{"id":"65dd2571-d8f7-41fa-aede-bd4148990f11","arxiv_id":"2502.00073","paper_version":1,"verdict":"UNVERDICTED","confidence":"MODERATE","novelty_score":1.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper reviews a theorem stating that any Metropolis-type lattice model has a finite hierarchical decomposition of metastable plateaus with Markov-chain limits at each level, and surveys four Ising model examples.","lead":"This review paper lays out a claimed multi-level hierarchy of metastable transitions in low-temperature lattice models and summarizes how it plays out in four Ising model settings. It is useful as an entry point to the metastability literature, because it condenses several recent results into one framework.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 3 and Lemma 4 are stated without proof; strict separation Γ*_h > Γ*_{h-1} is the unverified hinge of the hierarchy.","rationale":"The construction is a plausible pathwise hierarchy: recurrent classes of the level-(h−1) trace chain are closed, so exits between them must cost strictly more than Γ*_{h−1}; hence Lemma 4 likely follows, and Theorem 1 follows by merging at the minimum inter-class barrier. I do not see a concrete counterexample in the text; the four examples are consistent with the scheme, e.g., Section 3.4 gives Γ* = 1, 2, 4. The reader's UNVERDICTED verdict is therefore right: the combinatorial skeleton looks sound, but the paper does not provide proofs of the key lemmas or of the convergence theorem, and the central universal claim cannot be certified from this article. I would not move to accept or reject; the appropriate status remains unverified pending the full version. The exhaustive small-graph check is a cheap way to test whether Lemma 4 or Theorem 1 could fail; if it passes, attention should shift to auditing [22]'s proof of Theorem 3.","tokens_in":17370,"tokens_out":17912,"duration_ms":225268,"concrete_test":"Implement the recursive construction of Section 2.1 exactly and exhaustively enumerate all connected graphs on n ≤ 6 vertices with integer energies in [−3, 3]; for each graph compute Γ*_h and ν_h at every level and test Γ*_h > Γ*_{h−1} and ν_h < ν_{h−1}. If any counterexample appears, Lemma 4 or Theorem 1 is false; if none appears, the remaining decisive check is to audit the proof of Theorem 3 in the full version [22], verifying the mode of convergence and the negligibility statement for the general finite-state setting.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim is Theorem 3: for every level h, the e^{βΓ*_h}-accelerated trace process converges to the trace chain X^{*,h}, and time spent outside V^{*,h} is negligible. The proof is not present; the text defers to [22]. Within the construction, the specific condition that makes the hierarchy nontrivial is Lemma 4 (Γ*_h > Γ*_{h-1}), stated without proof immediately after (28). If two consecutive minimal depths are equal, the two time scales merge, the accelerated process at level h is not separated from level h−1, and the claimed 'simple Markov chain at each level' loses its meaning. Lemma 4 is plausible from recurrence-class closedness—an inter-class transition at depth Γ*_{h−1} would merge the classes—but this argument is not supplied, and Lemma 3 (equal energy inside P^h_i), Theorem 1 (ν_h < ν_{h−1}), and Theorem 2 (ground states recurrent) are also unproved. The universal 'any abstract lattice system' assertion therefore rests on a chain of statements that cannot be checked from this document alone.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper announces a general theorem (Theorem 3) asserting that for any finite-state Metropolis lattice model, the collection of stable plateaux admits a hierarchical decomposition into finitely many levels, with the accelerated trace process converging to a simple Markov chain at each level. It presents the formalism of stable plateaux, cycles, and contracted/trace Markov chains in Section 2, and then applies this framework to four Ising-model settings: Glauber dynamics with positive and zero external fields and Kawasaki dynamics with few and many particles, stated as Theorems 4–11. The paper explicitly states that most proofs are omitted and refers to the author's earlier preprint [22] for the hierarchical decomposition and to other works for the concrete cases.","tokens_in":17629,"tokens_out":7897,"duration_ms":74922,"significance":"If Theorem 3 and its supporting lemmas are correct, the proposed framework would provide a useful unifying perspective on metastable tunneling in low-temperature lattice systems. The four examples cover classical settings, and several of the example results are drawn from published works, which lends some plausibility. However, the central general theorem and the key structural lemmas are stated without proof, and the proof is deferred to the author's own unpublished preprint [22]. The manuscript does not contain machine-checked proofs, reproducible code, or an independent verification of the main claim. Therefore, the significance of the paper as a contribution to the mathematical literature cannot be assessed from this document alone; it currently functions as an announcement of results proved elsewhere.","major_comments":[{"comment":"Lemma 4, stated immediately after equation (28), asserts the strict separation of time scales, Γ*_h > Γ*_{h-1}, but no proof is provided. This lemma is load-bearing: if two consecutive minimal depths coincide, the exponential time scales