{"id":"f44f6890-8ce7-4efd-9213-5336547a07bb","arxiv_id":"1908.03076","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A weaker anomalous diffuser can prolong its survival against a stronger normal diffuser by restarting its memory at its peak, a 'selective recalling-forgetting' strategy.","lead":"This paper models a race between two spreading processes, where the faster one normally wins. It argues that the slower process, if it carries memory of its past, can survive longer by deliberately forgetting that memory at the moment it peaks.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The strategy's prescribed restart at the peak is chosen from the same simulated I2 curve that the strategy is claimed to improve; no comparison over restart times or derivation establishes that peak restart maximizes survival.","rationale":"The reader's weakest assumption is exactly the optimality of the peak restart, and I agree with that identification. The paper defines the strategy at the peak of the simulated I2 curve and then claims that this strategy extends survival; unless all restart times are compared, the claim of 'maximum survival' remains circular. This is a genuine gap, but it is testable and may be true, so conditional acceptance remains appropriate. No independent verification of the numerics is provided (Appendix Eq. 13), but that is secondary to the missing optimization over restart times. If the proposed scan is run and supports the peak as the optimal restart time, the central claim would be materially strengthened. I do not see an internal inconsistency in the fractional model itself that is more load-bearing than this unproven optimality assumption.","tokens_in":9468,"tokens_out":6786,"duration_ms":75736,"concrete_test":"For fixed gamma=0.995, alpha=0.5, choose a threshold epsilon (e.g., the minimum market share used in Fig. 6). For a grid of restart times tau in [0,T], solve the piecewise system (9)-(11) with the Caputo lower terminal moved to tau, and define T_epsilon(tau) as the last time I2 remains above epsilon (or the time I2 first crosses epsilon). Find argmax T_epsilon(tau). If it is not tau = t_peak, the paper's prescription is suboptimal; if it is t_peak, repeat for several gamma and alpha values and halve the step h in Eq. (13) to check that the optimal restart time and the added lifetime Delta_tau are stable.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that the weaker competitor maximizes survival by restarting its Caputo memory at the peak t* = argmax I2 (Sec. IV, Fig. 3). This is the load-bearing step: if another restart time gives a longer time above a pre-defined minimum share, then the 'selective recalling-forgetting strategy' as specified is not optimal, and the abstract's promise of 'maximum survival' fails. The paper never performs this comparison. It selects t* because it is the maximum of the constant-memory solution, so the strategy is defined by the outcome it is supposed to explain. The peak maximizes the starting value of the restarted process, not necessarily the survival duration: after restart the system is governed by a new Caputo derivative with lower terminal t*, and the subsequent trajectory depends on I1(t*) and S(t*) as well as on I2(t*). Without a scan over restart times, or an analytic argument, the claim 'continue until the peak point... then restart' is not established. The numerical evidence also lacks convergence checks (Appendix Eq. 13), so it is not known whether the small gains shown persist at smaller step sizes; however, the optimality gap is the more fundamental issue.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a deterministic competition model between a normal diffuser with higher growth rate I1 and an anomalous diffuser with lower growth rate I2, where I2 evolves under a fractional-order (Caputo-type) memory. For a memory order alpha=0.5 and relative growth rate gamma=0.995, the numerical solution shows that memory delays the decline of I2. The authors then introduce a 'selective recalling-forgetting strategy' in which the memory is reset at the peak of the I2 curve, and present numerical evidence that this restart extends the survival time of I2 above a threshold. The paper frames this as a general strategy for weaker competitors, with a business interpretation and a sweep over gamma for the added survival time. The central claim that this strategy achieves maximum survival is asserted rather than proven.","tokens_in":9741,"tokens_out":11398,"duration_ms":111591,"significance":"If the maximum-survival claim were rigorously established, the paper would offer a novel and practically relevant control strategy for weaker competitors in social, economic, or epidemiological competition models. The model is transparent, the fractional-order memory mechanism connects to a substantial prior literature, and the numerical implementation via the fractional Adams scheme is reproducible in principle. The gamma-sweep figures (Figs. 5 and 6) provide a useful predictive map for when the strategy is beneficial. The main weaknesses are that the optimality of the peak restart is not proven, only one memory order is demonstrated, the survival threshold is not specified, and no numerical convergence checks are reported. These are load-bearing for the paper's central promise but appear addressable within the scope of a revision.","major_comments":[{"comment":"The central claim that restarting the memory at the peak t* = argmax I2 yields 'maximum survival' is not established. The paper does not scan over restart times, nor does it provide an