{"id":"612a0aed-9db1-4150-86e3-635649786c3c","arxiv_id":"1908.03714","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A new list of seven graph moves is conjectured to generate all refined isomorphism relations among unital graph C*-algebras, with full proofs in two cases and partial gauge-simple results.","lead":"This paper proposes seven local graph moves and conjectures that, for six refined notions of isomorphism of graph C*-algebras, two graphs are equivalent exactly when one can be transformed into the other by moves preserving that notion. It proves the conjecture in two cases, gives partial results for gauge-simple algebras, and establishes which moves preserve which structure.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 100-invariance of (C+) and (P+) and the full 100-generation claim rest on the unproven classification theorem [ERRS, Theorem 3.5], making the central conjecture vulnerable to a gap in that external result.","rationale":"The reader's weakest assumption identifies exactly the same load-bearing point: the advanced moves (C+) and (P+) are proved 100-invariant only by appealing to the substantial self-cited classification theorem [ERRS, Theorem 3.5], and the full-generation theorems for 000 and 100 are outsourced to [ERRS] and [AER]. My independent reading confirms that the fifteen-page preprint establishes invariance for the elementary moves in detail, and the gauge-simple generation results in Section 6 give genuine evidence for the conjecture, but the two central full-generality claims are not self-contained. The concern is not an internal inconsistency but an external dependency: if the classification theorem is incorrect or incomplete, then the 100-invariance of (C+) and (P+) (Theorems 3.14 and 3.16) fails, and with it the 100 direction of Conjecture 5.1. The proposed test is a targeted, minimal check of the classification input on the concrete graph pair already featured in the paper, which would expose a failure at the critical step. Since the reader already conditioned the verdict on exactly this dependency, no change to the verdict is needed.","tokens_in":115,"tokens_out":2015,"duration_ms":87413,"concrete_test":"Verify the completeness statement [ERRS, Theorem 3.5] on the smallest nontrivial pair where the advanced moves are used: apply (C+) to the graph G(2,0) from Example 4.8 to obtain G, and compute the full ordered reduced filtered K-theory of both C*(G(2,0)) and C*(G) with the distinguished unit class. Check whether every invariant isomorphism sends [1] to [1] and can be lifted to a ∗-isomorphism, as required by the theorem; if a lifting fails for this pair, the proof of Theorem 3.14 collapses and the 100-generation claim is unsupported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central conjecture (Conjecture 5.1) asserts 100 = ⟨(O),(I+),(R+),(C+),(P+)⟩. The right-to-left inclusion is supported by Theorems 3.14 and 3.16, whose proofs terminate with 'By [ERRS, Theorem 3.5], C*(E_C+) ≅ C*(E)' and the analogous statement for (P+). That theorem is a substantial classification result for unital graph C*-algebras by reduced filtered K-theory, submitted for publication and not reproduced here. If the classification has a gap or a missing hypothesis, the 100-invariance of (C+) and (P+) is not established, and the 100-generation theorem (Theorem 5.3), which is deferred to the companion paper [AER], loses its stated support. Similarly, Theorem 5.2 for 000-generation is cited from [ERRS], so two of the six full-generation claims are not self-contained. The abstract's phrase 'in two of the six cases, we may prove the conjecture in full generality' overstates what this paper proves on its own. No internal error is apparent in the move definitions or the invariance proofs for (O), (I+), (I-), (R+), and (S); the load-bearing risk is concentrated at the two points where the argument invokes unpublished classification input.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper formalizes eight notions of isomorphism among unital graph C*-algebras encoded by three-bit words, where the bits record exact versus stable isomorphism, gauge-action equivariance, and diagonal preservation. It proposes seven graph moves and conjectures that, in six of the eight cases, the appropriate equivalence relation is generated by the moves respecting that structure. The authors prove invariance of each move, with Theorems 3.2, 3.4, 3.5, 3.7, 3.10, 3.12, 3.14, and 3.16 covering the seven moves, and they develop standard-form reductions that establish the conjecture for gauge-simple graph C*-algebras in substantial generality, including a full treatment of the unicore case. The two statements of full generation in the conjecture are not proved here: Theorem 5.2 is cited from [ERRS] and Theorem 5.3 is deferred to the companion paper [AER].","tokens_in":44159,"tokens_out":3380,"duration_ms":39742,"significance":"If the conjecture and its supporting theorems are correct, the paper provides a genuine geometric, move-based description of several operator-algebraic equivalence relations, connecting Williams' symbolic-dynamics moves with C*-classification and with rigidity results for Cuntz-Krieger algebras. The paper's own