merge, the trace process at level h is not separated from level h−1, and the claimed simple Markov chain limits at each level lose their meaning. The text defers to [22], an unpublished preprint by the author. The paper should either include a proof or a proof sketch of Lemma 4, or clearly identify this statement as a conjecture or as a result announced in [22] so that the reader can judge its status.","section":"Section 2.1, Lemma 4"},{"comment":"In the statement of Theorem 3, the notion of convergence is not specified. The phrase 'converges' should be made precise, for example as weak convergence of the finite-dimensional distributions of the accelerated process in the Skorokhod space D([0,T], P^{*,h}) with the J1 topology, or as convergence of the full process in a suitable path space. The second assertion about the negligible time outside V^{*,h} also needs a precise interpretation of the expectation over the initial state. Without a precise statement, the theorem cannot be checked or applied.","section":"Section 2.2, Theorem 3"},{"comment":"The inductive construction of the hierarchy depends on several statements that are asserted without proof: Lemma 3 (equal energy inside P_i^h), Lemma 5 (disjointness and cycle structure of V_i^h), Lemma 6 (identification of (C^h)^⋆ and (C^h)^♯), Theorem 1 (strict decrease of the number of recurrent components), and Theorem 2 (ground-state recurrence). These are not mere technicalities: Theorem 1 guarantees that the procedure terminates, and Theorem 2 identifies the terminal level. Since these results are central to the claimed hierarchy, the manuscript should at least provide proof sketches or exact references to proofs in [22] for each statement.","section":"Section 2.1, Theorems 1–2 and Lemmas 3, 5, 6"},{"comment":"The fourth example (Kawasaki dynamics with many particles) is stated as a theorem, but its content, including the definitions of the stable plateaux in Theorem 9 and the hierarchical decomposition in Theorem 10, is deferred entirely to the author's preprint [22]. As a result, the reader cannot verify this application from the present manuscript. The paper should clearly distinguish between results that are established in the published literature and results that are announced from [22], and should provide enough detail to make the statements self-contained or point to a verifiable source.","section":"Section 3, Theorems 9–11"}],"minor_comments":[{"comment":"The introduction refers to 'Section ??, where we discuss some possible extensions', but no such section appears in the manuscript; this placeholder should be resolved.","section":"Section 1"},{"comment":"In equation (15), the rate R_C(F(C), η) is defined using |F(C)|^{-1}; the paper should explicitly note that this corresponds to choosing an element of F(C) uniformly at random when leaving the contracted cycle.","section":"Section 2, Definition 4"},{"comment":"The notation P^{*,k} in Section 3.1 denotes a set of configurations, while in Section 2 the same symbol denotes a collection of plateau sets. This conflation should be clarified, for example by stating that in the positive-field example every stable plateau is a singleton so that the collection can be identified with its union.","section":"Section 3.1"},{"comment":"The condition expressing that a previous cycle does not intersect any new cycle is written with a symbol that may be misprinted as '= / 0'; the intended condition 'C ∩ V_i^h = ∅' should be stated unambiguously.","section":"Equations (30) and Lemma 6"},{"comment":"Several typographical errors are present, such as 't-em perature' in the abstract; the manuscript would benefit from a careful proofreading pass.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The main concern is that the central theorem and the key supporting lemmas are not proved in this manuscript and are deferred to the author's own preprint [22], which is not publicly verified. This makes it difficult to ascertain the correctness of the claimed universal hierarchical decomposition. The examples based on published works are useful, but the fourth example also depends on [22]. I would like the editor to consider whether a proceedings review article can carry an announcement of an unpublished theorem in this way, or whether the authors should be required to either include proofs/sketches or clearly relabel the results as announced from a preprint."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nQuick take: this is a proceedings review article, not a new research result. The central Theorem 3, and the separation Lemma 4 it depends on, are stated without proof and deferred to the author's own arXiv preprint [22]. If you want to verify the general hierarchy, you need to read that. That said, the paper is a useful synthetic overview of metastable hierarchy for Metropolis lattice systems, and the four Ising case studies are drawn from credible sources.\n\nWhat's good: the recursive construction—stable plateaux, initial depth, cycles, trace chains—is clearly laid out, and Lemmas 1 and 2 are actually proved. The exposition is clean and the examples (positive/zero field Glauber, few/many particle Kawasaki) show the framework in action. The author is upfront that it's a review, and the citation patterns look honest: the external attributions to Beltrán–Landim and Kim–Seo match the literature I know.