{"id":"d87dace3-1069-42d2-8575-914054643f38","arxiv_id":"1908.02359","paper_version":4,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Stochastic fusion builds higher-spin exclusion processes and duality functions from stationary measures of exclusion processes via Rogers-Pitman intertwining.","lead":"This paper introduces a new construction, stochastic fusion, that builds multi-occupancy interacting particle systems such as ASEP(q,m) out of single-occupancy exclusion processes, using only their stationary measures and the Rogers-Pitman intertwining. It then derives new duality functions and hydrodynamic limits for symmetric, asymmetric, and dynamic exclusion processes with open boundaries.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The finite-lattice q-exchangeability import from Kua19 is the load-bearing unproved step in the stochastic fusion construction.","rationale":"The reader's weakest assumption identifies exactly the same load-bearing dependency on Kua19 Proposition 4.5, and my reading agrees that the finite-lattice applicability is the point of maximum risk. I do not see a contradiction inside the paper's own generator computations; the issue is that the decisive preservation statement is imported from a self-cited prior work without adapting the proof to reflecting boundary conditions. If the imported result is true in the needed finite-lattice form, the main construction, the stationarity of π, and the q-exchangeability-based duality outputs all go through as written, and the paper's own caveats (Remark 9, Section 5.4) show that the author is aware of the limits of the general method. If the imported result is not true in that form, the finite-lattice construction and the duality corollaries built on it are unsupported. This is an addressable but genuine correctness risk, so the CONDITIONAL verdict is appropriate: the paper should either supply an independent proof of finite-lattice q-exchangeability preservation or explicitly state the range of validity of Kua19's result and restrict the theorems accordingly. My concrete test is deliberately small enough to be checked exactly, and it directly targets the boundary terms that distinguish the finite-lattice setting from Z. A positive result would resolve the concern; a negative result would force a substantial revision of Theorems 4.2 and 4.6.","tokens_in":47985,"tokens_out":17761,"duration_ms":193746,"concrete_test":"For the reflecting-boundary n=2, L=3 ASEP with one particle of each species and one hole, compute the generator action on the full q-exchangeable family p(x)q^{inv σ}/Z and verify symbolically or by exact enumeration that the family is invariant, that the stationary base p(x)=constant gives L^*(πΛ)=0 for the π of Theorem 4.2(i), and that QL_ASEP^(n)=L_ASEP^(p)Q holds at both boundary edges; if any of these fails, the finite-lattice version of Kua19 Prop. 4.5 is false and the finite-lattice construction is unproved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central construction is only valid on finite lattices through Theorem 4.2, and every part of that theorem feeding the duality output—stationarity of πΛ in 4.2(i), stationarity of π in 4.2(iii), q-exchangeability preservation of ASEP(q,m) in 4.2(iv), and the QL_ASEP^(n)=L_ASEP^(p)Q intertwining behind Theorem 4.6(a) and Corollaries 4.7/4.8—rests on the imported statement (Kua19, Prop. 4.5) that multi-species ASEP preserves q-exchangeability. The paper cites this result but does not prove it for the finite lattice with reflecting boundaries used here, where the generator has boundary terms and the usual infinite-line proof does not automatically apply. If q-exchangeability preservation fails, or holds only on Z, then Theorem 4.2(i) is not established, π is not known to be stationary, the fused semigroup is not known to preserve q-exchangeability, and the duality functions in Theorem 4.6(a) and its corollaries lose their proof. The paper's own Remark 9 notes that the direct fused candidate ΛDSchΦ is not a duality for generic m, so the positive duality output depends on this specific imported mechanism rather than on a robust direct check. The second limitation flagged by the reader, namely that Rogers–Pitman only gives Markovity and duality under the special initial measure πΛ, is partly mitigated by the paper's direct proofs of full duality in Theorems 4.11 and 4.15; the finite-lattice q-exchangeability import is the more consequential gap.