{"id":"b5da8f29-0e1b-4637-b854-e8d3c3272ccb","arxiv_id":"1908.02731","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The class I[D[I]] is composable but not splittable, answering Karpilovskij's question, and every infinite composable class avoiding an increasing or decreasing permutation is splittable.","lead":"This paper shows that some permutation classes can be broken into compositions of smaller classes even though they cannot be split by a two-color merge. It answers an open question about the relationship between two structural notions for permutation classes.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 3.1's unsplittability proof omits the bounded-length argument that makes 'no proper infinite subclasses' imply unsplittability of I and D; without it, the central claim is not fully established.","rationale":"The central claim has two parts: composability and unsplittability. The composability proof is sound: every element of C is a direct sum of blocks in D[I], and the explicitly constructed α and β are layered permutations with π=α∘β, so C⊆L∘L with L a proper subclass. The load-bearing gap is in the unsplittability half: the one-line justification for unsplittability of I and D omits the bounded-length argument, and that argument is essential before Lemma 2.3 can be applied. The gap is fixable and the assertion is true, so the concern lands as an incomplete proof rather than a false theorem. The reader's additional complaints about Lemma 3.2 and Theorem 3.3 concern secondary results and do not affect Theorem 3.1; notably, the 'if and only if' in Lemma 3.2 is indeed stronger than needed, but only the forward direction is used. Because the main construction appears correct and the missing justification is readily supplied, the conditional verdict remains appropriate.","tokens_in":5963,"tokens_out":36193,"duration_ms":381572,"concrete_test":"Perform the omitted verification for I: let A⊆I be proper, choose N with ι_N∉A; by heredity every element of A has length <N. For A,B with maxima a,b, prove every element of A⊙B has length ≤a+b, so ι_n∉A⊙B for n>a+b, contradicting I⊆A⊙B. Repeat for D. If this check succeeds, unsplittability of I and D (and hence Theorem 3.1) is established; if it fails, the central counterexample is not proven.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 3.1 (first paragraph) asserts: 'there are no proper infinite subclasses, so I and D are unsplittable.' This inference is not valid as stated. Splittability only requires two proper subclasses, and a finite class with proper subclasses can be splittable (e.g., {1,12,21} splits via Av(12) and Av(21)). The missing step is that every proper subclass of I, being hereditary, omits some ι_N and therefore contains only increasing permutations of length <N; likewise for D. Hence any two proper subclasses have bounded maximum lengths a,b, and any merge has length at most a+b, so arbitrarily long increasing (or decreasing) permutations cannot be covered. This step is the load-bearing bridge from finiteness of proper subclasses to unsplittability of I and D, and then via Lemma 2.3 to unsplittability of C=I[D[I]]. The assertion is true and the repair is straightforward, but the proof as written is incomplete.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies two decomposability notions for permutation classes: splittability (red-blue coloring into two proper hereditary subclasses) and composability (factorization under permutation composition into proper subclasses). The main result, Theorem 3.1, exhibits the class C = I[D[I]] and proves that it is composable but not splittable, thereby answering a question of Karpilovskij. The secondary result, Theorem 3.3, states that every infinite composable class avoiding an increasing or a decreasing permutation is splittable. The paper also introduces the notion of exact-splittability and poses several open questions.","tokens_in":6129,"tokens_out":22323,"duration_ms":206982,"significance":"If correct, Theorem 3.1 is the first example separating composability from splittability, a natural distinction in the area, and the construction via inflations is elegant. Theorem 3.3 gives a useful sufficient condition under which the two notions coincide. The paper is self-contained and does not rely on any numerical fitting; the core arguments use standard tools such as the Erdős–Szekeres theorem and known results on unsplittable classes. I regard the results as significant for the permutation-class community.","major_comments":[{"comment":"The inference 'there are no proper infinite subclasses, so I and D are unsplittable' is not valid as stated; finiteness of all proper subclasses does not by itself rule out a split into two finite proper subclasses (for example, the finite