{"id":"a15d7530-5aca-47ed-8ce0-c59eb6d5702d","arxiv_id":"1908.07242","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For a persistent random walk with arbitrary temporal correlations, multiple peaks in the displacement distribution occur only when forward persistence is present; temporal correlations reshape the peaks but do not create them.","lead":"This paper derives the exact short-time displacement distribution for an active random walker whose hops can repeat, reverse, or randomize direction, with arbitrary waiting-time correlations. It shows that extra peaks in the distribution require persistent forward motion, which could let experiments read directional persistence from displacement data.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Section IV's theorem omits the n=0 term: even with γf=0 and delta step lengths, the central Gaussian peak plus the one-hop peak can give two maxima, so the claim 'peaks imply persistence' is false as stated.","rationale":"The concern is load-bearing because the central message of the paper is the inverse implication 'more than one peak ⇒ positive directional correlations'. The n=0 term is part of the model (Eqs. (4)-(5)) and gives a peak at zero for any positive q0; the one-hop term gives a second peak near the preferred step length. This is not an edge case of extreme parameters: the three temporal distributions used in the paper all have q0,q1>0. In Fig. 3 the authors themselves identify the r=0 feature as a peak. Thus the theorem as written is internally contradicted, not merely incomplete. The reader's identified assumption about delta step lengths / Gaussian smearing is a related but distinct issue; it also invalidates the unqualified claim when s(l) is multimodal. Both are repaired by a clear qualification, so I would keep the conditional verdict rather than reject outright. The exact calculation in Section III is not affected.","tokens_in":16661,"tokens_out":11561,"duration_ms":114472,"concrete_test":"In d=3, evaluate the exact result, e.g. Eq. (12) with Eq. (5), for γf=γb=0, s(l)=δ(l-a), Poissonian q_n(τ) at τ=1, a=1, α=0.1. Count local maxima of Pr(r) on r∈[0,3]. If there are two maxima (at r=0 and near r=a), the Section IV theorem is false as stated. As a control, set q0=0 and check that the central maximum disappears, isolating the n=0 contribution.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section IV claims that γf=0 implies at most one peak in Pr(r,τ). The proof reduces the problem to showing p1 has one peak and pn≥2 are non-increasing, but it never treats p0, which contributes a peak at r=0. In the paper's own temporal distributions q0(τ)>0 and q1(τ)>0 (e.g. Poisson q0=e^{-τ}, q1=τe^{-τ}); with s(l)=δ(l-a) and small α, the n=0 contribution is a Gaussian centered at zero and the n=1 contribution is a shell peaked near r=a. Hence Pr has two local maxima with γf=γb=0, directly contradicting the abstract and Section IV. This does not rely on the smearing assumption the reader questioned. A separate gap is that for multimodal s(l), p1(r)=s(r)/(Ω_d r^{d-1}) already has several peaks at γf=0. Both failures show the central claim needs qualification (e.g. condition on at least one hop, exclude the r=0 peak, and restrict to single-peaked s).","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies a discrete-time random-walk model for active motion in d dimensions, in which each hop is either forward with probability γf, backward with probability γb, or uncorrelated (random length and direction) with probability γ0. The timing of hops is governed by arbitrary waiting-time distributions qn(τ), and a Gaussian continuous process is added by convolution. The authors derive a Fourier-space expression for the displacement distribution, give explicit formulas for small n and for three step-length distributions (Dirac, modified Gaussian, Cauchy), and present numerical plots for three temporal distributions (Poisson, binomial, geometric). The paper's central analytical claim is in Section IV: if γf=0, the displacement distribution has at most one peak, so the presence of more than one peak implies positive persistence. The abstract and summary generalize this to say that non-monotonicity of the displacement distribution occurs only when persistence is strong enough.","tokens_in":16888,"tokens_out":5708,"duration_ms":66473,"significance":"If the central theorem were valid, the paper would provide a clean and broadly applicable criterion: multiple peaks in a displacement distribution would diagnose forward persistence, independent of temporal correlations. The Fourier-space derivation is systematic and appears correct; the small-n formulas are explicit and checkable, and the Cauchy case is solved in closed form. The numerical exploration is useful and reproduces the expected qualitative dependence on step-size width and persistence. No fitted parameters are involved, and the model is simple enough that the main formulas could be rederived independently. However, the headline theorem as stated is not correct, and the abstract overstates what is proved. The paper's practical message can likely be repaired by restricting the statement, but the present version requires substantial revision of the central claim.","major_comments":[{"comment":"The proof that γf=0 implies at most one peak omits the n=0 