{"id":"cd78b492-8d88-43cb-b835-3793721b4c46","arxiv_id":"2501.02980","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"high","formal_verification":"none","parameter_count":6,"one_line_summary":"Multispecies traffic is claimed to belong to the directed percolation universality class, but the power-law fit is made to match the target exponent, so the confirmation is circular.","lead":"This paper claims that mixed car-and-motorcycle traffic behaves like water flowing through a porous medium, a phenomenon called directed percolation, based on drone data and a stochastic simulation model. The authors say the finding could unify traffic jams with other nonequilibrium systems and explain why mixed-traffic data is so scattered.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The DP confirmation is circular: Tc is chosen by minimizing |ν − ν⊥|, so the reported agreement ν ≈ ν⊥ is an artifact of the selection rule and provides no independent evidence for the DP universality class.","rationale":"I have read the paper in good faith. The authors build a sophisticated, data-calibrated simulation and report a number of interesting empirical findings, including a heavy-tailed time-to-collision distribution and a fundamental diagram that matches field data. These are real contributions. However, the abstract's central claim—that multispecies traffic is a member of the 1D directed-percolation universality class—rests on the exponent fit in Figure 4 and Eq. (11). The reader's weakest_assumption is correct and I agree with it: the optimality condition Tc = argmin |ν − ν⊥| makes the agreement ν ≈ ν⊥ an artifact of the fitting procedure. The paper even states this selection rule explicitly, so there is no internal inconsistency, but the consequence is that the 'unambiguous confirmation' is not a prediction; it is a parametrization. I considered whether the absence of an absorbing state is an even more fundamental objection, since DP is defined for absorbing-state transitions and this model has no absorbing state. This is a serious conceptual issue, but the concrete quantitative failure is already present in the exponent fit, so I focus on that as the load-bearing concern. I would keep the reader's REJECT verdict: the central claim is not supported by the evidence as presented. The public availability of code and data makes a decisive re-analysis possible, so a revised version with an independent Tc determination and full data collapse could change the verdict, but that is not the current paper.","tokens_in":10433,"tokens_out":5187,"duration_ms":47890,"concrete_test":"Re-analyze the archived simulation data for the same density permutations with Tc determined independently of ν, for example by the zero crossing or inflection point of ⟨ΔΦ⟩ as a function of T (the 'sharp discontinuity' highlighted in the Fig. 4a caption). Fix Tc to that value, fit ν in Eq. (11), and compare to ν⊥. If the fitted ν deviates from ν⊥ by more than the reported σν = 1.4×10⁻⁵ for a substantial fraction of the densities, the circular Tc selection is the source of the agreement and the DP claim fails. Additionally, test data collapse using all three 1D DP exponents; absence of collapse would further confirm the concern. Because the code and data are publicly available (GitHub and open-traffic.epfl.ch), this re-analysis is feasible.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing evidence for the central claim is the statement in the Results, 'Multispecies traffic as DP' (paragraph after Eq. 11), that the fitted exponent ν matches the 1D DP value ν⊥ with average R² = 0.993. But the fitting protocol is designed to produce this match: Eq. (11) contains two unknowns, Tc and ν, and the authors impose the optimality condition Tc = argmin |ν − ν⊥|, where ν⊥ is the known DP spatial-correlation exponent. For each density permutation, the critical temperature is therefore selected to make the fitted exponent as close as possible to the DP value. The reported ν is thus not an independent measurement of a critical exponent; it is the result of an optimization whose target is ν⊥. The R² merely reports that, after choosing Tc in this way, a power law describes the curves; it does not test whether the system lies in the DP universality class. No data collapse with the full set of DP exponents (β, ν⊥, ν∥) is shown, and the authors explicitly defer the measurement of the other two exponents and the negative-⟨ΔΦ⟩ branch to future work. The empirical time-to-collision power law and fundamental-diagram validation are useful contributions, but they do not bear on the DP identification. Because the only quantitative test of DP is circular, the central claim is unsupported.