{"id":"503a4482-621c-4e32-ae2c-64a91a1e97ca","arxiv_id":"2602.10782","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A new 'ghost particle' trick yields determinant formulas for exact coalescence probabilities in any nearest-neighbor Markov process, including Brownian motion.","lead":"This paper proves that exact probabilities for coalescing particle systems—where particles merge on collision—can be written as determinants, by introducing invisible 'ghost' particles that keep the particle count fixed. The result gives a closed formula for survivor positions that applies to random walks, birth-death chains, and Brownian motion.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 8.2 is false for arbitrary measurable A: on reversed survivor orderings the RHS is negative while the coalescing probability is zero.","rationale":"The reader's weakest assumption—the unproved continuous-time measure-preservation and uniqueness of successful castings in Theorem 7.2—is a real rigor gap and deserves attention. However, the most load-bearing issue is more direct: Theorem 8.2, the paper's central closed-form result, is false for arbitrary measurable A because the determinant is antisymmetric under swapping survivor coordinates while the coalescing-process support is the ordered sector. This gives an explicit counterexample with a negative RHS, not merely an unproved lemma. The defect is localized and repairable by adding 1_{y1<...<yk} to the density or restricting A, so the overall verdict remains CONDITIONAL rather than REJECT; the reader's CONDITIONAL judgment is preserved. I chose 'partial' agreement because I am not claiming the reader's continuous-time concern is wrong, just that the order-support flaw is a more immediately demonstrable failure of the central claim as stated.","tokens_in":25004,"tokens_out":32750,"duration_ms":328462,"concrete_test":"Take two standard Brownian motions with x1=0, x2=1, T=1, composition 1+1. Numerically or analytically compute ∫_{y1>y2} [p_0(y1)p_1(y2)-p_0(y2)p_1(y1)] dy1 dy2. The result is negative (it equals -P(no collision)), while the coalescing probability for the event {survivor positions (y1,y2): y1>y2} is exactly 0. Repeating with A restricted to y1<y2 gives the standard non-collision probability, confirming that the missing order indicator is the issue.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The coalescence determinant is stated as an identity between measures on all of S^k (Theorem 8.2). This fails as stated. For the no-coalescence pattern 1+1 (n=2, k=2), Definition 8.1 gives the Karlin–McGregor matrix [p_{x_i}(y_j)]. Take Brownian motion with x1<x2 and A={y1>y2}. The LHS is 0: if two survivors exist, they cannot have swapped order without having coalesced. But the RHS is ∫_{y1>y2} det[p_{x_i}(y_j)] dy = P(X1>X2) - P(X1<X2) = -P(no collision) < 0. Thus the formula produces a negative probability on a nonempty measurable set. The proof in Section 8 integrates out ghosts but never enforces the heir-order support y1<...<yk; the correct density is det(M~) · 1_{y1<...<yk}. Without this indicator (or an explicit restriction on A), the central theorem as stated asserts a measure identity that is literally false. This is a statement-level defect, independent of the (also sketchy) continuous-time measure-preservation argument.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a 'ghost particle' method for coalescing particle systems on a line. When two particles coalesce, one heir continues and one ghost random walk is created, preserving the total particle count. This restores a square matrix structure and allows the authors to state a determinant formula for the probability of specified coalescence patterns. The main results are Theorem 3.2, a discrete generating-function identity for coalescence performances on planar weighted DAGs; Theorem 7.2, a continuous-time measure identity under Karlin–McGregor assumptions; and Theorem 8.2, the 'coalescence determinant', which marginalizes ghosts and gives a closed-form determinant for survivor positions alone, intended to cover Brownian motion, birth-death chains, and lattice walks.","tokens_in":25324,"tokens_out":7728,"duration_ms":83246,"significance":"The ghost-particle idea is attractive and, if the measure-theoretic statements are corrected, would extend the Karlin–McGregor/LGV determinant formalism to coalescing systems at pattern-level resolution. The discrete proof in Sections 4–6 is self-contained and appears internally consistent: it constructs an explicit