class {1,12,21} is splittable via {1,12} and {1,21}). The missing step is that a proper subclass of I omits some ι_N and hence has bounded length, so a merge of two such subclasses has bounded length and cannot contain ι_N for arbitrarily large N; the analogous argument holds for D. Since Lemma 2.3 is then used to transfer unsplittability to C=I[D[I]], this step is load-bearing and should be written out.","section":"Section 3.1, Theorem 3.1"},{"comment":"The proof of the bound A∘B⊆I_{kℓ} contains a false equivalence. After defining α′ and β′, the text states that for u<v, β′_u<β′_v if and only if α′_{β′_u}>α′_{β′_v}. The reverse direction is false: for α=132, β=21, γ=α∘β=31, the decreasing subsequence at positions 1,2 gives β′=21 and α′=12 (with the natural sorted restriction), so β′_1<β′_2 is false while α′_{β′_1}>α′_{β′_2} is true. What the proof needs is only the forward implication, which yields that any increasing subsequence of β′ gives a decreasing subsequence of α of the same length; together with the fact that any decreasing subsequence of β′ is a decreasing subsequence of β, Erdős–Szekeres gives m≤kℓ. The definitions of α′ and β′ should also be spelled out, since the equality δ_m=α′∘β′ depends on taking α′ from the sorted indices rather than from the listed order.","section":"Section 3.2, Lemma 3.2"},{"comment":"The step 'C⊆A1∘I_{m^{k-1}} implies C is the merge of m^{k-1} copies of A1' relies on the asserted fact that composing a permutation with an element of I_n is equivalent to de-merging it into at most n subpermutations. This fact is stated without proof and the referenced figure is missing. A proof is necessary: if π=α∘η with η∈I_n, partition the positions of η into at most n increasing subsequences (possible because η has no decreasing subsequence of length n+1); each color class induces a subsequence of π that is a subpermutation of α and hence lies in A1. Without this argument, the splittability conclusion of Theorem 3.3 does not follow.","section":"Section 3.3, Theorem 3.3"}],"minor_comments":[{"comment":"The name 'Steinrímsson' should be 'Steingrímsson'.","section":"Section 1"},{"comment":"The phrase 'a fixed set set of permutations' contains a duplicated word.","section":"Section 2.1"},{"comment":"The notation π^r_i in the composability proof should be defined explicitly as the reversal of the block π_i, rather than left ambiguous with the reversal of the whole permutation.","section":"Section 3.1, Theorem 3.1"},{"comment":"The quantity 'δmk+1' should be typeset as δ_{m^k+1}, and the superscripts in I_{m^{k-1}} should be rendered unambiguously.","section":"Section 3.3"},{"comment":"The reference 'see Figure??' should be resolved: either the figure should be included or the reference removed.","section":"Section 3.3"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe headline: Zhang gives a genuine answer to Karpilovskij's question—there is a composable permutation class that is not splittable, namely C = I[D[I]]. The construction is neat and the composability proof is explicit: every element is a composition of two layered permutations. I checked the inflation argument and it works. If you work on permutation classes, this is the result to know from this paper.\n\nWhat's new: the example itself, and the sufficient condition in Theorem 3.3. The tools are standard (inflation, Lemma 2.3, Erdős–Szekeres), but the construction is not in the cited literature. The paper is honest about what it builds on.\n\nSoft spots, in order of severity:\n\n1. The unsplittability proof for I and D in Theorem 3.1 is too terse. The sentence 'there are no proper infinite subclasses, so I and D are unsplittable' skips the bounded-length argument: any proper subclass of I is finite, so a merge of two proper subclasses has bounded length and cannot cover arbitrarily long increasing permutations. The assertion is true, but as written the proof is incomplete. This is a one-sentence fix.\n\n2. Lemma 3.2's proof states an 'if and only if' that is false; only the forward direction is needed and used. The lemma itself is fine, but the exposition should be corrected.\n\n3. Theorem 3.3 relies on an unproved 'de-merge' claim about composition with an element of I_n, with a missing figure. The claim is standard and likely correct, but it needs a proof or a citation, not just a pointer to a non-existent figure.\n\nThe secondary theorem's proof has these gaps, but the main construction is solid. The flaws are all fixable, none is load-bearing. Citation pattern is clean: cites [7] and [8] appropriately, no self-citation, no fitted parameters.\n\nWho this is for: permutation-pattern specialists, and anyone interested in split/merge vs composition. It deserves a serious referee; a good referee would ask for the fixes above but should not reject the result. I'd send it to review and expect a revised version to be citable.