term. In Eq. (30), p0(r) is a delta function at r=0, and q0(τ)>0 for standard temporal distributions such as the Poisson distribution in Eq. (46), where q0=e^{-τ}. With s(l)=δ(l-a) and small α, the total distribution Pr(r,τ) contains a central peak near r=0 from the n=0 term and a shell peak near r=a from the n=1 term, giving two local maxima even when γf=γb=0. The induction argument for pn with n≥2 therefore does not establish at most one peak of the full sum. The theorem needs an additional condition, such as conditioning on at least one hop (or requiring q0=0), or an explicit exclusion of the r=0 feature.","section":"Section IV, Eqs. (30), (51), and (46)"},{"comment":"The reduction to a Dirac delta step-size distribution is not valid for general s(l). Equation (30) gives p1(r)=s(r)/(Ω_d r^{d-1}), so if s(l) is multimodal, p1 already has several peaks at γf=0. The statement that any other step-size distribution only smears peaks presupposes that s(l) is single-peaked. As written, the paper claims the theorem for arbitrary step-length distributions, but the argument only covers single-peaked ones; the theorem and abstract must be restricted accordingly.","section":"Section IV, third paragraph"},{"comment":"The paper conflates non-monotonicity with the presence of multiple peaks. A distribution with a single peak away from r=0 is non-monotonic, and the n=0 counterexample above shows non-monotonicity can occur without any persistence when both p0 and p1 contribute. What the (qualified) proof can support is, at most, a statement about the absence of multiple peaks under restricted conditions. The abstract's claim that 'non-monotonicity can occur only if the persistence is strong enough' is not established by the paper's own analysis and should be reformulated to match the actual theorem.","section":"Abstract and Section V"}],"minor_comments":[{"comment":"In the expression for p1(r), the argument of the delta function should be r-a rather than l-a; as written the notation is inconsistent with the left-hand side depending on r.","section":"Eq. (51)"},{"comment":"The plots show Pr(r) on a log scale, but the text does not state clearly whether the plotted quantity is the vector density, the radial density, or some other marginal. This distinction is essential for interpreting 'peaks', especially for the delta-shell p1 contribution.","section":"Figs. 3-5"},{"comment":"Reference [29] cites Phys. Rev. Lett. volume 188, which does not exist; the volume should likely be 118. Reference [47] cites Phys. Rev. Lett. 36, 823 (1930); this should be Phys. Rev. 36, 823.","section":"References"},{"comment":"The phrase 'It is a representative of general distributions with a single peak' should be explicitly connected to the restriction needed in Section IV; otherwise the reader may assume the theorem covers the Cauchy distribution and other heavy-tailed or multimodal step sizes.","section":"Section II.C"}],"recommendation":"major_revision","confidential_remarks":"The paper contains a solid derivation of the displacement distribution for a useful toy model, but the central theorem as stated is false because of the n=0 contribution and the unrestricted step-size distribution. I believe this is fixable within the manuscript's scope by adding the appropriate hypotheses and aligning the abstract and summary with the corrected statement. Once the claims are restricted, the paper could be publishable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear [Name],\n\nThe short version: the exact displacement PDF for a persistent random walk with arbitrary temporal correlations is a real, useful result. The headline claim—that multiple peaks in the displacement distribution imply positive persistence—is not true as stated, and the proof has a concrete hole even within the authors' own model.\n\nWhat is genuinely new: the operator recursion leading to Eq. (28) is systematic and checkable, the small-n expressions are correct in the cases I worked by hand, and the explicit evaluation for delta, modified Gaussian, and Cauchy step distributions with Gaussian noise is careful. The idea that the van Hove function's peak structure can be used to infer persistence is appealing and, under the right conditions, likely correct.\n\nThe problems are two. First, the proof of the no-multiple-peaks theorem in Section IV only treats p1 and p_n≥2; it never treats p0, the walkers that perform zero hops. With q0(τ)>0, as in their Poisson, binomial, and geometric models, the continuous Gaussian process turns p0 into a peak at r=0. With delta steps and small α, the q1 term is a shell peaked near r=a. So for γf=0 you get two local maxima, e.g., Poisson q0=e^{-τ}, q1=τe^{-τ}, a=1, α=0.1. The abstract's claim is not an overstatement; it is simply false as written. The second gap is the reduction to delta step sizes: smearing can only remove peaks if the step-length distribution is unimodal. If s(ℓ) has several preferred lengths, p1 already has several peaks at γf=0. That assumption is never stated.\n\nThere is also a semantic mismatch: the abstract says 'non-monotonicity', while the theorem actually concerns 'more than one peak'. A single peak at r=a already makes Pr non-monotonic, and that can happen at γf=0 (even with p0 ignored).