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces 'stochastic fusion,' a method for constructing higher-occupancy interacting particle systems from exclusion processes using stationary measures and the Rogers–Pitman intertwining. Theorem 3.1 gives conditions under which a projection of a Markov process is Markov and inherits a weighted duality. The method is applied to symmetric exclusion, ASEP(q,m), and dynamic ASEP, producing inhomogeneous fused processes, duality functions, stationary measures, and hydrodynamic limits for open boundaries. An appendix co-authored with Amol Aggarwal relates the construction to fusion of stochastic vertex models.","tokens_in":48271,"tokens_out":29247,"duration_ms":283111,"significance":"If the main construction holds, the paper provides a general mechanism for generating higher-spin exclusion processes and duality functions from stationary measures, generalizing results in [GKRV09], [CGRS16], and [CGR19] to inhomogeneous lattices and open boundaries. The paper includes explicit computations, honest disclosure of where the Ansatz fails (Remark 9, Section 5.4), and new duality and hydrodynamic-limit results. The main risk is the imported finite-lattice q-exchangeability result; additionally, the stationary measure conclusion in Theorem 4.12(b) is incorrect as stated.","major_comments":[{"comment":"Theorem 4.2(i),(iv) and Theorem 4.6(a) rest on the imported result (Kua19, Prop. 4.5) that multi-species ASEP preserves q-exchangeable measures. The finite-lattice, reflecting-boundary version of this statement is not proved in the manuscript, and the text does not specify whether the cited proposition covers the generator with boundary terms. Since stationarity of πΛ and the intertwining QL^(n)=L^(p)Q both use this import, please provide the precise statement of the cited result and either prove the finite-lattice version or give a reference that does.","section":"Theorem 4.2 and Theorem 4.6"},{"comment":"The proof via duality determines only the factorial moments E[∏ s(x_i)/m] = α^d, which are those of a product of Binomial(m,α) measures. The theorem's claimed 'i.i.d. product of Bernoulli measures of parameter α' with P(s∞(x_1)=1,...,s∞(x_d)=1)=α^d is false for m>1; for a Bernoulli(α) product the displayed probability would be (α/m)^d. The stationarity statement should be corrected to the i.i.d. product of Binomial(m,α) distributions (the m=1 case being Bernoulli(α)).","section":"Theorem 4.12(b)"}],"minor_comments":[{"comment":"Remarks 12 and 14 contain '[ ?]' placeholders with no matching bibliography entries; these citations should be completed or removed.","section":"Remarks 12 and 14"},{"comment":"The terms SEP and SSEP are used inconsistently across the abstract, Section 4.2, and Theorem 4.12; please standardize the nomenclature for the symmetric process.","section":"Abstract and Section 4.2"},{"comment":"The comparison '... > 1 > ...' is unclear; the intended comparison between the two jump rates should be written explicitly.","section":"Section 2.3.2"},{"comment":"The phrase 'identity equivalent to (21)' refers to an equation that is never labeled; the intended reference appears to be equation (13).","section":"Theorem 4.11 proof"},{"comment":"The phrase 'By direction computation' is a typo for 'By direct computation'.","section":"Theorem 4.12(a)"},{"comment":"The dual process in Lemma 4.16 is understood to have an absorbing empty state after a particle exits at -1; this convention should be stated explicitly when DSchP^{-1}(0,x(t)) is used.","section":"Section 4.3, Lemma 4.16"}],"recommendation":"major_revision","confidential_remarks":"The load-bearing point is the finite-lattice q-exchangeability import; if the author can supply the missing proof or a precise reference covering reflecting boundaries, and correct the stationary measure statement, the paper would be publishable. The stationary measure error is localized and fixable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Jeff,\n\nThis is a real contribution. The stochastic fusion construction is new: it takes an exclusion process with stationary measures and builds higher-spin processes (inhomogeneous ASEP(q,m), SSEP(m/2), and a dynamic version) via Rogers–Pitman intertwining. The generator calculations are explicit and careful, and the paper is notably honest about where the method stops: Remark 9 concedes the direct fused duality Ansatz fails for generic m, and Section 5.4 says plainly that the dynamic duality Ansatz does not produce a duality. That honesty earns trust.\n\nThe strongest parts are the open-boundary duality results and the hydrodynamic limits for SSEP(m/2) and ASEP. Theorems 4.11 and 4.15 give full dualities, not just the special-initial-condition ones that the Rogers–Pitman setup yields directly. The appendix with Aggarwal is a useful algebraic complement, and the comparison to vertex-model fusion is illuminating.\n\nThe main soft spot is exactly what the stress-test flags: Theorem 4.2 imports from Kua19 the fact that multi-species ASEP preserves q-exchangeability, and uses it on a finite lattice with reflecting boundaries. The paper does not prove that finite-lattice version. I do not think this is fatal. On a finite interval the generator is just the infinite-line generator restricted to internal edges, so a local proof of q-exchangeability preservation should transfer without trouble. But the paper should either state that reduction explicitly or give a direct check, because Theorem 4.2(i) and (iv) and the duality outputs depend on it.