' These statements cannot both be true on the reported evidence. Either the statistical tests are mis-specified or the comparison claims are over-stated; as written, the data do not support a significant advantage for COVO.","section":"IV-E, Tables VII-IX"},{"comment":"No experimental protocol is reported. The reader is not told the dimension of the benchmark functions, the number of independent runs, the iteration or fitness-evaluation budget, the initialization scheme, or the termination criterion, and there is no indication that the ten baseline algorithms were run under comparable budgets or parameter configurations. Table IV reports statistics for COVO only, and the comparison values appear only in Figure 5 and in scattered textual statements rather than in a complete numeric table. Without this information, the claim of superiority over ten baselines on 13 functions is not auditable or reproducible.","section":"IV, Tables IV-VI"},{"comment":"The novelty claim is weakened by the relation to CVOA. Table I is explicitly headed 'COVO PARAMETERS [47]', and the update equations and death/recovery/spread structure in Section III-B closely follow the CVOA workflow of reference [47]. The social-distancing operator of Eq. (5) is the only clearly new component, but no ablation study or direct COVO-vs-CVOA comparison isolates its effect. The paper therefore does not demonstrate what the social-distance mechanism contributes to the reported results, and the claimed novelty is not supported by the experiments.","section":"III-B, Table I"}],"minor_comments":[{"comment":"The conclusion states that 'three real-world engineering challenges from IEEE-CEC 2011 are used to validate' the method, but no CEC-2011 experiments appear anywhere in Section IV; either add those experiments or remove the sentence.","section":"V"},{"comment":"The conclusion also calls the model 'parameter-free', yet Table I lists eleven parameters including N, S_rate, SS_rate, travelP, P_die, D_rate, H_dist, T, L, U, and Delta; this wording should be corrected.","section":"V"},{"comment":"Tables I and II disagree on the initial values: Table I reports P_die=0.43597 and D_rate=0.13955, while Table II reports P_die=1 and D_rate=0.888557; the manuscript should state which configuration generated the reported results.","section":"Tables I and II"},{"comment":"The conclusion mentions comparisons with 'eight cutting-edge metaheuristic approaches', but Section IV-A lists ten baselines; the count should be made consistent.","section":"V"},{"comment":"Equation (5) is typeset unclearly and is not used transparently in Algorithm 1: the piecewise definition and the role of Delta are unclear, and the pseudo-code compares H_dist with T without explaining how H_dist is computed from pairwise distances at each iteration.","section":"III-B, Eq. (5)"}],"recommendation":"reject","confidential_remarks":"The central problems are internal contradictions rather than presentation issues. Table V alone invalidates the noise-removal contribution, and the reported p-values in Tables VII-IX contradict the claim of superiority over the baselines. These flaws cannot be repaired by modest rewriting; a proper experimental study with specified dimensions, run counts, budgets, and baseline tuning, plus a corrected denoising experiment, would be needed before the manuscript could be reconsidered."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the take. This is not a new algorithm; it's CVOA with a threshold on a social-distancing distance, which the authors themselves acknowledge by labeling Table II as inherited from [47]. The only genuinely new piece is Eq. (5)'s H_dist operator and the PZ formalism. That transparency is a point in its favor: they don't hide the lineage.\n\nWhat the paper does well: it runs 13 standard benchmark functions, includes convergence plots, and attempts a concrete application (ECG denoising). The literature review is broad, and the writing is clear enough to see what was done.\n\nThe problems are load-bearing, not cosmetic. First, there is no experimental protocol anywhere: no dimension, iteration budget, number of runs, or baseline tuning. That alone makes the central comparison meaningless. Second, the paper's own statistical tests contradict the headline: COVO's Friedman p=0.9259 and Wilcoxon p=0.1277 are both above 0.05, so the data show no significant difference. Yet the text claims COVO 'performed better than the traditional models in all 13 cases.'