{"id":"b9ae05bb-de25-4b9f-bcf6-111120a82f63","arxiv_id":"2411.17825","paper_version":2,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Every K-Lipschitz function from a closed subset of a metric space into any real interval extends to a K-Lipschitz function on the whole space, and analogous extension, decomposition, and selection results hold for locally Lipschitz functions.","lead":"This paper refines the classical McShane-Whitney Lipschitz extension theorem so that extensions into any real interval are possible, and extends the result to locally Lipschitz functions. It also proves that countable open covers of metric spaces admit Lipschitz partitions of unity without using Stone's theorem, yielding new locally Lipschitz selection theorems.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified","rationale":"The reader's ACCEPT verdict is well-founded. The proof of Theorem 1.3 is complete and correct; the closedness assumption is necessary, explicitly stated, and the paper demonstrates failure without it. The subsequent sections rely on Theorem 1.3 and the partition-of-unity results, whose proofs also appear sound. Minor typographical issues do not affect correctness. Therefore no adjustment to the verdict is needed.","tokens_in":18663,"tokens_out":7370,"duration_ms":59647,"concrete_test":"Independently re-derive inequality (3.1) from definitions (1.1) and verify that it forces the averaged extension in Proposition 3.2 to satisfy f(p)∈(a,b) for every p∈X\\A when A is closed, using only d(p,A)>0.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim, Theorem 1.3, is proved correctly. In Proposition 3.2, the K-Lipschitz McShane–Whitney extensions Φ− and Φ+ are clipped to Ψ−=max(Φ−,a) and Ψ+=min(Φ+,b), and their average f is taken. For p∈X\\A, closedness of A gives d(p,A)>0, so inequality (3.1) implies Ψ−(p)<b and Ψ+(p)>a, hence f(p)∈(a,b); for p∈A, f(p)=φ(p)∈Δ. Clipping and averaging preserve the K-Lipschitz property. Unbounded interval cases are handled similarly via Proposition 3.1. The paper explicitly notes that the closedness assumption cannot be dropped (identity on an open interval), so no hidden gap exists. The later local and selection theorems build on this solid foundation.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper refines the classical McShane-Whitney extension theorem by showing that a K-Lipschitz function from a closed subset A of a metric space into an arbitrary interval Δ ⊂ R can be extended to a K-Lipschitz function on all of X with values in Δ (Theorem 1.3). The proof (Section 3) clips and averages the two McShane-Whitney extensions, using closedness exactly to obtain strict containment in the open interval. The result is extended to locally pointwise Lipschitz functions (Theorem 4.1) and to locally Lipschitz functions (Theorem 7.1), the latter via a direct construction of locally finite Lipschitz partitions of unity for countable open covers (Theorem 5.1), avoiding Stone's paracompactness theorem. Applications include a decomposition of locally Lipschitz functions into locally finite sums of bounded Lipschitz functions (Theorem 6.1), a characterization of local Lipschitzness through a continuous Lipschitz-constant function (Theorem 6.5), and a locally Lipschitz version of Michael's selection theorem (Theorems 7.2 and 7.4).","tokens_in":109,"tokens_out":29626,"duration_ms":479013,"significance":"The central interval-extension theorem is a clean and useful refinement of a classical result, and the direct proof of the countable case of Frolík's Lipschitz partition-of-unity theorem is a genuine technical contribution. The later selection and decomposition results are natural and well motivated. The proofs are largely self-contained and carefully written; the paper's historical remarks and references are a strength. If the two proof gaps noted below are repaired, the paper would be a solid addition to the literature.","major_comments":[{"comment":"The step that sets A_n = A ∩ supp(ξ_n) and then applies Theorem 1.3 is not justified as written. In Section 5 the support is defined as coz(ξ) = {x : ξ(x) ≠ 0}, which is an open set. Hence A_n is generally not closed in X, and the hypothesis of Theorem 1.3 (A closed) fails. This matters because Theorem 1.3 is essential for the extension f_n: X → Δ. The gap can be repaired by using a partition of unity whose closed supports are contained in V_n, as the paper itself points out in Remark 5.6, or by another argument ensuring that the set to which Theorem 1.3 is applied is closed.","section":"§7, Proof of Theorem 7.1"},{"comment":"The same issue occurs in the proof of Theorem 6.1, where each g_n = f|supp(ξ_n) is said to be extendable by Proposition 3.2. Since supp(ξ_n) is the cozero set (open), it is not necessarily closed, and Proposition 3.2 requires a closed domain. The intended conclusion is nevertheless obtainable: one can use the McShane-Whitney extension (Theorem 1.1) and then clip to a