sign-reversing involution on castings, has no fitted parameters, and reduces to the classical LGV determinant in the no-coalescence case. The n=2 coalescence example checks out. However, the central continuous and ghost-free statements, as written, fail on non-ordered measurable sets, and the continuous-time proof is only a sketch. These issues are load-bearing for the paper's main claims, though they appear repairable within the manuscript's scope.","major_comments":[{"comment":"Theorem 8.2 asserts P_int(survivor positions ∈ A)=∫_A det(̃M) dν for every measurable A⊆S^k. This is false without an ordering restriction. For n=2, k=2, composition 1+1, ̃M=[p_{x_i}(y_j)]. Take x1<x2 and A={y1>y2}. In the coalescing system two survivors cannot have swapped order, so the left-hand side is 0. The right-hand side is ∫_{y1>y2} det[p_{x_i}(y_j)] dν = P(X1>X2)-P(X1<X2), which is strictly negative for Brownian motion. The marginalization proof in §8 integrates out ghosts but never enforces the chamber y1<...<yk. The theorem must restrict A to that ordered chamber, or include the indicator 1_{y_1<...<y_k} in the integrand. Since the paper explicitly calls det(̃M) a probability density/mass function, this is a statement-level defect, not a cosmetic one.","section":"Theorem 8.2 (§8.3)"},{"comment":"The same support issue appears in the continuous-time ghost formula. Definition 7.1 constrains ghosts relative to their heirs but does not constrain heir positions among themselves; when G=∅ the admissibility condition is vacuous. Taking A={y1>y2} in the two-particle no-coalescence case gives an admissible set for which the RHS of Theorem 7.2 is P(X1>X2)-P(X1<X2)<0 while the LHS is 0. Thus the claimed 'identity between measures' holds only on ordered survivor configurations. In addition, the proof sketch in §7.3 is substantially less complete than the discrete proof: the segment-swap map on the casting space must be shown to be a measurable, measure-preserving, sign-reversing involution, and the claim that each outcome with final positions in A has exactly one successful π needs a continuous analogue of Proposition 5.11 with the same tie-breaking conventions. The paper should either prov","section":"Theorems 7.1–7.2 (§7.2, §7.3)"}],"minor_comments":[{"comment":"The notation 'A⊆R R' in Definition 7.1 appears to be a typo; it should be a subset of the appropriate product state space, e.g. S^R or S^k.","section":"§7.2"},{"comment":"The abstract and introduction emphasize 'nearest-neighbor transitions', while Theorem 3.2 is stated for general planar weighted DAGs and §7 covers skip-free birth-death chains. The terminology should be harmonized (e.g., 'skip-free' or 'order-preserving').","section":"Abstract and §1.5"},{"comment":"The caption phrase 'heir column below step (i≥rank) above step (i<rank)' is confusing; the figure illustrates ghost columns, not heir columns, and the caption should be reworded to match the matrix display.","section":"Figure 2 caption"},{"comment":"The coefficient extraction notation [t^+], [t^-] is used in the introductory example before the formal definition in §3.1. Moving the formal definition earlier, or adding a forward reference, would improve readability.","section":"§1.3.2 vs §3.1"},{"comment":"For discrete state spaces, the matrix entries F_{x_i}(y_l)-[i<j] can be negative, so det(̃M) is a signed mass function on arbitrary sets. Please state explicitly that the formula is a signed measure identity and is a genuine probability mass/density only on the ordered chamber.","section":"§8.1"}],"recommendation":"major_revision","confidential_remarks":"The discrete determinant identity appears sound and is the paper's real contribution. The main revision should focus on the measure-theoretic statements: add the ordered-chamber restriction (or indicator) to Theorems 7.2 and 8.2, and either complete or precisely reference the continuous-time proof. I would not reject outright; the defects are substantial but repairable within the manuscript's scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a genuine piece of work, but the headline theorem overreaches. The stress-test note is correct: Theorem 8.2 asserts an identity between measures on all of S^k. For the no-coalescence composition 1+1 with Brownian motion and x1<x2, take A={y1>y2}. The left side is zero — two survivors cannot swap order without coalescing — while the right side is ∫_{y1>y2} det[p_{xi}(yj)] dy, which is P(X1>X2)-P(X1<X2) < 0. So the statement is literally false. The cause is visible in the proof: integrating out ghosts never enforces the survivor-order support. That is a one-line fix — restrict A to the chamber y1<...