analytic argument that this t* maximizes the time spent above a given threshold. Because the post-restart trajectory is governed by a new Caputo initial-value problem starting at t*, with state (I1(t*), S(t*), I2(t*)), a restart later than the peak might produce longer survival even if it starts from a lower I2. The selection of the peak in Fig. 3 is therefore an ansatz based on the same simulated curve that serves as the baseline, and the abstract's promise of maximum survival exceeds what is demonstrated. At minimum, the authors should compare survival times over a grid of restart times, or prove a condition under which the peak is optimal.","section":"Sec. IV, Fig. 3"},{"comment":"The strategy is tested numerically for only one memory order (alpha=0.5) and, in the illustrative comparison of Fig. 3, one growth rate (gamma=0.995). Figure 6 varies gamma but not alpha, and the heatmap in Fig. 5 uses the same alpha. The paper therefore does not show that the peak-restart strategy remains beneficial or near-optimal for other memory strengths, even though alpha is presented as a tunable parameter of the model. A scan over alpha (and over the survival threshold) is needed to support the general claims made in Sections I and VI.","section":"Sec. V, Figs. 5 and 6"},{"comment":"No numerical convergence study is reported. The step size h is not stated, and the reported improvements in survival time (e.g., Delta tau in Fig. 6) could be affected by discretization error in the fractional Adams scheme. The authors should report h, the error tolerance, and a convergence check (e.g., Delta tau versus h, or a Richardson estimate) to demonstrate that the observed strategy gains are not numerical artifacts.","section":"Appendix, Eqs. (13)-(14)"},{"comment":"Equation (8) is not the standard Caputo fractional derivative as written: for 0<alpha<1 the prefactor should be 1/Gamma(1-alpha) and the kernel should be (t-tau)^{-alpha}, not Gamma(alpha-1) with (t-tau)^{alpha-2}. Because Eq. (7) is used in the derivation and Eq. (11) is implemented numerically with Eqs. (13)-(14), this may be a typographical error, but as written it makes the model inconsistent: for alpha=0.5 the specified kernel is negative, which would reverse the sign of the memory forcing in Eq. (5). Please correct the definition or state explicitly which fractional derivative is used.","section":"Sec. III, Eq. (8)"}],"minor_comments":[{"comment":"The intersection term (I1 ∩ I2) appears in the normalization condition Eq. (1) but is never used in the subsequent equations; please define it and state whether it is negligible or remove it to avoid confusion.","section":"Eq. (1)"},{"comment":"The model is an ODE system without spatial derivatives, so the terms 'diffusion' and 'reaction-diffusion' may mislead readers; please clarify that the competition is temporal, with spatial language used only as a metaphor.","section":"Title and Sec. II"},{"comment":"The notation c t* D^alpha 103 is unclear; the terminal time appears as '103' and the starting and ending points of the fractional derivative are ambiguous. Please define the piecewise operators explicitly with proper limits.","section":"Sec. IV"},{"comment":"The parameter Delta tau is described in the caption as the added lifetime for a predefined minimum proportion, but the threshold is never specified in the text or figure; please state the threshold used to compute survival time in Figs. 3 and 6.","section":"Fig. 3 and Sec. V"}],"recommendation":"major_revision","confidential_remarks":"The paper has a promising core idea, but the main claim of optimality is not supported. I would ask the authors to add a scan over restart times, a scan over alpha, convergence checks, and to fix the fractional-derivative definition. If these are supplied, the paper could become publishable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The new thing here is the restart idea: the weaker competitor evolves with a Caputo-fractional memory until its share peaks, then the memory origin resets and the process continues. That piecewise 'selective recalling-forgetting' strategy is not in the earlier fractional SIR work, and it is a legitimate extension worth thinking about.\n\nWhat the paper does well: the model is simple and internally consistent, the memory effect visibly slows the weaker competitor's decline, and the authors are honest that the business case is a proof of concept. The cumulative-market-share comparison and the scan over gamma (Figs. 4-6) are reasonable exploratory steps. The numerics use a standard fractional Adams method, and the appendix points to the right references, though without convergence checks.\n\nThe soft spot is load-bearing. The central claim is that restarting at the peak gives 'maximum survival,' but t* is chosen as the argmax of the no-strategy memory curve. That means the strategy is defined by the outcome it is supposed to explain. The paper never scans over alternative restart times, nor does it offer an optimality argument. A peak point maximizes the starting value of I2 after restart, but survival duration also depends on I1(t*) and S(t*), so there is no reason to expect argmax I2 to be the argmax of survival time. For all the simulations show, restarting a bit earlier or later could do better. The main demonstration also uses only one fractional order (alpha=0.5) and one growth rate (gamma=0.995); Fig. 6 is helpful but does not fix the optimality gap.