\n\nSoft spots: the load-bearing parts are missing. Lemma 4 (Γ*_h > Γ*_{h-1}) is the hinge—if two levels have equal minimal depth, the exponential time scales merge and the claimed Markov-chain limits stop being separate. It is stated without proof, and the same goes for Theorems 1, 2, and the main convergence result Theorem 3. All deferred to [22], which is a self-citation. Plausible, but unverifiable from this document alone. Also minor: Theorem 8 says \"Glauber dynamics\" where it should say Kawasaki, and there is a placeholder broken reference (\"Section ??\") in the intro. Nothing that affects the survey's content, but it is sloppy.\n\nWho gets value: someone wanting a map of metastable hierarchy results and a pointer to the literature; a PhD student or a researcher entering the area. Not someone needing a proof of the abstract theorem.\n\nMy view: as a survey for a proceedings volume, it is fine and useful. If it were a research submission to a math journal, I would want the full version [22] to be refereed or at least publicly available before considering the main theorem. I would accept it for peer review as a review article, with the expectation that the referee checks the statements against [22] and the typos get fixed.","headline":"A clean, useful survey of metastable hierarchy in Metropolis lattice systems, but the main theorem and its key separation lemma are unproved here and relegated to the author's own preprint.","tokens_in":18108,"tokens_out":2394,"would_cite":true,"duration_ms":25007,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60K35","82C22","82C20"],"pacs":[],"model":"deepseek-v4-flash","headline":"For any abstract lattice system with Metropolis-type dynamics, the stable plateaus organize into a finite hierarchy, and at each level the accelerated tunneling dynamics converges to a simple Markov chain.","keywords":["metastability","low-temperature lattice models","hierarchical decomposition","stable plateaux","energy barriers","Metropolis-type dynamics","Ising model","trace Markov chain"],"falsifier":"Compute the first two minimal energy barriers $\\Gamma^{\\star,1}$ and $\\Gamma^{\\star,2}$ in any finite Metropolis lattice model; equality would make the two exponential time scales coincide, so the accelerated processes could not converge to distinct level-1 and level-2 limit chains, contradicting the claimed hierarchy.","tokens_in":90,"feed_emoji":"🧊","tokens_out":10665,"duration_ms":172495,"temperature":0.7,"pith_summary":"This paper claims that metastability in low-temperature lattice systems is layered rather than a single rare event. For any finite lattice configuration space with a Hamiltonian and Metropolis-type jump rates, the stable plateaus can be grouped recursively into cycles according to the smallest energy barrier separating them. Each level $h$ has its own time scale $e^{\\Gamma^{\\star,h}\\beta}$; after accelerating by that factor, the tunneling between the deepest valley groups converges to a simple finite Markov chain, while the original process spends negligible time outside those valleys. The number of recurrent groups strictly decreases level by level, so the procedure terminates at a unique class that consists exactly of the ground states. If this is right, a complicated high-dimensional energy landscape is reduced to a finite stack of simple tunneling processes.","feed_headline":"Metastable lattice systems reduce to simple Markov chains per level","feed_subtitle":"At each of finitely many time scales, tunneling between stable plateaus converges to a simple Markov chain.","key_machinery":"The load-bearing mechanism is an iterative cycle contraction. A stable plateau is a connected set of equal-energy configurations whose boundary has strictly higher energy; the communication height $\\Phi(\\eta,\\xi)$ is the minimum over paths of the maximum energy along the path. At each level $h$, the plateaux are grouped into disjoint cycles $V_i^h$ whose depth equals the current minimal barrier $\\Gamma^{\\star,h}$, and the contracted graph and trace-chain construction produces a Markov chain on the recurrent groups at that level. The argument depends on the strict monotonicity $\\Gamma^{\\star,h}>\\Gamma^{\\star,h-1}$, which keeps the exponential time scales $e^{\\Gamma^{\\star,h}\\beta}$ separated; deeper plateaux act as absorbing states while the trace process encodes tunneling between plateaux of depth at least $\\Gamma^{\\star,h}$.","core_discovery":"On the paper's own terms, the discovery is that every abstract Metropolis-type lattice model carries a canonical hierarchical decomposition of metastability. Starting from the stable plateaux $P^1$, define the depth of a plateau as the smallest energy barrier that must be crossed to reach any other plateau; plateaux of minimal depth form the first recurrent groups, and the trace of the accelerated dynamics on them converges to a limit chain $X^{\\star,1}$. Repeating the construction on the irreducible components produces levels $h=1,\\dots,\\mathfrak{m}$ with strictly increasing minimal depths $\\Gamma^{\\star,h}$; Theorem 3 asserts that for each $h$ the $e^{\\Gamma^{\\star,h}\\beta}$-accelerated trace process on the recurrent plateau groups converges to the finite Markov chain $X^{\\star,h}$, and the original process spends negligible time outside the union of the level-$h$ valleys. The terminal level's unique recurrent class is exactly the set of ground states. The second half of the paper checks that this hierarchy is realized in four Ising-model settings and identifies in each case the explicit time scales and plateau classes.","pith_inferences":["A