{"id":"8b7d9940-2b38-4e3e-a1e1-2e1f0bd6d6b3","arxiv_id":"2508.10403","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"A colored sandpile with axis-locked particle motion is non-abelian, has exact mean avalanche size (L+1)/8, and shows scaling exponents consistent with a new universality class.","lead":"A new sandpile variant is introduced where grains of different colors move only along fixed lattice axes, making the dynamics non-abelian. Simulations show spatially structured steady states and avalanche exponents distinct from earlier sandpile models.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Non-abelian update protocol is not specified for the avalanche-size data; the claimed τ=1.5 and τ_T=1.84 universality class may be an artifact of one arbitrary toppling order.","rationale":"I evaluated the exact mean avalanche size first, because a flaw there would be fatal. Eq. (1) survives scrutiny: define D as the sum over all particles of their remaining distance to the boundary along their assigned color axis. Every particle jump (internal or exiting) decreases D by exactly 1, and adding a particle of random color at a random site increases D by the average distance (L+1)/2. In the stationary state E[ΔD]=0, so E[total jumps]=E[d]=(L+1)/2; each toppling produces 4 jumps, hence E[s]=(L+1)/8. So the exact result is not the weak point.\n\nThe real soft spot is the non-abelian protocol ambiguity. The paper defines the model with FIFO emission and fixed color trajectories, but the avalanche outcome depends on the order in which unstable sites are toppled, as the paper itself says. No unique update rule is specified for the size-distribution measurements, and no test of alternative protocols is reported. Since the exponents are extracted from three system sizes by visual data collapse, even the claimed numerical values are not quantitatively supported. The reader's weakest assumption identifies exactly this. The correct response is to keep the CONDITIONAL verdict: the exact mean size and density profiles give partial support, but the universality-class claim requires either a specified protocol with demonstrated robustness or a quantitative collapse analysis. This does not move the reader's verdict.","tokens_in":8228,"tokens_out":17428,"duration_ms":194520,"concrete_test":"Run the same colored-sandpile rules on L=4096 and L=16384 under at least three protocols: (i) random-sequential update (at each step choose one unstable site uniformly and topple); (ii) parallel update with a fresh random permutation per time step and FIFO insertion at receiving sites; (iii) parallel update with LIFO insertion or with a fixed site-index toppling order. For each protocol, measure D(s,L) and D(T,L), and estimate β, α, τ, τ_T using a quantitative collapse metric (e.g., minimize the integrated squared deviation between scaled curves for the three L values). If the inferred exponents differ by more than ~0.05, or if the collapsed curves do not overlap across protocols, the claimed universality class is protocol-dependent and the central claim is not established.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The colored sandpile is explicitly non-abelian: the paper states that \"the sequence in which different unstable sites have been updated\" affects the color distribution in the stable state. Yet the main-text definition never specifies the toppling order used to measure D(s,L) in Fig. 3. The only protocol mention is in Supp03, for lifetimes: \"all the unstable sites are ordered in a random sequence and then simultaneously toppled,\" which is itself ambiguous (random then simultaneous?) and is not stated to apply to the size-distribution runs.\n\nBecause the dynamics are non-abelian, different update rules—random-sequential vs. parallel with a random permutation, FIFO vs. LIFO queue insertion, fixed vs. randomized site order within a sweep—can lead to different stationary measures and different avalanche statistics. The exact result ⟨s(L)⟩=(L+1)/8 is protocol-independent: a potential argument (each particle jump decreases the sum of distances-to-exit by 1, and additions increase it by the random drop distance) gives E[s]=E[d]/4=(L+1)/8. Thus Fig. 2 cannot constrain the exponents. The universality-class claim rests entirely on the scaling collapse of D(s,L) and D(T,L), which are computed under one unspecified protocol and judged by eye with only three system sizes. If the exponents shift under an alternative protocol, the model does not define a single universality class. This is the load-bearing weakness, and it is acknowledged in the paper's own admission of order-dependence.