{"id":"cb10210a-3efa-4d5b-87f5-ab6b4e89f1e8","arxiv_id":"2411.16756","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For every r>=2, the Martin boundary of the r-differential Young-Fibonacci graph is described explicitly by boundary words with a parameter beta, plus the Plancherel measure, and all these measures are ergodic.","lead":"This paper describes all invariant probability measures on infinite paths of the r-differential Young-Fibonacci graph, a natural family of graphs built from words. It extends the known Martin boundary and ergodicity results from the classical case r=1 to every r>=2, with explicit formulas.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Distinctness of the listed Martin-boundary measures is announced as 'proved later' in Remark 7, but no proof appears; Theorem 4+Remark 7 therefore only gives exhaustion, not the claimed exact classification.","rationale":"The reviewer's weakest-assumption pick (Theorem 2) is reasonable, but the counting formula itself has strong support: it is derived by explicit bijections (Assertions 1-2) and reduces to known YF counts. I did not find a concrete error in the termwise limiting step used in Theorem 4. The actual gap that threatens the central claim is the unproved distinctness assertion. A classification theorem needs both directions; Remark 7 explicitly promises a proof that never materializes. Ergodicity and projection do not remove the need for injectivity. This is a missing proof rather than a contradiction, so the appropriate disposition is CONDITIONAL, matching the reader's verdict; hence verdict_should_be is UNCHANGED. The proposed test is a direct finite computation using the paper's own formula, so it can settle the issue without new theory.","tokens_in":13062,"tokens_out":15028,"duration_ms":123466,"concrete_test":"Carry out the missing injectivity check. For r=2, let v=...21_1 and v'=...21_2 be infinite words with the same positions of 1s and 2s, so s(v)=s(v'), and take β=1. Using the limit formula in Theorem 4, compute μ_{2,v,1}(1_1) and μ_{2,v',1}(1_1). The terms not involving 1_{w,v} coincide, while 1_{1_1,v}=0 and 1_{1_1,v'}=1, so the values differ iff the subtracted term lim_n d1(ε,vn{2})/d1(ε,vn)·2 is nonzero; evaluate this limit explicitly. It is a ratio of products of g-values for a fixed finite suffix of v and is positive when π(v)>0. This would establish injectivity on the index of the rightmost unit; combining with Corollary 4 and Theorem 3(3) then separates all parameters. If the term vanished, repeat with w=1_21_1, which detects the next unit index.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is two-fold: every Martin-boundary measure lies in the list (exhaustion) and every list element is a distinct boundary measure (injectivity). Theorem 4 establishes exhaustion by taking termwise limits. Injectivity is never proved. Remark 7 says 'Различность этих мер будет доказана позже' ('their distinctness will be proved later'), but Sections 3-5 contain no such proof; Theorem 10 only proves ergodicity, which does not imply that different parameters give different measures. The projection identity in Corollary 4 reduces a hypothetical equality μ_{r,v,β}=μ_{r,v',β'} to μ_{s(v),β}=μ_{s(v'),β'} on YF, hence by Theorem 3(3) to s(v)=s(v') and β=β'. This does not separate words with the same forgetting map but different unit indices. The formula in Theorem 4 contains the term 1_{w,v}·d1(w{e(w,v)+1},v{e(w,v)+1})·r^{e(w,v)}, which is sensitive to whether w and v end in units of the same index; without a concrete check that this term makes v↦μ_{r,v,β} injective on each fiber of s, the classification may be redundant and the 'consists exactly' claim is unsupported.