contributions include a useful nomenclature, careful invariance proofs for the elementary moves, explicit move-realizations of many equivalences, and a thorough treatment of the gauge-simple and unicore cases. The significance is partly conditional, however, because the 100-invariance of the advanced moves and two of the six full-generation claims rest on external, not-yet-published results.","major_comments":[{"comment":"The proofs of Theorem 3.14 and Theorem 3.16 terminate by invoking [ERRS, Theorem 3.5], a classification theorem for unital graph C*-algebras by reduced filtered K-theory that is only submitted and not reproduced here. Since the right-to-left inclusion of the 100 part of Conjecture 5.1 is directly based on these two theorems, a gap or missing hypothesis in [ERRS, Theorem 3.5] would invalidate the claimed invariance of (C+) and (P+) and hence the supporting evidence for 100-generation. The manuscript should state the classification theorem and its precise standing, or supply a proof or an independent verification of the two isomorphisms.","section":"§3.14, §3.16, and [ERRS, Theorem 3.5]"},{"comment":"The full-generation claims in two cases are not proved in this paper: Theorem 5.2 is quoted from [ERRS] and Theorem 5.3 is deferred to the companion paper [AER]. The abstract's assertion that 'in two of the six cases, we may prove the conjecture in full generality' therefore overstates what the present manuscript establishes on its own. Please either include the relevant proofs, provide the exact theorem statements and the standing of the external references, or reword the claims so that the reader knows which parts are proved here.","section":"§5, Theorem 5.2, Theorem 5.3, and Abstract"},{"comment":"The 111-generation result for finite graphs with no sinks is referenced as [Bri], which is listed as a private communication. This is another load-bearing external input for the table of generation results and for the 111 column of the gauge-simple summary. The dependence should be made explicit and the result should be upgraded to a verifiable preprint or stated with proof if it is needed for the paper's conclusions.","section":"§6, Theorem 6.1"}],"minor_comments":[{"comment":"The notation '10z' and '11z' in the proof of Theorem 6.12, Case IV, is inconsistent with the paper's three-bit convention; these should be '100'/'101' and '110'/'111', or explicitly defined at first use.","section":"§6.2, Theorem 6.12, Case IV"},{"comment":"The '−' entries in the Invariance and Non-invariance columns are not explained in the caption; a sentence stating that '−' means no non-invariance example is claimed would help.","section":"Figure 2"},{"comment":"Several references are listed as 'In preparation' or 'Private communication' ([AER], [Bri], [ERRS]), which makes verification difficult; please replace these with stable references, preprints, or an appendix stating the needed results.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The paper is a well-written programmatic contribution, but its central 100-invariance theorem and two of the six full-generation claims depend on external submitted or in-preparation work. Editorial judgement is needed on whether deferring such load-bearing results to companions is acceptable. I recommend requesting the authors to state the external theorems explicitly and clarify which claims are proved in this paper versus in [ERRS] and [AER]."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, this paper gives a clean 3-bit nomenclature for eight isomorphism notions and proves which of seven local moves preserve which structure. Second, the main conjecture—that refined equivalences are generated by the structure-preserving moves—is genuinely new, but this preprint only proves part of it; the rest is outsourced to [ERRS] and [AER].\n\nWhere credit is due: the move framework is useful. The (I+) move, a rethinking of in-splitting that allows empty partition cells, does real work, and the invariance proofs in Section 3, especially for (O), (I-), (I+), (R+), and (S), are detailed and appear careful. The partial generation results for gauge-simple graphs, including the unicore case in Section 6.2, are nontrivial and well-supported by explicit reductions. The paper does not claim more than it proves at the level of individual theorems—the caveats about the conjecture being open are stated.\n\nSoft spots: the two full-generation theorems are not proved here. Theorem 5.2 cites [ERRS]; Theorem 5.3 is deferred to [AER]. More specifically, the 100-invariance of (C+) and (P+) (Theorems 3.14 and 3.16) ends with an appeal to [ERRS, Theorem 3.5], a classification result that is submitted for publication and not reproduced. If that classification has a gap, the 100 column loses its support. The abstract's phrase 'in two of the six cases, we may prove the conjecture in full generality' overstates what this paper contains on its own; one of those is a companion paper. That is worth fixing before publication, but it is a wording issue, not a mathematical error.