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper introduces a 'colored sandpile' on the square lattice in which each grain has one of four colors fixing its direction of motion along a lattice axis. Topplings transfer n_c grains from the bottom of an unstable column to neighboring sites according to their color. The model is explicitly non-abelian. The author derives and numerically verifies the exact stationary mean avalanche size ⟨s(L)⟩=(L+1)/8 (Eq. 1), measures avalanche-size and lifetime distributions for system sizes up to L=16384, and claims a new universality class with exponents τ=1.5 and τ_T=1.84 in two dimensions. One-dimensional and flush-all variants, steady-state density profiles, and density finite-size corrections are also reported.","tokens_in":8695,"tokens_out":9269,"duration_ms":106335,"significance":"If the exponent claims are robust, the model is a genuinely new non-abelian sandpile universality class with a nontrivial steady-state structure. The paper has a notable strength: the exact mean avalanche size follows from a simple conservation/current argument and is verified numerically over a wide range of L (Fig. 2). The explicit identification of the order-dependence of the stable state is also useful. However, the universality-class claim rests on scaling collapses that are judged visually, and the avalanche dynamics are not fully specified because of the non-abelian update issue. The significance is therefore conditional on a more rigorous definition and quantitative collapse analysis.","major_comments":[{"comment":"The model is non-abelian (the paragraph beginning 'Since the sand particles are colored...' says so explicitly), so avalanche statistics are not well defined until the order in which unstable sites are updated is specified. The main text never states the update rule used to measure D(s,L) in Fig. 3(a,b). The only protocol statement, in Supp03, concerns lifetimes and says unstable sites are 'ordered in a random sequence and then simultaneously toppled', which is ambiguous and is not stated to apply to the size data. Different schedules (random-sequential, FIFO, LIFO, parallel with a random permutation) can produce different stationary color configurations and different avalanche histories. The exact result Eq. (1) is schedule-independent and cannot be used to fix the exponents. Please give a complete algorithmic specification for every simulation and show that the reported τ and τ_T are s","section":"Definition of the model; Fig. 3; Supp03"},{"comment":"The finite-size scaling collapses are assessed visually from three system sizes (L=1024, 4096, 16384 for sizes and a similar set for lifetimes). No error bars, no quantitative collapse criterion, and no range of α, β consistent with the data are provided. The statement that τ_T=2 is 'ruled out' is based on a visual comparison. Since the claim of a new universality class rests entirely on these collapses, the paper needs a quantitative analysis (e.g., moments, a collapse error functional, or a goodness-of-fit test) and a statement of statistical uncertainty.","section":"Fig. 3(a,b); Supp04"}],"minor_comments":[{"comment":"The derivation of Eq. (1) is compressed to one sentence. A formal potential/current argument (each toppling moves n_c grains one step; in stationarity the average total distance traveled equals the average distance to exit) would make the exact result fully transparent.","section":"Eq. (1)"},{"comment":"The paper repeatedly refers to Supp01–Supp08. If these are separate files, they were not part of the manuscript under review; at minimum the algorithmic details in Supp03 and the collapse analysis in Supp04 need to be in the main text or included in the submission.","section":"Supplements"},{"comment":"The text says the f_k values are 'extrapolated using Eqn. 1'; this should be Eqn. (5). The hand-tuned 1/ν values in Table I have no uncertainties or fitting criterion; if these are descriptive fits, say so.","section":"Table I"},{"comment":"The text says ρ1(x,L) is plotted against x, but the axis label is x/L. Please clarify. For L=128 and 256 to overlap, the abscissa must be x/L.","section":"Fig. 5(b)"},{"comment":"The flush-all version has two different scaling branches (flat small-s region scaled with β=1.105, α=1 and large-s region with β=2.55, α=1.7). Please state whether this is a crossover and define the crossover scale, otherwise the effective τ=1.5 for large sizes is not fully characterized.","section":"Flush-all version; Supp08"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"I read the colored sandpile paper. Short version: it introduces a genuinely new variant, gets one exact result, and makes a universality-class claim that is currently not well supported because the toppling order is underspecified.\n\nWhat's new: the model itself — grains with fixed colors moving along assigned lattice axes, with FIFO toppling — is not in the earlier two-component sandpile papers, and it's non-abelian in an interesting way. The exact mean avalanche size ⟨s(L)⟩=(L+1)/8 is clean and convincing; it follows from a simple conservation argument and the numerics in Fig. 2 are tight. The density profiles with color gradients are also a nice observation, and the paper is honest about the deviations from the simple ratio prediction.\n\nThe soft spots are real. The most important one is protocol dependence. The model is non-abelian: the paper says the sequence of updates matters. But the main text never says what toppling order was used to measure D(s,L) in Fig. 3. The only protocol mention is in Supp03 for lifetimes — 'ordered in a random sequence and then simultaneously toppled' — which is itself ambiguous and not stated to apply to the size data. Different orders (random sequential, parallel with random permutation, FIFO vs LIFO) can give different stationary measures and different avalanche statistics in non-abelian systems. Since ⟨s(L)⟩ is protocol-independent, Fig. 2 cannot constrain the exponents. So the claimed τ=1.5 and τ_T=1.84 rest entirely on by-eye collapses of three system sizes under one unspecified protocol. The stress-test note is right about this. The paper should either specify the protocol and show the exponents are stable across reasonable alternatives, or drop the universality-class claim to 'these exponents under this rule.'