\n\nNone of this kills the exact expression, which is worth having. But the necessary-condition theorem needs a qualified restatement: for a single-peaked step distribution and conditioning on at least one hop, at most one peak away from r=0 exists at γf=0. I would send this to a referee, because the main calculation is genuinely useful and the error is fixable with major revision. I would not cite it in its current form.\n\nBest,","headline":"Genuine exact calculation, but the central 'peaks imply persistence' theorem is false as stated because the proof ignores the n=0 term and assumes unimodal step lengths.","tokens_in":17350,"tokens_out":6392,"would_cite":false,"duration_ms":63485,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["05.40.Fb"],"model":"deepseek-v4-flash","headline":"The paper proves that in a homogeneous system a short-time displacement distribution with more than one peak exists only when walkers have positive forward persistence, and derives the exact distribution for arbitrary temporal correlations.","keywords":["active random walk","persistent random walk","displacement distribution","van Hove distribution","non-monotonicity","directional correlations","temporal correlations","run-and-tumble motion"],"falsifier":"Simulate the model with $\\gamma_f=0$, $\\gamma_b=0$, and a two-peaked step length $s(\\ell)=\\frac{1}{2}[\\delta(\\ell-a)+\\delta(\\ell-2a)]$ in a homogeneous three-dimensional system; if $P_r(r,\\tau)$ has two local maxima for some small noise width $\\alpha$, the theorem in its general form is false.","tokens_in":16461,"feed_emoji":"🧭","tokens_out":8652,"duration_ms":86451,"temperature":0.7,"pith_summary":"Active particles in a homogeneous fluid can still show a non-monotonic displacement distribution, a short-time probability curve with several humps, and this paper pins down exactly what causes those humps. For a minimal persistent random-walk model with arbitrary step timing, the authors prove that more than one peak can appear only if the walker has a positive tendency to repeat its previous step in the same direction. Temporal correlations between hops, by contrast, change only the size, shape, and decay of the peaks, not whether they exist. The practical payoff is a diagnostic: peaks in a measured short-time displacement distribution reveal directional persistence, while their spacing and heights report the step length and the timing statistics.","feed_headline":"Multiple displacement peaks reveal a walker's directional memory","feed_subtitle":"Peaks appear only with forward persistence, so experiments can read direction memory from short-time displacements","key_machinery":"The load-bearing object is a three-channel persistent random-walk model: at each hop the walker either repeats its previous hop exactly (forward, $\\gamma_f$), retraces it (backward, $\\gamma_b$), or takes a fresh step whose length comes from $s(\\ell)$ in a random direction ($\\gamma_0=1-\\gamma_f-\\gamma_b$). The derivation writes the Fourier transform $\\tilde p_n(k)$ of the $n$-hop displacement as a matrix product of operators $M_0,M_f,M_b$ acting on the first-hop distribution, which yields exact closed forms for small $n$ and an exact convolution formula for the total PDF with a Gaussian thermal bath. For the no-peaks theorem the essential tool is an integral recursion, $p_{n+1}(r)=\\frac{1}{2a}\\int_{|r-a|}^{r+a}\\frac{r'}{r}p_n(r')\\,dr'$ for fixed step length $a$, which lets the authors prove by induction that every $p_n(r)$ with $\\gamma_f=0$ is non-increasing in $r$; Gaussian noise can only smooth further, so the full $P_r$ has at most one peak.","core_discovery":"On the paper's own terms, the central claim is a no-go result with a positive counterpart. For an isotropic active random walk whose hops are either forward repeats of the previous hop (probability $\\gamma_f$), exact reversals (probability $\\gamma_b$), or fresh random hops of length drawn from $s(\\ell)$, the displacement probability density $P_r(r,\\tau)$ after a short interval $\\tau$ has at most one local maximum whenever $\\gamma_f=0$, regardless of the timing correlations $q_n(\\tau)$ and regardless of the step-length distribution. Therefore any displacement distribution with more than one peak implies $\\gamma_f>0$, i.e. positive directional correlations. The same calculation gives an exact Fourier-space expression for $P_r$ for arbitrary $q_n$ and $s(\\ell)$, and the numerical examples show that peak spacing is set by the most likely hop length, while the decay of peak heights with $r$ distinguishes Poissonian from correlated timing.","pith_inferences":["A testable extension the paper leaves implicit is computing the minimal persistence $\\gamma_f$ needed for a first side peak as a function of noise width $\\alpha$ and step-length variance, which would turn the qualitative 'strong enough' condition into a quantitative threshold.","Because backward moves are shown to be equivalent to reweighting the temporal correlations, the no-peaks theorem implies that purely anti-persistent walkers ($\\gamma_b>0$, $\\gamma_f=0$) also cannot produce multiple peaks, so experimental anti-persistence should show a monotone displacement curve.","If the step-length distribution itself has several preferred lengths, the 'peaks imply persistence' slogan could fail; an experimenter using this diagnostic should first check that the single-step length distribution