\n\nIf the claim were softened to 'restarting at the peak extends survival in these simulations,' I would have little complaint. As written, 'maximum survival' is an overclaim on the current evidence. The paper is not sloppy in its equations, and the strategy is plausible as a heuristic, so this is a fixable issue rather than a fatal one.\n\nWho this is for: people working on fractional-order models of social or economic competition, and anyone interested in memory-reset mechanisms. It is a modest but genuinely new contribution, and it deserves a serious referee—one who should ask for either a one-dimensional optimality argument in restart time or a clean scan over restart times and parameters, plus numerical convergence checks.","headline":"A coherent fractional-competition model with a plausible but unproven restart-at-peak strategy; the survival claim needs a restart-time scan or a proof before it can be taken literally.","tokens_in":10241,"tokens_out":1512,"would_cite":false,"duration_ms":18775,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"By resetting its memory at the moment its market share peaks, a weaker competitor can postpone its decline and survive longer than constant memory would allow.","keywords":["anomalous diffusion","fractional calculus","Caputo derivative","memory effects","competition model","survival strategy","SI model","market share"],"falsifier":"Run the same competition model with the memory restart applied at a different time, before or after the peak, and compare the survival time $\\Delta\\tau$ and cumulative share of $I_2$ with the peak-restart values; if any other restart time consistently outperforms the peak, the central claim that the peak is the right restart point fails. A second check is numerical: the claim relies on the fractional solver being accurate at $\\alpha=0.5$ and $\\gamma=0.995$, so a convergence test over decreasing step sizes would settle whether the reported peak exists in the continuous system at all.","tokens_in":9323,"feed_emoji":"🧠","tokens_out":10028,"duration_ms":94810,"temperature":0.7,"pith_summary":"The paper asks how a diffusion process with a lower growth rate can survive a competition against a faster normal-diffusion rival, and proposes that selectively forgetting its own history at the right moment does the trick. In the model, the weaker competitor is described by a fractional-order derivative that remembers its past states, and without any strategy it is doomed to vanish once the shared resource is exhausted. The central finding is the \"selective recalling-forgetting strategy\": run the anomalous diffusion with full memory up to its peak, then restart the memory at that peak as if the past had been forgotten. This reset postpones the weaker competitor's decline and lengthens the time it holds a given minimum share. If the result holds, any system that can tune its own memory—a firm in a market, a species, an epidemic—may buy extra survival time against a stronger rival.","feed_headline":"A losing diffuser can outlast its rival by forgetting at the peak","feed_subtitle":"A slower competitor holds its minimum share longer by restarting memory at the peak.","key_machinery":"The load-bearing mechanism is a piecewise restart of the Caputo fractional derivative in the equation for the weaker competitor. The Caputo derivative of order $\\alpha$ replaces the ordinary time derivative with an integral over the whole past weighted by a power-law kernel, so it carries memory of previous states; the paper uses $\\alpha=0.5$ for long-lasting memory. The strategy is implemented by splitting the fractional operator at the peak time $t^*$: $c_0 D_t^\\alpha$ before $t^*$ and $c_{t^*} D_t^\\alpha$ after it. At $t^*$ the accumulated history is discarded, and the past no longer pulls the declining trajectory downward. This reset, localized at the peak of the memory curve, is the mechanism the paper credits for the extra survival time.","core_discovery":"The discovery, stated on the paper's own terms, is that the weaker competitor can use its memory as a survival resource. The competition is modeled by three coupled equations: the normal diffuser $I_1$ grows with rate $1$, the anomalous diffuser $I_2$ grows with rate $\\gamma < 1$ and follows a Caputo fractional derivative of order $\\alpha = 0.5$, and a shared source $S$ feeds both until it is exhausted at the critical time $t_c$. Without memory, $I_2$ peaks at $t_c$ and then decays to zero. With constant memory, the decline is slower but still terminal. The paper's result is that restarting the memory at the peak $t^*$—forgetting everything before that moment and running a new fractional process from the peak—extends the lifetime $\\Delta\\tau$ of $I_2$ and raises its cumulative share, most clearly when $\\gamma$ is close to $1$. The same mechanism is cast as a business proof of concept in which a small firm preserves a minimum market share for longer.","pith_inferences":["Inference: the peak-restart rule is chosen from the simulated curve, not derived; optimizing the restart time for each pair $(\\gamma, \\alpha)$ would show whether the peak is truly the best reset point, which is a direct testable extension.","Inference: the reset mechanism resembles restart protocols in stochastic search, where resetting a process to a favorable state can shorten hitting times; reading the peak as a favorable restart state connects this deterministic model to a broader class of restart problems, a connection the paper does not draw.","Inference: if one reset helps, repeating it at each successive peak could extend survival even further; the paper tests only a single reset, so a periodic memory-reset strategy is an unexamined corollary of