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the r-differential version of the Young–Fibonacci graph YF_r for integers r ≥ 2. It first proves a path-count formula (Theorem 2) that expresses the number of downward paths d_r(w,v) in terms of the ordinary Young–Fibonacci counts d_1 and powers of r, using the map s that forgets the indices of unit symbols. It then uses this formula, together with the known Martin boundary classification for YF, to show that every Martin-boundary measure is a limit of one of the two types μ_{r,v,β} (for infinite boundary words v with π(v)>0 and β∈(0,1]) or μ_{r,P} (the Plancherel measure). The paper further claims that all these measures are ergodic, with proofs for the Plancherel case and for the general case carried out by analogy with earlier works. The central aims are a complete classification of central measures on YF_r and their ergodicity.","tokens_in":13328,"tokens_out":5417,"duration_ms":49822,"significance":"If the classification is correct, the paper gives a complete description of the Martin boundary of the path space of YF_r for every r ≥ 2, extending the fundamental results of Goodman–Kerov [4] and the ergodicity results [7,8,9] to the r-differential family. The path-count transfer via the forgetting map s is a natural and clean reduction, and the computations in Theorems 2 and 4 are coherent and well structured; the use of previously established YF results as lemmas is appropriate. The claimed result would be a valuable contribution to the asymptotic theory of graded graphs and central measures. However, as detailed below, the exactness of the classification is not fully established because the distinctness of the listed measures is only announced and never proved.","major_comments":[{"comment":"The classification theorem is incomplete because the distinctness of the listed measures is asserted but not proved. Theorem 4 establishes only exhaustion: every sequence v_n has a subsequence whose limit measure is of the form μ_{r,v,β} or μ_{r,P}. It does not show that different pairs (v,β) give different measures, nor that μ_{r,P} is distinct from the μ_{r,v,β}. Remark 7 states 'Различность этих мер будет доказана позже', but no such proof appears in Sections 3–5. Corollary 4 reduces a hypothetical equality μ_{r,v,β}=μ_{r,v',β'} to the equality μ_{s(v),β}=μ_{s(v'),β'} on YF, which by Theorem 3(3) gives s(v)=s(v') and β=β', but this does not separate words with the same forgetting map but different unit indices. The term 1_{w,v} d_1(w{e(w,v)+1}, v{e(w,v)+1}) r^{e(w,v)} in the formula of Theorem 4 is sensitive to such differences, but no concrete check of injectivity on each fiber of s is provided. Consequently the statement in Remark 7 that the Martin boundary 'состоит только из таких мер' is unsupported as an exact description; at present the paper establishes only that every boundary measure is among the listed ones, not that the list is irredundant.","section":"Section 3, Remark 7"},{"comment":"The proof of ergodicity of all Martin-boundary measures is delegated: Theorem 10 is stated to be 'fully analogous' to Corollary 11 of [9], and Theorem 5 is likewise attributed to [7] without a detailed adaptation. Since ergodicity is one of the two headline claims, the reader cannot verify these results from the manuscript alone. In particular, Theorem 10 relies on Theorem 9 and on the classification of Theorem 4; if the classification is only one-sided (as noted above), the ergodicity statement may refer to a list with possible redundancies. The author should either write out the adaptation or give a precise dictionary between the parameters v, β, r here and the corresponding parameters in [9].","section":"Section 4, Theorem 10"}],"minor_comments":[{"comment":"The phrase 'известный естественную r-дифференциальную' contains a grammatical error; it should read 'известную естественную r-дифференциальную'.","section":"Section 1, last paragraph"},{"comment":"The statement 'Пусть r ∈ Nn⩾2' contains a typo; it should read 'Пусть r ∈ N, r ⩾ 2'.","section":"Corollary 4"},{"comment":"In the limit computation, the quantities e(w,v_n) and h(w,v_n) are replaced by e(w,v) and h(w,v) without an explicit note; this is justified by the symbolwise convergence v_n→v, but the stabilization should be stated for clarity.","section":"Proof of Theorem 4"},{"comment":"References [5] and [6] appear to be two versions (journal and arXiv) of the same path-enumeration result; the relationship between them should be clarified or one should be cited.