\n\nWho this is for: anyone working on graph C*-algebras, Cuntz-Krieger rigidity, or shift equivalence. It is a good research announcement, and the examples comparing the eight notions are valuable. I would send it to a serious referee. The dependence on unpublished work is a reason for caution, not for rejection; many classification papers are like this.\n\nRecommendation: engage with it. Make sure the companion papers are visible to the referee, and ask the authors to sharpen the abstract.","headline":"A careful research announcement that reframes six isomorphism notions for graph C*-algebras as move-generated equivalences; the invariance proofs are solid, but two of the six full-generation claims rest on unpublished companion work.","tokens_in":44739,"tokens_out":1893,"would_cite":true,"duration_ms":19608,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["46L05","46L35","46L55","37B10"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper's project is to show that six refined isomorphism notions among graph C*-algebras are each generated by a short list of local graph moves, and to prove as much of that as possible.","keywords":["graph C*-algebras","Cuntz-Krieger algebras","gauge action","diagonal-preserving isomorphism","shift equivalence","symbolic dynamics","local moves","classification"],"falsifier":"Compute the move-generated equivalence relation from $(O)$ and $(I+)$ on all finite graphs with, say, at most four vertices and compare it with $111$-equivalence as detected by the pointed ordered K-theory data of $C^*(E\\times_1\\mathbb{Z})$ described in Corollary 2.5; any pair that the invariants declare $111$-equivalent but that no sequence of $(O)$ and $(I+)$ moves connects would disprove the $111$ part of Conjecture 5.1.","tokens_in":43652,"feed_emoji":"🔄","tokens_out":17123,"duration_ms":159887,"temperature":0.7,"pith_summary":"This paper asks whether the many recently distinguished notions of sameness for $C^*$-algebras built from directed graphs can be understood as purely geometric operations on the graphs themselves. The authors encode eight isomorphism notions by three-bit words $xyz$: the first bit says whether the isomorphism is exact or only stable, the second whether it preserves the gauge action, and the third whether it sends the diagonal to the diagonal. They propose seven local moves and conjecture that in six of these settings the equivalence relation is generated by those moves that respect the relevant structure; they prove the conjecture in two cases in full generality, give gauge-simple results in the other four, and solve completely the case where the irreducible core is a single vertex. If the conjecture holds, questions about algebraic isomorphism reduce to watching which local rewrites are allowed, in the same spirit that Reidemeister moves describe knot equivalence.","feed_headline":"Six isomorphism notions may reduce to seven local graph moves","feed_subtitle":"A 3-bit code separates exact vs stable, gauge, and diagonal; two cases are already proved.","key_machinery":"The load-bearing objects are the seven moves themselves, and the genuinely new one is $(I+)$, unital in-splitting: when two or more vertices have exactly the same future, $(I+)$ redistributes their pasts among them, and Corollary 3.7 shows that this preserves the unit, the gauge action, and the diagonal. The second engine is the reduced filtered K-theory classification of unital graph $C^*$-algebras; to prove that $(C+)$ and $(P+)$ are 100-invariant, the paper writes explicit unimodular matrices $U,V$ that transform the defining matrix of the altered graph into that of the original, including the class of the unit, and then invokes classification. For distinctions among the $1yz$ notions, the machinery is the fixed-point algebra of the gauge action, realized as $C^*(E\\times_1\\mathbb{Z})$ with a canonical translation, whose pointed ordered K-theory records exactly what a diagonal- and gauge-preserving isomorphism must preserve.","core_discovery":"The central claim, stated as Conjecture 5.1, is that the six equivalence relations are generated as follows: $000=\\langle(O),(I-),(R+),(S),(C+),(P+)\\rangle$, $001=\\langle(O),(I-),(R+),(S)\\rangle$, $011=\\langle(O),(I-)\\rangle$, $100=\\langle(O),(I+),(R+),(C+),(P+)\\rangle$, $101=\\langle(O),(I+),(R+)\\rangle$, and $111=\\langle(O),(I+)\\rangle$. The paper's original contribution is the refined move list, especially the unital in-splitting $(I+)$ defined by redistributing the pasts of vertices with identical futures, together with a systematic proof of which moves preserve which structure. It records the full generation statement in the 000 case from earlier work, cites the companion paper for the full 100 case, and proves here the invariance of the two advanced moves $(C+)$ and $(P+)$ by reducing them to matrix identities in reduced filtered K-theory and invoking classification. The remaining gauge-simple cases are handled by reducing graphs to a small standard form using only the relevant moves.","pith_inferences":["The same 'which moves preserve the structure?' question applies to the two remaining corners, 010 and 110; the paper's Remark 5.4 shows the current move list cannot generate 010, so solving that case will require a new move of a more arithmetic character.","If the reduced filtered K-theory classification used for (C+) and (P+) is later extended to