\n\nRecommendation: engage with it; it answers the open question.","headline":"A genuine answer to Karpilovskij's question, with a clean construction; the main result holds up, but several proofs are too terse and need repair.","tokens_in":6583,"tokens_out":11222,"would_cite":true,"duration_ms":103668,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["05A05","05D10"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper shows that the permutation class $I[D[I]]$ is composable but unsplittable, settling an open question about the relation between the two properties.","keywords":["permutation classes","splittability","composability","inflation","layered permutations","pattern avoidance","merges of permutations","subsequence bounds"],"falsifier":"Show a split of $C=I[D[I]]$: find proper subclasses $A,B\\subsetneq C$ such that every element of $C$ can be colored red and blue with the red subsequence in $A$ and the blue subsequence in $B$. Equivalently, by the paper's Proposition 2.1, find patterns $\\pi,\\pi'\\in C$ such that every element of $C$ has a two-coloring with no red $\\pi$ and no blue $\\pi'$. Finding such a split would disprove Theorem 3.1.","tokens_in":5748,"feed_emoji":"🔀","tokens_out":11279,"duration_ms":113643,"temperature":0.7,"pith_summary":"Permutation classes are hereditary sets of permutations. This paper compares two ways of decomposing one: a class is splittable if every permutation can be colored red and blue so that the red and blue subsequences lie in two proper subclasses, and composable if every permutation is a composition of permutations from finitely many proper subclasses. The paper's main claim is that composability does not force splittability: the class $C=I[D[I]]$ — inflations of increasing permutations by blocks from the class $D[I]$ — is composable but unsplittable. This supplies the missing combination in a classification of composable/splittable classes and answers the open question that motivated the paper. The same paper proves that an infinite composable class that avoids either an increasing or a decreasing permutation is always splittable, so a composable unsplittable class must contain arbitrarily long increasing and decreasing permutations.","feed_headline":"A permutation class can be composable yet unsplittable","feed_subtitle":"Built from increasing and decreasing blocks, the class settles the missing case in the composability-splittability question.","key_machinery":"The mechanism that carries the argument is the inflation operation, in which each point of one permutation is replaced by a block from another permutation class, together with the lemma that the inflation of unsplittable classes is unsplittable. The class $C = I[D[I]]$ is obtained by inflating an increasing permutation by blocks from $D[I]$, and the lemma transfers unsplittability from $I$ and $D$ up to $C$. For composability, the direct-sum structure of $C$ is exploited: reversing each block of an element turns it into a layered permutation, and composing with a direct sum of decreasing permutations of matching block lengths produces the original element, so $C$ is contained in $L \\circ L$ with $L = I[D]$ the layered class. For the positive splittability theorem, a composition lemma bounds the length of decreasing subsequences in a composition of two pattern-avoiding classes, which lets a $k$-fold composition be rewritten as a merge of copies of a single proper subclass.","core_discovery":"The paper's central discovery is that composability does not imply splittability: the class $C=I[D[I]]$ is composable and unsplittable. The classes $I$ and $D$ of all increasing and all decreasing permutations are unsplittable because each has exactly one permutation of each length, and the inflation lemma passes unsplittability to $C$. At the same time, every element of $C$ is a composition of two layered permutations, so $C$ lies inside $L\\circ L$ for a proper subclass $L$; hence $C$ is composable. The paper also proves a partial converse: an infinite composable class that avoids some increasing or decreasing permutation must be splittable, so the new example necessarily contains arbitrarily long increasing and decreasing permutations.","pith_inferences":["A natural next test is whether other inflations, such as $D[I[D]]$ or iterated inflations $I[D[I[D]]]$, remain composable while unsplittable; the same inflation lemma would preserve unsplittability if the base classes do.","The proof of Theorem 3.3 separates even from odd numbers of composition factors; this suggests that composability by an odd number of proper subclasses may be the only route to novel unsplittable examples, and questions