\n\nThird, and most damning, the ECG noise removal table (Table V) contradicts the application claim. In all four rows, MSEo-r (original vs reconstructed) is larger than MSEo-c (original vs contaminated), which means the 'denoising' moved the signal farther from the original. The surrounding text says the opposite. That is an internal contradiction in the one piece of quantitative evidence offered for the application.\n\nThere are also smaller inconsistencies: parameter tables I and II give different initial values for the same symbols (e.g., H_dist 13.77 vs 0.873, P_die 0.43597 vs 1), and the conclusion mentions IEEE-CEC 2011 validation that never appears in the results.\n\nWho is this for? Not many people. A reader looking for a new optimizer gets a renamed CVOA; a reader looking for a noise-removal method gets a table that shows the method hurting the signal. The core claims are unsupported, and the application result is self-contradictory. I would not send this to referees; it is a desk reject. If the authors redo the experiments with a proper protocol and correct Table V, there might be a minor variant paper in it, but as submitted it does not warrant referee time.","headline":"A thin relabeling of CVOA whose own tables and significance tests contradict the paper's central claims; desk reject.","tokens_in":20046,"tokens_out":2979,"would_cite":false,"duration_ms":27732,"reading_group":"no","serious_thinker":"no","would_accept_peer_review":false},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A new optimizer built on infection, recovery, and distancing rules reports the best fitness error on all 13 standard benchmarks, and the same engine is used to clean ECG signals.","keywords":["COVID-19","Social Distancing Induced Coronavirus Optimization Algorithm","Metaheuristic Algorithms","Optimization Algorithm","Coronavirus Optimization Algorithm","Standard Benchmark Functions","ECG noise removal","convergence analysis"],"falsifier":"Run COVO and the ten baselines on F1–F13 with fixed dimensions, equal numbers of fitness evaluations, and many independent repeats, using the paper's listed parameter values, and check whether COVO's claimed best errors ($1.23\\times10^{-18}$ on F1, $0$ on F9) and its 100.5-second runtime reproduce. Since the paper's own Friedman and Wilcoxon tests give COVO p-values of 0.9259 and 0.1277 — both above 0.05 — recomputing those tests from the reported tables would already show that the data do not support a statistically significant difference.","tokens_in":18981,"feed_emoji":"🦠","tokens_out":13166,"duration_ms":106476,"temperature":0.7,"pith_summary":"The paper proposes COVO, a population-based metaheuristic whose update rules translate the COVID-19 epidemic response — infection, super-spreading, travel, recovery, death, and social distancing — into search moves. The authors claim that on thirteen standard benchmark functions, covering unimodal, step, quartic-noise, and multimodal landscapes, COVO converges faster and reaches lower fitness error than ten published optimizers, including a best error of $1.23\\times 10^{-18}$ on F1 and $0$ on F9. They further apply COVO to remove Gaussian noise from an ECG signal by optimizing the unmixing matrix of independent component analysis, reporting a mean square error of $0.18787$ between original and reconstructed signal. If the claims hold, COVO would be a competitive, gradient-free optimizer for complex multimodal problems and signal-denoising tasks. The paper also argues that the social-distancing rule lets the search terminate without a pre-set stopping criterion.","feed_headline":"COVID-inspired optimizer claims best score on all 13 benchmarks","feed_subtitle":"It models search agents as people who infect, recover, or stay apart, then uses the same engine to clean ECG signals.","key_machinery":"The load-bearing object is the social-distancing parameter $H_{dist}$, a distance-gated scalar computed from $\\lVert X_i^t - X_j^t\\rVert$ and the constant $\\Delta$ via equation (5). It selects among three position updates: a plain random restart within the bounds, a spread-rate-scaled restart $L + (U-L)S_{rate}$, and a super-spreader restart $L + (U-L)S_{Srate}$. The death probability $P_{die}$ doubles as an acceptance threshold