\n\nThat said, the paper is not incoherent. The model is well-defined in principle, the exact result is solid, and the simulations are plausible. The hand-tuned 1/ν values and lack of error bars are weaker but secondary. This is a useful addition to the SOC sandpile zoo, and the community would care about it even without the universality-class framing.\n\nMy recommendation: send it to peer review. It deserves referee time. But the review needs to insist on a precise toppling order and protocol robustness tests, or a clear argument that exponents don't depend on it. Without that, the main claim is not established.","headline":"New non-abelian sandpile with one exact result; the universality-class claim is undermined by an underspecified toppling order.","tokens_in":9077,"tokens_out":2619,"would_cite":false,"duration_ms":27306,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["05.65.+b"],"model":"deepseek-v4-flash","headline":"The colored sandpile, in which each color moves along one axis only, is non-abelian and has avalanche exponents τ=1.5 in two dimensions, distinct from earlier sandpile classes.","keywords":["self-organized criticality","sandpile model","colored grains","non-abelian dynamics","universality class","avalanche size distribution","density profile","directed motion"],"falsifier":"Run the colored sandpile on an $L=4096$ square lattice under a fixed deterministic toppling order, for example always toppling the lowest-index unstable site first, and measure $D(s,L)$. If the same collapse with $\\beta=3$, $\\alpha=2$, and $\\tau\\approx1.5$ does not reproduce, or if $\\langle s(L)\\rangle$ deviates from $(L+1)/8$ in the stationary state, the claim that this is a well-defined universality class independent of update details fails.","tokens_in":8191,"feed_emoji":"🏜️","tokens_out":8844,"duration_ms":90665,"temperature":0.7,"pith_summary":"The paper introduces a sandpile variant in which every grain is painted with one of a few colors and each color is permanently locked to one lattice axis. Because the order in which neighboring unstable sites topple changes the color sequence of a receiving column, the dynamics is non-abelian, unlike the standard abelian sandpile. The paper claims that this model reaches a steady state with a non-trivial spatial density profile and an avalanche size distribution $D(s) \\sim s^{-\\tau}$ with $\\tau=1.5$ in two dimensions, and that the average avalanche size is exactly $\\langle s(L)\\rangle=(L+1)/8$. These exponents differ from previously studied sandpile universality classes, so the paper argues that the colored sandpile belongs to a new universality class. The model is intended as a starting point for describing granular heaps with grains of different colors or properties.","feed_headline":"Colored sandpile avalanches scale with τ = 1.5, a new class","feed_subtitle":"Each grain color is locked to one axis, so toppling order matters and avalanche statistics leave classic sandpile classes","key_machinery":"The central mechanism is the color-axis locking rule: a grain of color $\\kappa$ is permanently assigned the unit lattice step $\\hat{e}_\\kappa$, and a toppling removes the bottom $n_c$ grains of an unstable column and sends each one step along its own axis, using a first-in-first-out sequence within the column. This makes toppling order relevant because a site receiving grains from two unstable neighbors inherits a color sequence that depends on which neighbor topples first; hence the dynamics is non-abelian. The exact result $\\langle s(L)\\rangle=(L+1)/8$ follows from the current identity $j_\\kappa(x+\\hat{e}_\\kappa,y)=j_\\kappa(x,y)+1/L^2$, which integrates to a linear per-color current profil","core_discovery":"The central claim is that making sand particles distinguishable by color and locking each color's motion to one lattice axis changes the universality class of self-organized criticality. For the square lattice with four colors and threshold $n_c=4$, an avalanche spreads isotropically even though each grain moves along a straight line, because the colors are uniformly mixed. The average avalanche size is derived from a conserved-current argument as $\\langle s(L)\\rangle=(L+1)/8$ and confirmed numerically to within $0.00002$. The size distribution satisfies the scaling form $D(s,L)L^3 \\sim G(s/L^2)$, giving $D(s)\\sim s^{-3/2}$; the lifetime distribution satisfies $D(T,L)L^{1.84} \\sim G(T/L)$, g","pith_inferences":["The non-abelian property means the fixed synchronous update order is not neutral: a deterministic sequential update, such as always toppling the lowest-index unstable site first, could change the stationary color mixture and possibly the exponents. Comparing protocols would separate robust universality from rule-dependent behavior.","The per-color current identity $j_\\kappa(x+\\hat{e}_\\kappa,y)=j_\\kappa(x,y)+1/L^2$ holds for any stationary state, so the exact linear current profile and $\\langle s(L)\\rangle=(L+1)/8$ should survive alternative updates; the fragile statements are the scaling exponents, not the mean.","The color-axis locking makes each grain perform straight-line motion, reminiscent of ballistic or directed-walk dynamics. The model may connect to directed