\n\nTwo smaller issues. The hydrodynamic limit arguments import external Brownian approximation results (Erhard–Franco–da Silva) without derivation; that is standard in this literature, but a referee should ask for a precise statement. And the abstract oversells the generality of the duality output: the general construction only gives duality for special initial distributions, and the full dualities are proven separately. That distinction should be in the abstract.\n\nBottom line: this deserves a serious referee. I would send it out, and ask the referee to pressure-test the Kua19 import and the hydrodynamic-limit statements. I would also bring it to our reading group.\n\n— Candidly, your colleague.","headline":"Original, honest, and mostly correct; the main risk is an under-proved import of q-exchangeability preservation on finite lattices.","tokens_in":48806,"tokens_out":2346,"would_cite":true,"duration_ms":29132,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60K35","82C22","60J27"],"pacs":[],"model":"deepseek-v4-flash","headline":"Stochastic fusion builds multi-occupancy versions of exclusion processes using only their stationary measures, and carries Markov duality over to the fused process.","keywords":["stochastic fusion","interacting particle systems","Markov duality","q-exchangeability","ASEP(q,j)","SSEP(m/2)","hydrodynamic limit","dynamic ASEP"],"falsifier":"Take a two-site lattice with capacities (m_1,m_2)=(2,3), fix q∈(0,1), run the multi-species ASEP(q,⃗m) from the proposed stationary measure π, and compare the empirical occupation distribution at large time with π; any persistent discrepancy contradicts Theorem 4.2(iii). Alternatively, check the q-exchangeability identity (3) on a single two-site jump of the multi-species ASEP(q,⃗m): a violation there would break the construction's core premise.","tokens_in":47747,"feed_emoji":"🎲","tokens_out":11193,"duration_ms":104910,"temperature":0.7,"pith_summary":"Stochastic fusion is a general recipe for converting an exclusion process—one in which each lattice site holds at most one particle—into a process in which sites may hold several particles. The recipe needs only the stationary measures of the original process on a finite lattice, together with a classical intertwining criterion for Markov functions. When the original process has a Markov duality, the fused process inherits a weighted duality, at least for the special initial measures the construction uses. The paper works this out for symmetric exclusion, asymmetric exclusion, and dynamic models, producing multi-occupancy versions of ASEP(q,j) and SSEP(m/2), new duality functions, and hydrodynamic limits for open-boundary versions.","feed_headline":"Fusing exclusion processes yields multi-particle lattice systems","feed_subtitle":"Only stationary measures are needed to get multi-occupancy versions of ASEP and SSEP, with duality.","key_machinery":"The central object is the pair (Φ,Λ) with ΛΦ=id. Φ is deterministic fusion: at each lattice site x it collapses m_x single-occupancy micro-sites into one site with occupancy vector (k_x^(1),...,k_x^(n)), total at most m_x. Λ is stochastic fission: it expands an occupancy vector back into a micro-configuration by placing the k_x^(i) particles of each species on the m_x micro-sites, choosing the placement with probability proportional to $q^{{sum of positions}}$ $q^{{inversions}}$ divided by a q-multinomial. These probabilities are exactly the q-exchangeable reversible measures of the original multi-species process. Because ΛΦ=id, the identity π̂ΛP_t=π̂ΛP_tΦΛ follows whenever π̂Λ is stationary, and then the projected process has semigroup Q_t=ΛP_tΦ and inherited duality functionals ΛD and ΛDΦ; this is the mechanism that carries the argument.","core_discovery":"The central claim is Theorem 3.1: given a Markov process X_t with semigroup P_t, a deterministic fusion map Φ, and a stochastic fission kernel Λ with ΛΦ=id, if some π̂Λ is stationary then φ(X_t) is Markov with semigroup Q_t=ΛP_tΦ, and every duality P_tD=D\\tilde P_t^* of X_t yields the weighted duality π̂Q_t(ΛD)=π̂(ΛD)\\tilde P_t^*. The paper applies this to the multi-species interchange process: the induced fused process is the inhomogeneous multi-species ASEP(q,⃗m) (Theorem 4.2), which has stationary measure π, preserves q-exchangeability, and admits duality functionals that specialize to previously known ones. The symmetric limit gives multi-species SSEP(⃗m), whose boundary reservoirs in the infinite-capacity limit lead to open-boundary SSEP with a duality, stationary measures, and hydrodynamic limit. In the asymmetric case the same mechanism yields open-boundary ASEP duality and its hydrodynamic limit, and fusing the dynamic ASEP produces an inhomogeneous dynamic ASEP(q,⃗m) that interpolates between ASEP(q,⃗m) and its space reversal.","pith_inferences":["Beyond the paper: the same recipe should apply to any finite-state Markov process with an explicitly known stationary measure—zero-range and inclusion processes are natural test cases—and would yield fused multi-occupancy versions with dualities.","Beyond the paper: the hydrodynamic limits found here indicate that each fused process should lie in the same macroscopic universality class as its