in equation (8), deciding whether a candidate is updated or discarded, while chaotic maps generate $P_{die}$ and the death rate and opposition-based learning doubles the initial population. The intended effect is a balance of exploration through distancing-driven restarts and exploitation through spread-scaled moves, which the authors identify as the source of COVO's convergence speed and low error.","core_discovery":"On its own terms, COVO treats each candidate solution as a person who may be susceptible, infected, recovered, dead, or traveling. Infected individuals die with probability $P_{die}$; those who survive spread the virus at a normal rate $S_{rate}$ or, for super-spreaders, at $S_{Srate}$, and they may travel with probability $P_{travel}$. The key control is the social-distancing parameter $H_{dist}$, obtained from the pairwise distance $dist_{ij}^t = \\lVert X_i^t - X_j^t\\rVert$ and a constant $\\Delta$: $H_{dist}=\\Delta$ when the distance is at least $\\Delta$, otherwise $H_{dist} = |dist_{ij}^t - \\Delta|$. If $H_{dist}$ is below the threshold $T$, a zero-infected patient is restarted as $X = L + (U-L)$; if above, the restart is scaled by the spread rate, $X = L + (U-L)S_{rate}$; traveling individuals use the super-spreading rate instead. After fitness evaluation, a solution with fitness larger than $P_{die}$ is refined as $X_{new} = X_{old} \\pm P_{die} Fit$, while a solution with fitness below $P_{die}$ is declared dead and the population is re-initialized. The paper reports best, median, worst, mean, and standard deviation for all thirteen functions, states that COVO's errors are lower than EHO, SSA, SSO, SFO, BOA, BWO, SMO, CVOA, SRO, and GBRUN in all thirteen cases, and concludes that the algorithm is highly convergent.","pith_inferences":["Editorial inference: The paper reports one best value per function; the $1.23\\times10^{-18}$ figure is at most a single-trajectory statement until repeated runs with means and standard deviations are published.","Editorial inference: If $H_{dist}$ drives the claimed convergence, then an ablation that fixes $H_{dist}=0$, forcing only the random-restart branch, should measurably worsen results on multimodal functions F9–F13; this ablation is not in the paper and is a direct test of the mechanism.","Editorial inference: The conclusion's references to three IEEE-CEC 2011 engineering problems and to 'eight' comparison algorithms are not supported by the reported experiments, which cover thirteen benchmark functions and ten baselines; those statements should not be read as validated results.","Editorial inference: The same social-distancing gate could be attached to other population algorithms as a restart rule driven by population dispersion; whether the benefit transfers is an open, testable question."],"forward_implications":["If COVO's reported errors are reproducible, it would give practitioners a single optimizer that outperforms ten established metaheuristics on thirteen standard test functions spanning unimodal, step, quartic-noise, and multimodal landscapes.","The ECG experiment suggests a concrete use case: COVO-optimized ICA could lower reconstruction error from about 0.998 for the contaminated signal to 0.18787, indicating meaningful noise removal in blind source separation.","COVO's reported runtime of 100.5 seconds, lower than all ten compared methods, would make it attractive when fitness evaluations are expensive.","Because the algorithm uses no gradient information, it would apply to discrete as well as continuous and non-smooth optimization problems, as the paper claims.","The distance-gated restart mechanism, if verified, would be a reusable design pattern for other population-based optimizers on multimodal problems."],"supporting_citations":[{"why":"Supplies the Elephant Herding Optimization baseline whose fitness values COVO is compared against on all thirteen benchmark functions.","marker":"[35]"},{"why":"Supplies the Salp Swarm Algorithm baseline used in the convergence and statistical comparisons.","marker":"[40]"},{"why":"Supplies the Sparrow Search Algorithm baseline used in the convergence and statistical comparisons.","marker":"[44]"},{"why":"Supplies the Sailfish Optimizer baseline in the comparison tables.","marker":"[42]"},{"why":"Supplies the Butterfly Optimization Algorithm baseline in the comparison