its logic.","Inference: the strategy assumes a diffuser can deliberately wipe its own memory, which in a market setting means a firm can erase accumulated customer history; the practical benefit would depend on how cleanly such a reset can actually be executed."],"forward_implications":["In a tight competition with $\\gamma$ close to $1$, restarting the memory at the peak lengthens the interval $\\Delta\\tau$ over which the weaker diffuser holds a minimum share and raises its cumulative share above both the no-memory and constant-memory cases.","For relative growth rates in roughly $0.6 < \\gamma < 0.7$ and near $\\gamma \\simeq 1$, the cumulative-share ratio of strategy-with-memory to memory-only exceeds $1$, so the paper recommends running the strategy in those regimes.","As $\\gamma$ approaches $0$, the added survival time $\\Delta\\tau$ approaches $0$, so the strategy offers almost nothing when the rival is overwhelmingly stronger.","With $\\alpha = 1$, the fractional system reduces to the integer-order memoryless competition, so the standard model is a special case of the generalized one.","In the business interpretation, a smaller firm facing a dominant market can treat the memory reset at the peak as a concrete policy for prolonging the lifetime of a minimum market share."],"supporting_citations":[{"why":"Supplies the earlier memory-effect epidemic model that motivates representing the weaker competitor's evolution with a fractional derivative.","marker":"[1]"},{"why":"Shows the competition equations are a Lotka-Volterra competition model, validating the structure of the dynamics.","marker":"[3]"},{"why":"Defines the Caputo fractional derivative used in the anomalous diffusion equation.","marker":"[14]"},{"why":"Supports the claim that fractional derivatives represent power-law memory in the system.","marker":"[15]"},{"why":"Provide the fractional Adams discretization and its stability analysis, which produce the numerical solutions from which the peak-restart strategy is read.","marker":"[31, 32]"},{"why":"Models competition and evolution in restricted space, the same squeezed-resource setting the paper uses to describe the post-peak contest.","marker":"[29]"},{"why":"Motivates the idea that weaker competitors can use strategy to survive, the background context for the selective recalling-forgetting strategy.","marker":"[8]"}],"fun_headline_variants":["Slow diffuser outlasts rival by forgetting at its peak","Anomalous diffusion survives longer with memory reset at peak","Forgetting the past at the peak extends a loser's survival","Reset memory at peak to boost survival in diffusion race"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The paper assumes that restarting the memory at the peak is the best possible restart time, but it chooses that moment because the simulated curve peaks there rather than proving it is optimal.","fun_headline_variants_meta":{"raw":{"variants":["Slow diffuser outlasts rival by forgetting at its peak","Anomalous diffusion survives longer with memory reset at peak","Forgetting the past at the peak extends a loser's survival","Reset memory at peak to boost survival in diffusion race"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000165,"raw_usage":{"total_tokens":1240,"prompt_tokens":926,"completion_tokens":314,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":542,"completion_tokens_details":{"reasoning_tokens":246}},"tokens_in":542,"tokens_out":314,"duration_ms":3789,"temperature":1.0,"reasoning_tokens":246,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:37:33.929433+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same competition model with the memory restart applied at a different time, before or after the peak, and compare the survival time $\\Delta\\tau$ and cumulative share of $I_2$ with the peak-restart values; if any other restart time consistently outperforms the peak, the central claim that the peak is the right restart point fails. A second check is numerical: the claim relies on the fractional solver being accurate at $\\alpha=0.5$ and $\\gamma=0.995$, so a convergence test over decreasing step sizes would settle whether the reported peak exists in the continuous system at all.","supporting_citations":[{"cited_title":"stronger","cited_arxiv_id":null,"evidence_quote":"Supplies the earlier memory-effect epidemic model that motivates representing the weaker competitor's evolution with a fractional derivative."},{"cited_title":"Gossip: Identifying central individuals in a so- cial network,","cited_arxiv_id":null,"evidence_quote":"Shows the competition equations are a Lotka-Volterra competition model, validating the structure of the dynamics."},{"cited_title":"Crowd-funding: Cash on demand,","cited_arxiv_id":null,"evidence_quote":"Defines the Caputo fractional derivative used in the anomalous diffusion equation."},{"cited_title":"Fractional differential equations, acad,","cited_arxiv_id":null,"evidence_quote":"Supports the claim that fractional derivatives represent power-law memory in the system."},{"cited_title":"Emergence of scaling in random networks,","cited_arxiv_id":null,"evidence_quote":"Models competition and evolution in restricted space, the same squeezed-resource setting the paper uses to describe the post-peak contest."},{"cited_title":"Fractional dynamics of net- work growth constrained by aging node interactions,","cited_arxiv_id":null,"evidence_quote":"Motivates the idea that weaker competitors can use strategy to survive, the background context for the selective recalling-forgetting strategy."}],"review_version":1}