about $k$-composability for odd $k$ are where new obstructions would appear.","Because $C$ is written as $L\\circ L$ with $L$ a proper subclass, ordinary splittability is not needed to cover it; this example could be used to compare other decomposition notions, such as exact-splittability, on a class that is unsplittable in the merge sense."],"forward_implications":["The two notions of decomposability are independent: each of the four combinations of composable/uncomposable and splittable/unsplittable is now realized by some permutation class.","Every infinite composable class that avoids some finite increasing or decreasing pattern is splittable, so any further composable unsplittable examples must contain arbitrarily long increasing and decreasing permutations.","The class $I[D[I]]$ is covered by two copies of the layered class under composition, giving a concrete class whose members have a two-factor factorization by layered permutations.","The inflation lemma provides a reusable way to build unsplittable classes from unsplittable components, so classes of the form $A[B]$ with $A,B$ unsplittable are immediately unsplittable."],"supporting_citations":[{"why":"Gives the criterion that splittability is equivalent to the existence of two forbidden patterns that every coloring must avoid, which the unsplittability proof uses.","marker":"[7]"},{"why":"Proves the lemma that inflating unsplittable permutation classes yields an unsplittable class; this is the step that transfers unsplittability from $I$ and $D$ to $I[D[I]]$.","marker":"[2]"},{"why":"Introduces composability of permutation classes, studies combinations of composability and splittability, and poses the open question answered here.","marker":"[8]"},{"why":"Supplies the standard length bound for increasing and decreasing subsequences used in Lemma 3.2, which converts compositions into merges in Theorem 3.3.","marker":"[5]"}],"fun_headline_variants":["Composable but not splittable: a new permutation class","Permutation class that is composable and unsplittable","Counterexample: composable permutation classes need not split","Answering Karpilovskij: composable does not imply splittable","Infinite composable classes: when must they split?"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The unsplittability of $I$ and $D$, and hence of $C$, rests on the fact that these classes have exactly one permutation of each length, so every proper subclass has bounded length and two proper subclasses cannot together cover arbitrarily long increasing or decreasing permutations; this fact is used without proof.","fun_headline_variants_meta":{"raw":{"variants":["Composable but not splittable: a new permutation class","Permutation class that is composable and unsplittable","Counterexample: composable permutation classes need not split","Answering Karpilovskij: composable does not imply splittable","Infinite composable classes: when must they split?"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000228,"raw_usage":{"total_tokens":1439,"prompt_tokens":870,"completion_tokens":569,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":486,"completion_tokens_details":{"reasoning_tokens":483}},"tokens_in":486,"tokens_out":569,"duration_ms":5745,"temperature":1.0,"reasoning_tokens":483,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:39:21.149641+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Show a split of $C=I[D[I]]$: find proper subclasses $A,B\\subsetneq C$ such that every element of $C$ can be colored red and blue with the red subsequence in $A$ and the blue subsequence in $B$. Equivalently, by the paper's Proposition 2.1, find patterns $\\pi,\\pi'\\in C$ such that every element of $C$ has a two-coloring with no red $\\pi$ and no blue $\\pi'$. Finding such a split would disprove Theorem 3.1.","supporting_citations":[{"cited_title":"Jelínek & P","cited_arxiv_id":null,"evidence_quote":"Gives the criterion that splittability is equivalent to the existence of two forbidden patterns that every coloring must avoid, which the unsplittability proof uses."},{"cited_title":"Albert & V","cited_arxiv_id":null,"evidence_quote":"Proves the lemma that inflating unsplittable permutation classes yields an unsplittable class; this is the step that transfers unsplittability from $I$ and $D$ to $I[D[I]]$."},{"cited_title":"Karpilovskij","cited_arxiv_id":null,"evidence_quote":"Introduces composability of permutation classes, studies combinations of composability and splittability, and poses the open question answered here."},{"cited_title":"Erdős & G","cited_arxiv_id":null,"evidence_quote":"Supplies the standard length bound for increasing and decreasing subsequences used in Lemma 3.2, which converts compositions into merges in Theorem 3.3."}],"review_version":1}