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This manuscript claims, on the basis of the pNEUMA drone data and a stochastic two-lane simulation model, that heterogeneous multispecies traffic belongs to the 1D directed percolation (DP) universality class. The authors calibrate a nonlinear optimal-velocity model with an anticipatory steering module, validate the aggregate fundamental diagram against field measurements, and report a power-law divergence of a speed-difference order parameter with mean view range; the fitted exponent is claimed to match the 1D DP value ν⊥. The abstract and Results present this as an 'unambiguous' confirmation of the DP hypothesis.","tokens_in":10686,"tokens_out":4218,"duration_ms":41929,"significance":"If the DP claim were independently supported, this would be a notable contribution that links urban traffic heterogeneity to nonequilibrium universality and offers an explanation for observed scatter in mixed-traffic data. The paper has genuine strengths: it draws on a large field dataset, provides an open simulation framework, reports a new empirical time-to-collision power law, and validates flow-speed relations against measurements. However, the quantitative evidence for DP is compromised by a circular fitting procedure, and the additional exponents and scaling collapse needed to identify a universality class are absent. I therefore cannot recommend publication without a substantially different analysis.","major_comments":[{"comment":"The central test of the DP hypothesis is not independent. Equation (11) has two unknowns, Tc and ν, and the authors impose the optimality condition Tc = argmin |ν − ν⊥|, where ν⊥ is the theoretical 1D DP exponent. Because Tc is chosen for each density permutation to make the fitted ν as close as possible to ν⊥, the reported agreement ν ≈ ν⊥ is a built-in consequence of the selection rule, not an empirical confirmation. The average R² = 0.993 only shows that, conditional on this choice of Tc, a power law describes the curves; it does not test whether the system is in the DP universality class. An unbiased test would fit Eq. (11) with free Tc and ν and report the resulting exponent distribution, or fix ν = ν⊥ and test for data collapse; neither is provided.","section":"Multispecies traffic as DP, Eq. (11) and following paragraph"},{"comment":"Only one of the three DP exponents is measured, and no scaling collapse is shown. DP in 1D is characterized by the exponents β, ν⊥, and ν∥; the paper measures only ν (fitted as described above) and explicitly defers β, ν∥, and the negative-⟨ΔΦ⟩ branch to future work. A universality-class assignment requires at least the full set of exponents or a data collapse with the DP scaling form; a single exponent obtained from a circular fitting procedure is insufficient to support the abstract's claim that 'multispecies traffic is a member of ... directed percolation in one spatial dimension.'","section":"Multispecies traffic as DP, Eq. (11) and Discussion"},{"comment":"The identification of the mean view range T as a temperature-like control parameter and of the speed difference ΔΦ as the percolation order parameter is asserted rather than derived. No finite-size scaling analysis is reported despite the small system size (L = 90 m), no correlation-length scaling is shown, and the non-universal thresholds pc are not tested against any independent prediction. Without such tests, even a non-circular power-law fit would not by itself establish the existence of a genuine phase transition in the DP class.","section":"Multispecies traffic as DP, Eqs. (9)-(10) and Fig. 4"}],"minor_comments":[{"comment":"The word 'micovehicles' should be 'microvehicles'.","section":"Fig. 3d caption"},{"comment":"The word 'nonequilibirum' should be 'nonequilibrium'.","section":"Discussion, first paragraph"},{"comment":"The bracket structure in the exponential term is confusing; please rewrite with clear parentheses, e.g., exp[−λ v0^{-1}(s − s0)].","section":"Eq. (6)"},{"comment":"The target direction is written both as a0 and as α0; please use a single consistent symbol throughout.","section":"Eq. (4) and surrounding text"},{"comment":"'arch traveled' should be 'arc traveled'.","section":"Methods, Maneuver detection"},{"comment":"The power-law distribution q ∝ Δt^{−μ} is stated without reporting the estimated exponent μ or the statistical test used to support the power-law claim; please provide the estimate, its uncertainty, and the goodness-of-fit measure.","section":"Fig. 1b and main text"},{"comment":"The reported standard deviation σν = 1.4 × 10^{-5} for the fitted exponents seems implausibly small given the 256 independent runs and the bootstrapped error bars shown in Fig. 4a; the provenance of this value should be clarified.","section":"Fig. 4c and related text"}],"recommendation":"reject","confidential_remarks":"The paper has a valuable empirical component, but the DP identification is the central claim and it rests on a circular fit. This is not a presentation issue; it would require a new analysis. If the authors can provide an independent test—for example, a free two-parameter fit, a data collapse with all three DP exponents, or finite-size scaling—the work might be reconsidered as a new submission."