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The paper is a Russian-language arXiv preprint; if it is under consideration for an English-language journal, a translation and stylistic revision will be needed. The main technical content seems sound in outline, but the missing distinctness proof is a genuine gap in the central classification claim, and the ergodicity proofs are too terse to verify. I would ask the author to supply a full proof of injectivity of the map (v,β)↦μ_{r,v,β} and a more explicit treatment of Theorems 5 and 10 before the paper can be accepted."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"New here: explicit Martin boundary formulas for YF_r, r≥2, and ergodicity of the resulting measures. The transfer via the forgetting map s, especially Theorem 2's path-count formula with the r-power correction, is genuinely useful and I think correct. Theorem 4's limit computations are coherent, and Corollary 4's projection identity is a nice tool that reduces many questions to the classical YF case. The Plancherel-type formula also looks right.\n\nSoft spots, in order of size. The distinctness claim is not proved. Remark 7 says the measures will be shown distinct later, but no such proof appears in Sections 3–5. Theorem 4 establishes only that the limit depends on (v,β), not that different pairs give different measures. Corollary 4 reduces an equality in a fiber of s to equality in YF, but it does not separate words inside the same fiber S_r(u). The term 1_{w,v}·d1(...)·r^{e(w,v)} in Theorem 4 depends on unit indices, so a direct check might work, but it is absent. This is a genuine gap in the central classification statement: as it stands, you get exhaustion but not the claimed exact list. Second, the ergodicity proof of Theorem 10 is deferred as \"fully analogous\" to [9]; the bounds in Theorem 9 are plausible but compressed, and a referee should verify the tail estimates transfer cleanly for r≥2. Minor: the proof leans on the author's own path-count papers [5,6] as black boxes; that is acceptable since they are published, but it means Theorem 2 is not self-contained.\n\nOverall the extension is real and the method is sound, but the classification as written is incomplete. This should not be desk-rejected; a serious referee could chase the distinctness proof and the ergodic transfer. I would send it to review with a request to fill the gap or state the classification as exhaustion only.","headline":"A genuine extension of the Young–Fibonacci boundary results to all r≥2, with a solid path-count transfer, but the classification is incomplete because distinctness of the listed measures is promised but never proved.","tokens_in":13858,"tokens_out":2003,"would_cite":true,"duration_ms":20801,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["05A15","05E10","31C35","60J50"],"pacs":[],"model":"deepseek-v4-flash","headline":"The Martin boundary of the r-differential Young–Fibonacci graph consists exactly of the $\\mu_{r,v,\\beta}$ measures plus a reweighted Plancherel measure, and every one of them is ergodic.","keywords":["central measures","Martin boundary","Young–Fibonacci graph","r-differential graph","graded graph path space","ergodicity","Plancherel measure","path enumeration"],"falsifier":"Compute both sides of Theorem 2 for a small pair in $YF_2$, for example $w=1_1$ and $v=1_1 2 1_2$, by listing all downward paths directly; any disagreement falsifies the formula. Equivalently, evaluate the Martin ratio $d_2(\\varepsilon,w)d_2(w,v_n)/d_2(\\varepsilon,v_n)$ along a sequence $v_n$ converging to a boundary word with $\\pi(v_n)\\to 0$ and check that the limit equals $\\mu_P(s(w))/r^{e(w)}$ for every $w$; one counterexample overturns the classification.","tokens_in":12860,"feed_emoji":"📐","tokens_out":14508,"duration_ms":125212,"temperature":0.7,"pith_summary":"For each integer $r\\ge 2$, the paper gives a complete description of the Martin boundary of the path space of the $r$-differential Young–Fibonacci