practical consequence that the paper leaves implicit is that the hierarchy could be computed algorithmically from finite portions of the energy landscape, turning metastability classification into a combinatorial search over droplet geometries.","If a model were found with equal consecutive minimal depths, the levels would merge rather than disappear, so the hierarchy would survive in coarsened form rather than being refuted outright.","The same recursive grouping should apply to any reversible finite Markov chain whose transition rates are exponentially small in a parameter, with stable plateaus replaced by the bottom sets of cycles."],"forward_implications":["If the central theorem is correct, metastable behavior in any Metropolis lattice model is described by a finite list of energy barriers and a limit Markov chain at each level.","The limiting chain at each level is finite-state and built from the trace process, so asymptotic hitting probabilities and transition rates become finite-dimensional, computable objects.","The ground states are recurrent at every level, and the unique recurrent class of the terminal level consists exactly of the ground states.","In the four Ising examples, the levels correspond to distinct physical mechanisms: edge flips and droplet shrinkage, droplet growth, strip formation, and the final transition to the ground state.","The construction uses only the Hamiltonian and the Metropolis rates, so the paper suggests the same hierarchy should transfer to broader rare-transition dynamics beyond the reversible Metropolis class."],"supporting_citations":[{"why":"Defines the trace Markov chain construction used to produce the level-$h$ limit processes.","marker":"[2]"},{"why":"Provides the classification of stable plateaus and the one-step absorbing chains for the positive-field Glauber case.","marker":"[3]"},{"why":"Provides the plateau classes and two-level structure for the few-particle Kawasaki case.","marker":"[7]"},{"why":"The full version that proves the general hierarchical decomposition and supplies the many-particle Kawasaki classification.","marker":"[22]"},{"why":"Provides the two-level structure and the random-walk limit for the zero-field Glauber case.","marker":"[24]"},{"why":"Supplies the cycle theory and exit-time asymptotics that justify treating plateaus as metastable elements.","marker":"[37]"}],"fun_headline_variants":["Lattice metastability: each tier reduces to a Markov chain","Every lattice model's metastable levels become simple Markov chains","Hierarchical metastable plateaus: each level converges to a Markov chain","Universal metastable hierarchy: tunneling limits are Markov chains"],"cache_read_input_tokens":20224,"weakest_assumption_plain":"The level-by-level hierarchy depends on the unproved assertion that the minimal energy barrier at each level is strictly larger than the one before it, since equal barriers would merge the exponential time scales and collapse the separate Markov-chain limits.","fun_headline_variants_meta":{"raw":{"variants":["Lattice metastability: each tier reduces to a Markov chain","Every lattice model's metastable levels become simple Markov chains","Hierarchical metastable plateaus: each level converges to a Markov chain","Universal metastable hierarchy: tunneling limits are Markov chains"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000304,"raw_usage":{"total_tokens":1747,"prompt_tokens":949,"completion_tokens":798,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":565,"completion_tokens_details":{"reasoning_tokens":728}},"tokens_in":565,"tokens_out":798,"duration_ms":8263,"temperature":1.0,"reasoning_tokens":728,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T22:00:23.342710+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the first two minimal energy barriers $\\Gamma^{\\star,1}$ and $\\Gamma^{\\star,2}$ in any finite Metropolis lattice model; equality would make the two exponential time scales coincide, so the accelerated processes could not converge to distinct level-1 and level-2 limit chains, contradicting the claimed hierarchy.","supporting_citations":[{"cited_title":"Beltr´ an and C","cited_arxiv_id":null,"evidence_quote":"Defines the trace Markov chain construction used to produce the level-$h$ limit processes."},{"cited_title":"Beltr´ an and C","cited_arxiv_id":null,"evidence_quote":"Provides the classification of stable plateaus and the one-step absorbing chains for the positive-field Glauber case."},{"cited_title":"Beltr´ an and C","cited_arxiv_id":null,"evidence_quote":"Provides the plateau classes and two-level structure for the few-particle Kawasaki case."},{"cited_title":"Metastable hierarchy in abstract low-temperature lattice models: an application to Kawasaki dynamics for Ising lattice gas with macroscopic number of particles","cited_arxiv_id":"2405.08488","evidence_quote":"The full version that proves the general hierarchical decomposition and supplies the many-particle Kawasaki classification."},{"cited_title":"Approximation method to metastability: an application to non-reversible, two-dimensional Ising and Potts models without external fields","cited_arxiv_id":"2212.13746","evidence_quote":"Provides the two-level structure and the random-walk limit for the zero-field Glauber case."},{"cited_title":"Olivieri and M","cited_arxiv_id":null,"evidence_quote":"Supplies the cycle theory and exit-time asymptotics that justify treating plateaus as metastable elements."}],"review_version":1}