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know this paper before the abstract sells it. There are two genuinely useful pieces here: a new empirical power-law for time-to-collision in mixed traffic (pNEUMA data, Fig. 1b), and a carefully calibrated simulation with heterogeneous optimal-velocity parameters, Ornstein-Uhlenbeck noise, and dynamic steering. The fundamental-diagram validation is credible, and the code and data are public. The nonlinear drop in car flow with motorcycle share, contradicting an earlier cellular-automaton study, is a nice finding in its own right. Credit where due: the empirical work and the modeling effort are real.\n\nThe soft spot is the central claim, and it is load-bearing. The paper says the 1D DP hypothesis is \"unambiguously confirmed\" with average R² = 0.993, but the test is circular. Equation (11) has two unknowns, Tc and ν, and the authors impose Tc = argmin |ν − ν⊥|, where ν⊥ is the known DP exponent. That means the fitted ν is selected to be as close as possible to ν⊥; the reported agreement is a consequence of the selection rule, not an independent measurement. The R² only says a power law fits after Tc is chosen that way. They also measure only one of three DP exponents, use a non-standard order parameter (a speed difference), discard the negative-⟨ΔΦ⟩ branch, and show no data collapse with the full set of exponents. The authors list these omissions as future work in the Discussion, so they are aware, but the abstract and results overstate the support. A further conceptual issue: standard DP is an absorbing-state transition, and their stochastic model has no absorbing state, so the link to DP universality is even more tenuous than the fitting issue alone.\n\nI agree with the reader's verdict. The paper deserves serious peer review because the empirical and modeling contributions are substantive and the DP claim is ambitious enough to warrant close scrutiny, but as it stands the central claim is not demonstrated. A proper test would estimate Tc from data collapse using all three DP exponents, or fix ν⊥ and measure β and ν∥ independently. If the authors did that, the paper could be much stronger. For now, treat the DP identification as an interesting but unsupported conjecture.\n\nWho benefits: traffic flow researchers, complexity scientists, and anyone working on micromobility policy. I would engage with it and would cite the time-to-collision power-law and the simulator, but not the DP universality claim. Send it to a serious referee, and the referee should ask for major revision or reframing.","headline":"A serious, well-documented modeling and empirical paper whose central DP claim rests on a circular fitting procedure: Tc is chosen to make the fitted exponent match the DP value, so the confirmation is not independent.","tokens_in":11281,"tokens_out":1900,"would_cite":true,"duration_ms":110256,"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":"Heterogeneous city traffic belongs to the directed-percolation universality class and, near its critical point, behaves like water seeping through a porous medium.","keywords":["directed percolation","multispecies traffic","nonequilibrium phase transition","micromobility","porous media flows","optimal velocity model","urban traffic","self-organized criticality"],"falsifier":"Run the same simulations and extract the remaining one-dimensional directed-percolation exponents, the order-parameter exponent $\\beta$ and the temporal-correlation exponent $\\nu_\\parallel$, together with a data collapse of $\\langle\\Delta\\Phi\\rangle_T^+$, without imposing $T_c = \\arg\\min|\\nu - \\nu_\\perp|$; if either exponent disagrees with the 1D DP values or the collapse fails, the universality claim is refuted.","tokens_in":10132,"feed_emoji":"🚦","tokens_out":14956,"duration_ms":134828,"temperature":0.7,"pith_summary":"This paper sets out to show that heterogeneous urban traffic, where cars share the road with scooters and motorcycles, is a member of directed percolation: the best-known