graph, the natural $r$-parameter generalization of the classical Young–Fibonacci lattice. The boundary consists of the measures $\\mu_{r,v,\\beta}$, one for every infinite boundary word $v$ with $\\pi(v)>0$ and every $\\beta\\in(0,1]$, together with a single reweighted Plancherel measure $\\mu_{r,P}$. Each measure is obtained as a limit of passage probabilities through a finite vertex along sequences of vertices converging to a boundary word, and the limiting value depends only on the finite vertex, the boundary word, and $\\beta$. The paper then proves that every measure in this list is ergodic, so the Martin boundary is exactly the ergodic list of Martin limits for the $r$-differential family.","feed_headline":"Every Martin-boundary measure of the r-YF graph is ergodic","feed_subtitle":"The boundary is the μ_{r,v,β} family plus a reweighted Plancherel measure for every r≥2.","key_machinery":"The machinery is bijective path counting. Theorem 2 expresses the number $d_r(w,v)$ of downward paths between finite vertices as $r^{d(v)-\\#w}$ times a finite combination of ordinary Young–Fibonacci path counts $d_1$ after forgetting unit indices: $d_r(w,v)=r^{d(v)-\\#w}\\bigl(d_1(w,v)+\\sum_{l=1}^{e(w,v)}d_1(w\\{l\\},v\\{l\\})(r^l-r^{l-1})-\\mathbf{1}_{w,v}d_1(w\\{e(w,v)+1\\},v\\{e(w,v)+1\\})r^{e(w,v)}\\bigr)$. The $g$-function encodes the positions and lengths of the blocks of twos in a word, and the product $\\pi(v)$ built from $g$ is the parameter that decides whether a boundary sequence contributes a $\\beta$-family or collapses to the Plancherel limit. Assertion 4 converts the path-count formula into the termwise limits used in every Martin-boundary calculation.","core_discovery":"The central claim is Theorem 4 together with Remark 7 and Theorem 10. For every $r\\ge 2$, fix a boundary word $v$ over the alphabet $\\{1_1,\\dots,1_r,2\\}$ with $\\pi(v)=\\prod_{g(v,i)>1}\\frac{g(v,i)-1}{g(v,i)}>0$ and a parameter $\\beta\\in(0,1]$. If a sequence of finite vertices $v_n$ converges symbolwise to $v$ and $\\pi(v_n)/\\pi(v)\\to\\beta$, then for every finite vertex $w$ the Martin ratio $d_r(\\varepsilon,w)d_r(w,v_n)/d_r(\\varepsilon,v_n)$ converges to a value $\\mu_{r,v,\\beta}(w)$ depending only on $w$, $v$, and $\\beta$. If instead $\\pi(v_n)\\to 0$, the same ratio converges independently of $v$ to the Plancherel-type measure $\\mu_{r,P}(w)=\\mu_P(s(w))/r^{e(w)}$, where $s(w)$ is $w$ with every unit index erased and $\\mu_P$ is the ordinary Young–Fibonacci Plancherel weight $d_1(\\varepsilon,s(w))^2/|w|!$. Remark 7 states that no other Martin-boundary measures exist, and Theorem 10 states that every measure in this list is ergodic.","pith_inferences":["A consequence not drawn in the paper is that, since the Martin boundary is ergodic, every central measure on $YF_r$ should admit an ergodic decomposition supported on this boundary, not only those that arise directly as Martin limits.","The quotient relation of Corollary 4 suggests that the Martin boundary of $YF_r$ is a fibration over the classical Young–Fibonacci boundary, with the parameter $\\beta$ and the index choices carried in the fibre; a testable formulation would be a bijection between the two boundary sets with the forgetting map.","For $r=2$ one could numerically simulate random walks conditioned to pass through growing vertices and compare the empirical limits with the closed formulas obtained from Theorem 2; this would serve as a computational check of the boundary classification."],"forward_implications":["For every $r\\ge 2$ the Martin boundary of the $r$-differential Young–Fibonacci graph is fully classified; no other Martin-central measures exist.","The Plancherel measure on $YF_r$ is the ordinary Young–Fibonacci Plancherel weight divided by $r^{e(w)}$, namely $\\mu_{r,P}(w)=\\mu_P(s(w))/r^{e(w)}$.","Every boundary measure $\\mu_{r,v,\\beta}$ and $\\mu_{r,P}$ is ergodic, so almost every infinite path under such a measure sees the cylinder weights as limiting empirical frequencies.","Summing $\\mu_{r,v,\\beta}$ over all liftings of a Young–Fibonacci vertex