is unimodal.","The result suggests a two-point assay for living cells: measure the short-time displacement distribution in a homogeneous environment; if it is non-monotonic, the underlying motion has positive velocity autocorrelation even when long-time trajectories look diffusive."],"forward_implications":["In a homogeneous active suspension, detecting two or more peaks in the short-time displacement distribution is direct evidence that the particles persist forward, and the short-time data can therefore be used to infer bounds on orientational correlations.","The spacing between peaks gives the most likely single-step length, while the relative heights of successive peaks give information about temporal correlations: exponential height decay signals Poissonian timing, and faster or slower decay signals negative or positive timing correlations.","Temporal correlations alone cannot produce non-monotonicity in a homogeneous system; they only set how pronounced the peaks are.","For heavy-tailed step lengths such as a Cauchy distribution the displacement PDF stays monotonic regardless of persistence, so the non-monotonicity diagnostic must be applied to systems with a well-defined single preferred hop length.","The exact formula for $P_r$ applies to any temporal correlation pattern $q_n$ and any step-length distribution in three dimensions, with analogous formulas in one and two dimensions."],"supporting_citations":[{"why":"Supplies the persistent random walk model with forward and backward step repeats that the paper generalizes to arbitrary step lengths and timing.","marker":"[46]"},{"why":"Gives the Gaussian displacement distribution for simple random walks, the monotone baseline against which non-monotonicity is defined.","marker":"[51]"},{"why":"Reports non-monotonic displacement PDFs for active particles in harmonic traps, the confined setting that motivates asking when such peaks appear without confinement.","marker":"[58]"},{"why":"Shows an aging Levy walk with non-monotonic displacement, a temporal-correlation mechanism the paper contrasts with persistence-driven peaks.","marker":"[60]"},{"why":"Shows peaks in the velocity distribution of active biological matter, motivating the search for equivalent signatures in displacement data.","marker":"[61]"},{"why":"Provides the experimental run-and-tumble observations that motivate the correlated active random walk model.","marker":"[31]"}],"fun_headline_variants":["Displacement peaks signal a walker's forward persistence","Short-time walker displacements reveal directional memory","Active walker's peaks tied to forward hop persistence","Multiple peaks arise from forward directional bias"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof's load-bearing premise is that a fresh random hop has a single preferred length; if a walker instead hops with several distinct favored lengths, peaks could appear from the hop-length distribution alone even with no forward persistence, which would break the paper's broad claim.","fun_headline_variants_meta":{"raw":{"variants":["Displacement peaks signal a walker's forward persistence","Short-time walker displacements reveal directional memory","Active walker's peaks tied to forward hop persistence","Multiple peaks arise from forward directional bias"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000594,"raw_usage":{"total_tokens":2721,"prompt_tokens":825,"completion_tokens":1896,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":441,"completion_tokens_details":{"reasoning_tokens":1838}},"tokens_in":441,"tokens_out":1896,"duration_ms":14798,"temperature":1.0,"reasoning_tokens":1838,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:24:16.129626+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate the model with $\\gamma_f=0$, $\\gamma_b=0$, and a two-peaked step length $s(\\ell)=\\frac{1}{2}[\\delta(\\ell-a)+\\delta(\\ell-2a)]$ in a homogeneous three-dimensional system; if $P_r(r,\\tau)$ has two local maxima for some small noise width $\\alpha$, the theorem in its general form is false.","supporting_citations":[{"cited_title":"Pouget, E","cited_arxiv_id":null,"evidence_quote":"Supplies the persistent random walk model with forward and backward step repeats that the paper generalizes to arbitrary step lengths and timing."},{"cited_title":"Montroll and G","cited_arxiv_id":null,"evidence_quote":"Gives the Gaussian displacement distribution for simple random walks, the monotone baseline against which non-monotonicity is defined."},{"cited_title":"Sposini, A","cited_arxiv_id":null,"evidence_quote":"Reports non-monotonic displacement PDFs for active particles in harmonic traps, the confined setting that motivates asking when such peaks appear without confinement."},{"cited_title":"Gnacik, A","cited_arxiv_id":null,"evidence_quote":"Shows an aging Levy walk with non-monotonic displacement, a temporal-correlation mechanism the paper contrasts with persistence-driven peaks."},{"cited_title":"Ben-Isaac, E","cited_arxiv_id":null,"evidence_quote":"Shows peaks in the velocity distribution of active biological matter, motivating the search for equivalent signatures in displacement data."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the experimental run-and-tumble observations that motivate the correlated active random walk model."}],"review_version":1}