universality class of nonequilibrium phase transitions, the same family as water seeping through a porous medium. Using a large drone-tracked field dataset and a stochastic agent-based model calibrated to it, the authors report that the transition between disordered and ordered multiclass flow follows a power law with the one-dimensional directed-percolation exponent, with an average fit quality of $R^2 = 0.993$. The claim matters because it would give urban traffic with micromobility a quantitative bridge to nonequilibrium statistical physics, and it would make the onset of microvehicle filtering through car traffic predictable from a universal exponent rather than from case-specific details.","feed_headline":"Mixed city traffic behaves like water in porous media","feed_subtitle":"Cars and microvehicles share the same phase-transition power law as porous flows.","key_machinery":"The load-bearing mechanism is a two-module agent-based model: a steering module with a speed-dependent dynamic view range $\\gamma_{\\max}(v) = \\exp(bv + c)$ that acts as a stochastic, temperature-like porosity or percolation probability, and a heterogeneous optimal-velocity module with a mean-reverting Gaussian noise term and an anticipation weight based on time to collision. The order parameter $\\Delta\\Phi = \\Phi_{\\mathrm{moto}} - \\Phi_{\\mathrm{car}}$ and the control parameter T (the mean view range) are used to extract the critical power law, with the critical temperature fixed by the condition $T_c = \\arg\\min|\\nu - \\nu_\\perp|$ so that the fitted exponent reports the theoretical 1D directed-percolation value.","core_discovery":"The central claim is that multispecies traffic is an instance of one-dimensional directed percolation. The authors define a temperature-like control parameter T as the mean dynamic view range of microvehicles and an order parameter ΔΦ = Φ_moto − Φ_car that measures whether microvehicles are on average faster than cars. Near a critical temperature Tc the ensemble-averaged order parameter vanishes, and on the supercritical side it follows $\\langle\\Delta\\Phi\\rangle_T^+ \\sim |T - T_c|^{\\nu}$ with $\\nu \\approx \\nu_\\perp$, the spatial-correlation exponent of 1D directed percolation. The paper reads this as macroscopic equivalence to porous flows: smaller vehicles percolate through lanes of cars much as water infiltrates a porous bed, and states that the hypothesis is unambiguously confirmed with an average $R^2 = 0.993$ across statistically significant density and motorcycle-count permutations.","pith_inferences":["Editorial inference: the reported agreement $\\nu \\approx \\nu_\\perp$ is steered by the fitting rule $T_c = \\arg\\min|\\nu - \\nu_\\perp|$ introduced right after Eq. 11 in the Results, so the exponent value alone is not independent evidence for directed percolation; the empirical content is the clean power-law form and the density-dependent thresholds.","Editorial inference: because the paper itself notes that all three DP exponents should be measured, a decisive test would extract $\\beta$ and $\\nu_\\parallel$ plus a data collapse from the same simulations without the Tc-optimization; failure there would falsify the universality assignment.","Editorial inference: if the universality is real, the same critical exponent should appear in other multimodal trajectory datasets beyond the one studied here, which is a testable prediction that does not require rerunning the simulator."],"forward_implications":["If the universality claim holds, mixed urban traffic near its phase transition has no characteristic length scale, and macroscopic flow statistics obey a universal power law independent of behavioral details and vehicle geometry.","The calibrated nonequilibrium model predicts a strongly nonlinear drop in car flow as the motorcycle share grows, contradicted only by an earlier unvalidated simulation study that predicted a linear reduction.","The heavy-tailed time-to-collision distribution observed in the field implies that equilibrium is a rare event, so equilibrium-based traffic theories cannot describe the phase transition; nonequilibrium, noisy, driven models are required.","The same modeling logic should apply to other heterogeneous crowds, such as pedestrians mixed with e-scooters, where universality could be tested outside the vehicle context."],"supporting_citations":[{"why":"It supplies the one-dimensional directed-percolation universality class and the theoretical spatial-correlation exponent $\\nu_\\perp$ used to anchor the fit.","marker":"[3]"},{"why":"It frames the earlier conjecture that heterogeneous traffic behaves as a porous flow, which this paper