larger graph classes, the explicit matrix identities in Theorems 3.14 and 3.16 would automatically extend the 100-invariance and the 100 generation theorem.","The (I+) move is essentially a one-sided analogue of classical in-splitting, so the 111 and 101 generation results suggest a parallel move description for one-sided shifts of finite type, a direction noted in the paper as being explored elsewhere.","A practical testable extension is to implement the standard-form lemmas computationally for all finite graphs up to a fixed size and compare the move-generated relations with the K-theoretic invariants from Corollary 2.5; any mismatch would localize where the conjecture fails."],"forward_implications":["In the classes where the 011 case is proved, 011-equivalence is exactly the relation generated by out-splitting and in-splitting, recovering a geometric characterization of conjugacy of shifts of finite type.","If Conjecture 5.1 holds, exact isomorphism of unital graph C*-algebras is generated by the five moves (O), (I+), (R+), (C+), and (P+), so every exact isomorphism can be exhibited as a finite sequence of local graph changes.","The strongest notion, 111, is conjectured to be generated by just (O) and (I+); for finite graphs with no sinks, Theorem 6.1 already proves this generation within that class.","The paper establishes the conjecture for all graphs defining gauge-simple C*-algebras in the 001 and 101 cases, and for the 011 and 111 cases whenever the graph is finite or has at most one vertex allowing a path back to itself.","When the irreducible core is a single vertex with c loops, the paper's unicore theorem gives explicit move recipes implementing every xyz-equivalence for c = 0,1,2,...,∞."],"supporting_citations":[{"why":"Supplies the classification theorem by reduced filtered K-theory used to prove the 100-invariance of (C+) and (P+), and the earlier result that 000 is generated by the older move list.","marker":"[ERRS]"},{"why":"Companion paper cited for the full 100 generation theorem, that 100 equals the move-generated relation.","marker":"[AER]"},{"why":"Shows diagonal-preserving gauge-invariant isomorphisms of stabilized graph C*-algebras correspond to conjugacy of two-sided shifts of finite type, the key input for the 011 case.","marker":"[CR17]"},{"why":"The classical theorem that in- and out-splittings generate conjugacy of two-sided shifts of finite type, basis for the 011 generation statement.","marker":"[Wil73]"},{"why":"Introduced out-splitting and in-splitting moves for graph C*-algebras; provides the original (O) move and the version of in-splitting that (I-) refines.","marker":"[BP04]"},{"why":"Gives the flow-equivalence rigidity result used to identify 001-equivalence with flow equivalence of the associated shifts for the covered finite essential cases.","marker":"[MM14]"},{"why":"Provides matrix factorizations in SL(2,N) used in the proof that 10z-equivalence of the graphs G(c,n) is generated by (O), (I+), (R+).","marker":"[Hua94]"},{"why":"Source of the polycephaly case analysis adapted for the paper's complete solution of xyz-equivalence when the core is a single vertex.","marker":"[Haz13]"},{"why":"Identifies the fixed-point algebra of the gauge action as a corner of the skew graph algebra, giving the K-theoretic data used to distinguish 11z-equivalent graphs.","marker":"[Cri08]"}],"fun_headline_variants":["Six isomorphism notions from seven graph moves?","Seven moves may generate six C*-algebra isomorphisms","Refined moves could unify graph C*-algebra equivalences","Partial proof: six isomorphisms via seven local moves","Graph C*-algebras: seven moves for six notions?"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument for the 100 case rests on a classification theorem taken from a companion paper, which is cited but not proved here; if that classification result is wrong, the claimed invariance of the two advanced moves collapses and with it the full 100 generation statement.","fun_headline_variants_meta":{"raw":{"variants":["Six isomorphism notions from seven graph moves?","Seven moves may generate six C*-algebra isomorphisms","Refined moves could unify graph C*-algebra equivalences","Partial proof: six isomorphisms via seven local moves","Graph C*-algebras: seven moves for six notions?"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000165,"raw_usage":{"total_tokens":1259,"prompt_tokens":964,"completion_tokens":295,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":580,"completion_tokens_details":{"reasoning_tokens":216}},"tokens_in":580,"tokens_out":295,"duration_ms":3659,"temperature":1.0,"reasoning_tokens":216,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:04:12.825346+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the move-generated equivalence relation from $(O)$ and $(I+)$ on all finite graphs with, say, at most four vertices and compare it with $111$-equivalence as detected by the pointed ordered K-theory data of $C^*(E\\times_1\\mathbb{Z})$ described in Corollary 2.5; any pair that the invariants declare $111$-equivalent but that no sequence of $(O)$ and $(I+)$ moves connects would disprove the $111$ part of Conjecture 5.1.","supporting_citations":[],"review_version":1}