under the index-forgetting map gives the corresponding ordinary boundary measure $\\mu_{v,\\beta}$, as stated in Corollary 4.","The explicit path-count formula gives finite-vertex Martin kernels that can be used to compute boundary measures numerically for concrete words and parameters."],"supporting_citations":[{"why":"Supplies the Martin-boundary theorem and the asymptotic proposition for the ordinary Young–Fibonacci graph that the $r$-differential proof reduces to.","marker":"[4]"},{"why":"Gives the path-count formula $d_1(w,v)$ for the Young–Fibonacci graph, quoted as Theorem 1 and used inside Theorem 2.","marker":"[5]"},{"why":"Provides the extended version of the same Young–Fibonacci path-count formula, cited together with [5] as the base of Theorem 2.","marker":"[6]"},{"why":"Proves ergodicity of the Plancherel measure on the Young–Fibonacci graph, whose argument Theorem 5 transfers to $\\mu_{r,P}$.","marker":"[7]"},{"why":"Supplies the vanishing-sum estimates (Corollary 10 and Corollary 11) used in Theorems 6, 8, and the ergodicity proof of Theorem 10.","marker":"[9]"}],"fun_headline_variants":["Martin boundary of r-YF graph: all measures ergodic","Every Martin measure on r-YF paths is ergodic","r-YF boundary: explicit measures, all ergodic","Explicit Martin boundary for r-YF: all measures ergodic","r-differential YF: boundary measures all ergodic"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole boundary classification rests on Theorem 2's path-count formula, which assumes the earlier Young–Fibonacci enumeration and counts every index choice by a power of $r$; if that count omits or double-counts any lifting, the Martin limits and hence the classification collapse.","fun_headline_variants_meta":{"raw":{"variants":["Martin boundary of r-YF graph: all measures ergodic","Every Martin measure on r-YF paths is ergodic","r-YF boundary: explicit measures, all ergodic","Explicit Martin boundary for r-YF: all measures ergodic","r-differential YF: boundary measures all ergodic"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001079,"raw_usage":{"total_tokens":4466,"prompt_tokens":847,"completion_tokens":3619,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":463,"completion_tokens_details":{"reasoning_tokens":3535}},"tokens_in":463,"tokens_out":3619,"duration_ms":22936,"temperature":1.0,"reasoning_tokens":3535,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T13:56:42.293089+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute both sides of Theorem 2 for a small pair in $YF_2$, for example $w=1_1$ and $v=1_1 2 1_2$, by listing all downward paths directly; any disagreement falsifies the formula. Equivalently, evaluate the Martin ratio $d_2(\\varepsilon,w)d_2(w,v_n)/d_2(\\varepsilon,v_n)$ along a sequence $v_n$ converging to a boundary word with $\\pi(v_n)\\to 0$ and check that the limit equals $\\mu_P(s(w))/r^{e(w)}$ for every $w$; one counterexample overturns the classification.","supporting_citations":[{"cited_title":"бесконечно удалённую вершину","cited_arxiv_id":null,"evidence_quote":"Supplies the Martin-boundary theorem and the asymptotic proposition for the ordinary Young–Fibonacci graph that the $r$-differential proof reduces to."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the path-count formula $d_1(w,v)$ for the Young–Fibonacci graph, quoted as Theorem 1 and used inside Theorem 2."},{"cited_title":"бесконечно удалённой вершине","cited_arxiv_id":null,"evidence_quote":"Provides the extended version of the same Young–Fibonacci path-count formula, cited together with [5] as the base of Theorem 2."},{"cited_title":"Тогда ∀a ∈ 1, d(v), i∈ N : lim n→∞ d(vn)Y j=a g (vn, j) − i g (vn, j) = βi · d(v)Y j=a g (v, j) − i g (v, j)","cited_arxiv_id":null,"evidence_quote":"Proves ergodicity of the Plancherel measure on the Young–Fibonacci graph, whose argument Theorem 5 transfers to $\\mu_{r,P}$."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the vanishing-sum estimates (Corollary 10 and Corollary 11) used in Theorems 6, 8, and the ergodicity proof of Theorem 10."}],"review_version":1}