sets out to substantiate.","marker":"[1]"},{"why":"It provides the large-scale drone-trajectory field data from a congested city used for calibration and for documenting the power-law time-to-collision distribution.","marker":"[6]"},{"why":"It supplies the off-lattice, active-matter nonequilibrium phase-transition framework that the traffic model builds on.","marker":"[14]"},{"why":"It introduces the optimal-velocity model that the heterogeneous speed module extends with distributed parameters and noise.","marker":"[29]"},{"why":"It gives the exponential speed-spacing functional form adopted for the calibrated response curves.","marker":"[30]"},{"why":"It provides the anticipatory time-to-collision steering heuristic behind the dynamic view-range mechanism.","marker":"[18]"},{"why":"It underlies the mean-reverting Gaussian noise treatment used to model stochastic perturbations in the velocity dynamics.","marker":"[32]"},{"why":"It supplies empirical arterial fundamental diagrams used to confirm the simulated effective lane capacity of about 1600 vehicles per hour.","marker":"[34]"}],"fun_headline_variants":["Mixed traffic is universal: like water in porous media","Cars and microvehicles percolate like porous flow","Directed percolation explains mixed city traffic","Mixed traffic obeys water-percolation power law"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The identification of mixed traffic with directed percolation rests on a fitting rule that chooses the critical temperature to make the measured exponent match the theoretical value; without that rule, the data alone do not establish the universality class.","fun_headline_variants_meta":{"raw":{"variants":["Mixed traffic is universal: like water in porous media","Cars and microvehicles percolate like porous flow","Directed percolation explains mixed city traffic","Mixed traffic obeys water-percolation power law"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000486,"raw_usage":{"total_tokens":2375,"prompt_tokens":903,"completion_tokens":1472,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":519,"completion_tokens_details":{"reasoning_tokens":1418}},"tokens_in":519,"tokens_out":1472,"duration_ms":12166,"temperature":1.0,"reasoning_tokens":1418,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T21:59:16.259385+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same simulations and extract the remaining one-dimensional directed-percolation exponents, the order-parameter exponent $\\beta$ and the temporal-correlation exponent $\\nu_\\parallel$, together with a data collapse of $\\langle\\Delta\\Phi\\rangle_T^+$, without imposing $T_c = \\arg\\min|\\nu - \\nu_\\perp|$; if either exponent disagrees with the 1D DP values or the collapse fails, the universality claim is refuted.","supporting_citations":[{"cited_title":"Non-equilibrium critical phe- nomena and phase transitions into absorbing states","cited_arxiv_id":null,"evidence_quote":"It supplies the one-dimensional directed-percolation universality class and the theoretical spatial-correlation exponent $\\nu_\\perp$ used to anchor the fit."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It frames the earlier conjecture that heterogeneous traffic behaves as a porous flow, which this paper sets out to substantiate."},{"cited_title":"& Geroliminis, N","cited_arxiv_id":null,"evidence_quote":"It provides the large-scale drone-trajectory field data from a congested city used for calibration and for documenting the power-law time-to-collision distribution."},{"cited_title":"& Shochet, O","cited_arxiv_id":null,"evidence_quote":"It supplies the off-lattice, active-matter nonequilibrium phase-transition framework that the traffic model builds on."},{"cited_title":"& Sugiyama, Y","cited_arxiv_id":null,"evidence_quote":"It introduces the optimal-velocity model that the heterogeneous speed module extends with distributed parameters and noise."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It gives the exponential speed-spacing functional form adopted for the calibrated response curves."},{"cited_title":"& Theraulaz, G","cited_arxiv_id":null,"evidence_quote":"It provides the anticipatory time-to-collision steering heuristic behind the dynamic view-range mechanism."},{"cited_title":"& Schadschneider, A","cited_arxiv_id":null,"evidence_quote":"It underlies the mean-reverting Gaussian noise treatment used to model stochastic perturbations in the velocity dynamics."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It supplies empirical arterial fundamental diagrams used to confirm the simulated effective lane capacity of about 1600 vehicles per hour."}],"review_version":1}