<yk or multiply det(˜M) by the corresponding indicator — but it has to be made. The reader's conditional verdict did not flag this; I think it is the main issue.\n\nWhat is genuinely good: the ghost-particle construction is a nice idea. It restores the square matrix and the LGV-style proof in Sections 4–6 is careful and elaborate. The general composition determinant (Theorem 3.2) appears to be new: the pairwise case is in Urbán's thesis and Warren's formula gives the unordered joint law, so arbitrary-pattern resolution is a real step. The staircase matrix structure and the sign identity are internally consistent, and the n=2 check reproduces known coalescence probabilities. The literature discussion is honest and puts the work in context without overselling.\n\nSoft spots, in proportion. (1) The support error above makes the central theorem false as stated — that is the biggest problem. (2) The continuous-time theorem (Theorem 7.2) is sketched, not proved: the measure-preservation of the segment-swap on the casting space is asserted via the strong Markov property, but the measure-theoretic details are not supplied. This may be repairable, but as written it is a real gap. (3) The recovery of Warren's formula is deferred to a companion paper, so part of the paper's own motivation is unverified here. None of these undermine the discrete proof, but they do mean the paper is not ready in its current form.\n\nWho it is for: people working on coalescing random walks, voter-model duality, and determinantal formulas. It deserves a serious referee, not a desk reject. But any referee should require the support correction and a complete continuous-time argument before publication. My recommendation: send it to review, with the expectation of major revision.","headline":"The ghost method is real and the discrete proof is serious, but Theorem 8.2 as stated is false on reversed-order configurations; that is a fixable support error, not a fatal flaw.","tokens_in":25730,"tokens_out":2999,"would_cite":false,"duration_ms":33853,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["05A15","05A19","15A15","60C05","60J65","82C22"],"pacs":[],"model":"deepseek-v4-flash","headline":"Coalescing particles on a line have exact determinant formulas for every prescribed collision pattern.","keywords":["coalescing random walks","ghost particles","determinantal formulas","lattice paths","birth-death chains","Brownian motion","voter model","interacting particle systems"],"falsifier":"Enumerate all coalescing path tuples for three simple random walks on Z starting at 0, 1, 2, with the 2+1 coalescence pattern, up to a fixed small time; compute the survivor-position probabilities by exhaustive counting and compare with the determinant formula evaluated at the same transitions. A single mismatch would disprove the discrete claim.","tokens_in":24949,"feed_emoji":"👻","tokens_out":6856,"duration_ms":71441,"temperature":0.7,"pith_summary":"The paper proves that coalescing particles on a line have exact, closed-form probability formulas for every specified collision pattern—which initial particles merge into which survivors. The trick is to add an invisible 'ghost' particle at each collision, keeping the total count fixed so that a square matrix can be built. The resulting coalescence determinant has transition probabilities in survivor columns and cumulative distributions with a staircase shift in ghost columns. If true, this gives pattern-level exact probabilities wherever the classical non-colliding determinant formula applies: lattice walks, birth-death chains, and continuous diffusions such as Brownian motion.","feed_headline":"Ghost particles turn any coalescence pattern into a determinant","feed_subtitle":"A closed-form probability for which particles merge, valid for random walks, birth-death chains, and Brownian motion.","key_machinery":"The ghost particle method: when two particles collide, one survivor continues as usual and one invisible ghost starts from the collision point as an independent non-interacting walk. This keeps the number of entities at n, so the transition matrix stays square. Ghost columns carry a staircase sign pattern (t+ below the row rank, -t- above), and coefficient extraction selects exactly the terms matching the prescribed collision pattern. Failed path assignments cancel pairwise by segment-swapping at the first spurious crossing, leaving only successful castings; integrating out ghost positions converts transition-probability columns into cumulative-distribution columns.","core_discovery":"The central result is the coalescence determinant: for a composition c1+...