percolation or to multi-species branching processes, yielding analytic predictions for $\\tau=3/2$ that the paper does not attempt.","In a real granular heap, color could stand for grain size, shape, or friction coefficient. The predicted density gradients near boundaries suggest measurable segregation in mixtures with direction-biased motion, an experimental extension the paper does not develop."],"forward_implications":["If the two-dimensional exponents $\\beta=3$, $\\alpha=2$, $\\tau=3/2$ hold in the infinite-size limit, the colored sandpile constitutes a new universality class of self-organized criticality.","The exact relation $\\langle s(L)\\rangle=(L+1)/8$ is established for the stationary state and provides a quantitative benchmark against which any future theory of this model can be tested.","Because toppling order changes color sequences at receiving sites, the update rule is part of the model's definition; different update rules may produce different stationary states or avalanche statistics.","The linear per-color density gradient and the non-uniform total density profile imply a color-segregation mechanism in the steady state, which may be relevant to granular heaps of mixed grains."],"supporting_citations":[{"why":"Introduces the original sandpile model and self-organized criticality; the colored model changes grain distinguishability and motion while keeping the same toppling-triggering framework.","marker":"[1]"},{"why":"Defines the abelian property of sandpile automata; the colored model's order-dependent toppling is introduced in explicit contrast to this.","marker":"[3]"},{"why":"Presents a stochastic two-state sandpile with a flush-all toppling rule, which the paper adapts as the 'flush all' version of the colored model.","marker":"[2]"},{"why":"Analyzes multiple topplings and compact avalanche clusters in the standard sandpile, used here to show colored avalanches are non-compact and differently structured.","marker":"[19]"},{"why":"Provides an exactly solved directed sandpile model, used as the reference for directed particle motion that nevertheless yields no preferred avalanche direction here.","marker":"[20]"},{"why":"Gives theoretical results for sandpiles with multiple topplings and universality classes, providing a baseline for the new exponents.","marker":"[21]"},{"why":"Studies a two-component abelian sandpile, the closest earlier multi-species model against which the non-abelian colored behavior is compared.","marker":"[15]"}],"fun_headline_variants":["Colored grains locked to axes shift sandpile universality class","Paint sandpile grains, get new critical scaling: τ=3/2","Color-coded motion changes sandpile avalanche statistics","Sandpile with colored grains: a distinct universality class","When grains have colors, sandpile avalanches take a new form"],"cache_read_input_tokens":2816,"weakest_assumption_plain":"The critical exponents are estimated from finite-size scaling collapses of one update protocol, random-order synchronous toppling, with collapses judged by eye; if a different toppling order changes the stationary state or the scaling, the claimed distinct universality class with $\\tau=1.5$ and $\\tau_T=1.84$ is not established.","fun_headline_variants_meta":{"raw":{"variants":["Colored grains locked to axes shift sandpile universality class","Paint sandpile grains, get new critical scaling: τ=3/2","Color-coded motion changes sandpile avalanche statistics","Sandpile with colored grains: a distinct universality class","When grains have colors, sandpile avalanches take a new form"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000231,"raw_usage":{"total_tokens":1307,"prompt_tokens":712,"completion_tokens":595,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":456,"completion_tokens_details":{"reasoning_tokens":523}},"tokens_in":456,"tokens_out":595,"duration_ms":6628,"temperature":1.0,"reasoning_tokens":523,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T20:27:26.050467+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the colored sandpile on an $L=4096$ square lattice under a fixed deterministic toppling order, for example always toppling the lowest-index unstable site first, and measure $D(s,L)$. If the same collapse with $\\beta=3$, $\\alpha=2$, and $\\tau\\approx1.5$ does not reproduce, or if $\\langle s(L)\\rangle$ deviates from $(L+1)/8$ in the stationary state, the claim that this is a well-defined universality class independent of update details fails.","supporting_citations":[{"cited_title":"Grassberger and S","cited_arxiv_id":null,"evidence_quote":"Analyzes multiple topplings and compact avalanche clusters in the standard sandpile, used here to show colored avalanches are non-compact and differently structured."},{"cited_title":"Dhar and R","cited_arxiv_id":null,"evidence_quote":"Provides an exactly solved directed sandpile model, used as the reference for directed particle motion that nevertheless yields no preferred avalanche direction here."},{"cited_title":"Paczuski and K","cited_arxiv_id":null,"evidence_quote":"Gives theoretical results for sandpiles with multiple topplings and universality classes, providing a baseline for the new exponents."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Studies a two-component abelian sandpile, the closest earlier multi-species model against which the non-abelian colored behavior is compared."}],"review_version":1}