parent exclusion process (diffusive for symmetric, KPZ-type for asymmetric); the paper motivates but does not prove this.","Beyond the paper: since the intertwining criterion guarantees Markovity only for the special initial class π̂Λ, any claimed duality for arbitrary initial conditions would require extra work; the paper's Ansatze suggest this may be possible for specific models.","Beyond the paper: verifying whether the dynamic ASEP(q,⃗m) has the same weak-asymmetry limit as the dynamic ASEP—the question the paper poses—would test whether the dynamic structure survives fusion in the scaling limit."],"forward_implications":["For symmetric exclusion, the fusion produces the inhomogeneous SSEP(⃗m); taking some capacities to infinity yields an open-boundary SSEP whose duality gives the hydrodynamic limit (heat equation with Neumann boundary condition) and i.i.d. Bernoulli(α) stationary measures on the half-line.","For asymmetric exclusion, the fusion produces the inhomogeneous ASEP(q,⃗m), and as a by-product proves that multi-species ASEP(q,j) preserves q-exchangeable measures, yielding new duality functions for ASEP, ASEP(q,j), and the q-Boson.","The open-boundary ASEP obtained this way has a duality from which the hydrodynamic limit follows: the q-deformed particle count converges to the solution of the heat equation with convection and Neumann boundary.","Fusing the dynamic ASEP produces a dynamic inhomogeneous ASEP(q,⃗m) whose rates interpolate, as the dynamic parameter goes from 0 to ∞, between ASEP(q,⃗m) and its space reversal; all capacities equal to 1 recovers the dynamic ASEP.","Because the recipe only needs stationary measures rather than an underlying symmetry algebra, it can be applied to processes without a known algebraic structure, at the price that Markovity and duality are guaranteed only for the special initial measure π̂Λ."],"supporting_citations":[{"why":"Supplies the Markov-function intertwining criterion (ΛΦ=id and ΛP_t=ΛP_tΦΛ imply the projection is Markov) on which Theorem 3.1 is built.","marker":"[RP81]"},{"why":"Proves that multi-species ASEP preserves q-exchangeable measures (its Proposition 4.5), the step used to establish stationarity of πΛ and q-exchangeability of the fused process.","marker":"[Kua19]"},{"why":"Defines ASEP(q,j) and its self-duality; the fused ASEP(q,⃗m) and its duality functions are inhomogeneous generalizations of these.","marker":"[CGRS16]"},{"why":"Defines SEP(m/2) and its duality with boundary reservoirs; the symmetric fusion and open-boundary duality generalize these results.","marker":"[GKRV09]"},{"why":"Establishes the standard ASEP duality, the input duality that, after applying Λ and Φ, yields the fused duality functionals.","marker":"[Sch97]"},{"why":"Introduces the dynamic ASEP, which the paper fuses into the dynamic inhomogeneous ASEP(q,⃗m).","marker":"[Bor17]"},{"why":"Provides the stationary measure and duality for dynamic ASEP on the infinite lattice used to compare the dynamic fusion and the duality Ansatz.","marker":"[BC18]"},{"why":"Gives multi-species ASEP(q,j) duality and reversible measures that the new duality functions extend and compare against.","marker":"[Kua17]"}],"fun_headline_variants":["Stochastic fusion turns exclusion processes into multi-occupancy systems","New fusion method creates multiparticle versions of ASEP and SSEP","From exclusion to multi-particle: stochastic fusion with duality","Fusing exclusion processes gives dual multi-species ASEP and SSEP"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The construction depends on the prior theorem that multi-species ASEP preserves q-exchangeable measures; if that theorem were false, the stationarity of πΛ, the q-exchangeability of the fused process, and the duality outputs built on them would not follow.","fun_headline_variants_meta":{"raw":{"variants":["Stochastic fusion turns exclusion processes into multi-occupancy systems","New fusion method creates multiparticle versions of ASEP and SSEP","From exclusion to multi-particle: stochastic fusion with duality","Fusing exclusion processes gives dual multi-species ASEP and SSEP"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000798,"raw_usage":{"total_tokens":3615,"prompt_tokens":1153,"completion_tokens":2462,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":769,"completion_tokens_details":{"reasoning_tokens":2390}},"tokens_in":769,"tokens_out":2462,"duration_ms":16558,"temperature":1.0,"reasoning_tokens":2390,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:48:02.535358+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a two-site lattice with capacities (m_1,m_2)=(2,3), fix q∈(0,1), run the multi-species ASEP(q,⃗m) from the proposed stationary measure π, and compare the empirical occupation distribution at large time with π; any persistent discrepancy contradicts Theorem 4.2(iii). Alternatively, check the q-exchangeability identity (3) on a single two-site jump of the multi-species ASEP(q,⃗m): a violation there would break the construction's core premise.","supporting_citations":[],"review_version":1}