+ck=n, the probability that the k survivors land in a measurable set A equals ∫_A det(M̃(y1,...,yk)) dν. The matrix M̃ has transition densities or probabilities in the first column of each block and cumulative distributions F_{x_i}(y_l) - [i<j] in the remaining columns of the block. The same paper proves a finer ghost formula that keeps ghost positions as formal variables; marginalizing over ghost positions converts the transition-probability columns into cumulative columns and yields the coalescence determinant. The proof works by adding one invisible ghost per collision so the total particle count stays n, then canc","pith_inferences":["Beyond the paper, the same ghost bookkeeping could be applied to absorbing boundaries: replacing the transition kernel by a killed kernel in the staircase columns would give a coalescence determinant for finite intervals, a testable variant.","Beyond the paper, the measure identity suggests an exact sampling algorithm: draw survivor positions from the determinant density rather than simulating collisions, which could be considerably faster for large n.","Beyond the paper, the density-versus-cumulative column pattern may encode a general operation—'merge rows into survivor columns'—that could extend the formula to higher dimensions or to long-range coalescence wherever a planar cancellation exists.","Beyond the paper, the ghost coefficient-extraction formalism may transfer to other reactions, such as annihilation or branching, whenever a sign-reversing involution can be defined on path tuples."],"forward_implications":["Any specified coalescence pattern has a closed-form probability that can be evaluated by one determinant, with no simulation or recursion over collision orders.","The formula holds for all skip-free Markov processes satisfying planarity and the strong Markov property, so it applies to lattice walks, birth-death chains, and Brownian motion alike.","When no collisions occur, the ghost determinant reduces to the classical non-colliding determinant, making the formula a genuine extension of that theory.","Summing the pattern-level probabilities over all compositions recovers the previously known determinant for the joint distribution of coalescing Brownian survivors, but with finer information about which particles merged.","The measure form of the theorem, valid for both discrete and continuous state spaces, identifies the determinant as a Radon-Nikodym derivative ready for further integration or scaling limits."],"fun_headline_variants":["Ghost particles reduce coalescence to a determinant","Coalescence probabilities as a determinant","Ghost trick unlocks exact coalescence determinant","One ghost per collision yields a determinant","Determinants for coalescence from ghost particles"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The construction assumes particles on a line cannot pass each other without meeting, and that after two particles meet, swapping their future trajectories does not change the probabilities—if either assumption fails, the cancellation at the heart of the determinant identity breaks down.","fun_headline_variants_meta":{"raw":{"variants":["Ghost particles reduce coalescence to a determinant","Coalescence probabilities as a determinant","Ghost trick unlocks exact coalescence determinant","One ghost per collision yields a determinant","Determinants for coalescence from ghost particles"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00105,"raw_usage":{"total_tokens":4231,"prompt_tokens":713,"completion_tokens":3518,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":457,"completion_tokens_details":{"reasoning_tokens":3454}},"tokens_in":457,"tokens_out":3518,"duration_ms":26451,"temperature":1.0,"reasoning_tokens":3454,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T01:00:31.484273+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Enumerate all coalescing path tuples for three simple random walks on Z starting at 0, 1, 2, with the 2+1 coalescence pattern, up to a fixed small time; compute the survivor-position probabilities by exhaustive counting and compare with the determinant formula evaluated at the same transitions. A single mismatch would disprove the discrete claim.","supporting_citations":[],"review_version":1}