tables.","marker":"[41]"},{"why":"Supplies the Black Widow Optimization baseline in the comparison tables.","marker":"[43]"},{"why":"Supplies the Spider Monkey Optimization baseline in the comparison tables.","marker":"[32]"},{"why":"Supplies the coronavirus optimization model and the parameter values listed in Tables I and II that COVO inherits, and serves as the CVOA baseline.","marker":"[47]"},{"why":"Supplies the Ship Rescue Optimization baseline in the comparison tables.","marker":"[53]"},{"why":"Supplies the GBRUN baseline in the comparison tables.","marker":"[54]"}],"fun_headline_variants":["COVID-inspired optimizer beats 10 rivals on all 13 benchmarks","Social-distancing algorithm COVO outperforms 13 benchmark tests","COVID-inspired search algorithm wins all 13 optimization benchmarks","New optimizer models social distancing to beat 13 test functions","Pandemic-inspired optimizer claims top spot on all 13 benchmarks"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The central claim assumes the comparison is fair: all ten baseline algorithms ran under identical, properly tuned conditions with the same initialization, dimension, iteration budget, and number of runs, yet the results section (Section IV, Tables IV–VIII) reports none of those settings.","fun_headline_variants_meta":{"raw":{"variants":["COVID-inspired optimizer beats 10 rivals on all 13 benchmarks","Social-distancing algorithm COVO outperforms 13 benchmark tests","COVID-inspired search algorithm wins all 13 optimization benchmarks","New optimizer models social distancing to beat 13 test functions","Pandemic-inspired optimizer claims top spot on all 13 benchmarks"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000719,"raw_usage":{"total_tokens":3308,"prompt_tokens":1102,"completion_tokens":2206,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":718,"completion_tokens_details":{"reasoning_tokens":2123}},"tokens_in":718,"tokens_out":2206,"duration_ms":15138,"temperature":1.0,"reasoning_tokens":2123,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T12:19:10.142367+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run COVO and the ten baselines on F1–F13 with fixed dimensions, equal numbers of fitness evaluations, and many independent repeats, using the paper's listed parameter values, and check whether COVO's claimed best errors ($1.23\\times10^{-18}$ on F1, $0$ on F9) and its 100.5-second runtime reproduce. Since the paper's own Friedman and Wilcoxon tests give COVO p-values of 0.9259 and 0.1277 — both above 0.05 — recomputing those tests from the reported tables would already show that the data do not support a statistically significant difference.","supporting_citations":[{"cited_title":"Elephant Herding Optimization,","cited_arxiv_id":null,"evidence_quote":"Supplies the Elephant Herding Optimization baseline whose fitness values COVO is compared against on all thirteen benchmark functions."},{"cited_title":"The Sailfish Optimizer: A novel nature -inspired metaheuristic algorithm for solving constrained engineering optimization problems,","cited_arxiv_id":null,"evidence_quote":"Supplies the Sailfish Optimizer baseline in the comparison tables."},{"cited_title":"Butterfly optimization algorithm: a novel approach for global optimization,","cited_arxiv_id":null,"evidence_quote":"Supplies the Butterfly Optimization Algorithm baseline in the comparison tables."},{"cited_title":"Black Widow Optimization Algorithm: A novel meta-heuristic approach for solving engineering optimization problems,","cited_arxiv_id":null,"evidence_quote":"Supplies the Black Widow Optimization baseline in the comparison tables."},{"cited_title":"Coronavirus Optimization Algorithm: A Bioinspired Metaheuristic Based on the COVID -19 Propagation Model,","cited_arxiv_id":null,"evidence_quote":"Supplies the coronavirus optimization model and the parameter values listed in Tables I and II that COVO inherits, and serves as the CVOA baseline."},{"cited_title":"Ship Rescue Optimization: A New Metaheuristic Algorithm for Solving Engineering Problems,","cited_arxiv_id":null,"evidence_quote":"Supplies the Ship Rescue Optimization baseline in the comparison tables."},{"cited_title":"GBRUN: A Gradient Search-based Binary Runge Kutta Optimizer for Feature Selection,","cited_arxiv_id":null,"evidence_quote":"Supplies the GBRUN baseline in the comparison tables."}],"review_version":1}