bounded interval, or first pass to closed supports via Remark 5.6. The proof as written, however, contains an invalid application.","section":"§6, Proof of Theorem 6.1"}],"minor_comments":[{"comment":"The sentence \"The case of a bounded interval Δ ⊂ R is covered by Theorem 4.1\" is circular; it should refer to Theorem 4.2.","section":"§4, Proof of Theorem 4.1"},{"comment":"The nonstandard definition supp(ξ) = coz(ξ) (an open set) is the source of the gaps in Theorems 6.1 and 7.1. The paper should either adopt the standard closed support for partitions of unity or explicitly state and use the refinement in Remark 5.6 at every place where Theorem 1.3 or Proposition 3.2 is invoked.","section":"§5, Definition of support"}],"recommendation":"major_revision","confidential_remarks":"The core results appear correct and the technical gaps are local and easily fixable. In particular, the main interval-extension theorem (Theorem 1.3) is proved correctly, and the partition-of-unity construction is sound. I would not reject the paper, but the proofs of Theorems 6.1 and 7.1 need to be corrected before publication. The author should clarify the support convention early on and adjust the arguments accordingly."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis is a good paper. The main new result is Theorem 1.3: if A is closed in a metric space and φ: A → Δ is K-Lipschitz with Δ an interval, then φ extends to a K-Lipschitz f: X → Δ. The proof for bounded intervals is a clever averaging of the two McShane-Whitney extensions clipped to the interval, and it works because closedness of A gives d(p,A) > 0 outside A. The unbounded cases follow by the same inequalities. The paper also shows that closedness is necessary, so the theorem is not merely a technicality.\n\nThe other main contributions are Theorem 6.1, representing locally Lipschitz functions as locally finite sums of Lipschitz functions, and Theorem 7.2, a locally Lipschitz version of Michael's selection theorem for open-convex-valued l.s.c. maps. These depend on a direct construction of countable Lipschitz partitions of unity (Theorem 5.1) that avoids Stone's theorem. I checked the partition-of-unity construction: it's correct and genuinely elementary. The selection theorems are natural applications, and they look right.\n\nI agree with the reader's assessment: the proofs are detailed and mostly self-contained, and the reliance on the author's earlier paper [21] is not circular because direct proofs are given for the imported tools. The main theorem is derived from the classical McShane-Whitney formulas, which is fine.\n\nSoft spots are minor. There is a typo in the proof of Theorem 4.1: the bounded interval case is said to be covered by Theorem 4.1, but it should be Theorem 4.2. Also, Proposition 6.6 relies on Hausdorff's theorem on approximation of semi-continuous functions by continuous functions; the paper cites it but does not prove it. That is acceptable for a research paper, but a reader who wants full self-containedness will need to go to the source.\n\nThe paper is not a revolution, but it refines a classical theorem and provides tools that are likely to be reused. The interval extension theorem, in particular, fills a small gap in the literature and should be cited.\n\nRecommendation: send it out for review. It deserves a serious referee, and with minor corrections it should be accepted. I'd bring it to reading group, and I'd cite Theorem 1.3 if I work on Lipschitz extensions.","headline":"A solid, useful refinement of McShane-Whitney with a neat partition-of-unity argument; the interval extension theorem is new and the proofs check out.","tokens_in":19333,"tokens_out":8235,"would_cite":true,"duration_ms":63366,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["26A16","54C20","54C60","54C65","54E35","54E40"],"pacs":[],"model":"deepseek-v4-flash","headline":"A K-Lipschitz function on a closed subset of a metric space with values in any interval of the real line extends to a K-Lipschitz function on the whole space with values in the same interval.","keywords":["Lipschitz function","locally Lipschitz function","locally pointwise Lipschitz function","Lipschitz extension","interval-valued extension","partition of unity","selection theorem","metric space"],"falsifier":"Take $X=\\mathbb{R}$, $A=(0,1)$, $\\Delta=(0,1)$, and $\\phi(t)=t$. This $\\phi$ is 1-Lipschitz, but no continuous $f:\\mathbb{R}\\to(0,1)$ extends it: continuity at 0 and 1 would force $f(0)=0$ and $f(1)=1$, which lie outside $\\Delta$.","tokens_in":18460,"feed_emoji":"📏","tokens_out":12732,"duration_ms":108891,"temperature":0.7,"pith_summary":"This paper proves a sharp refinement of the classical Lipschitz extension theorem: a K-Lipschitz function defined on a closed subset of a metric space and taking values in any interval Δ of the real line can be extended to a K-Lipschitz function on the whole space that still takes values in Δ. The construction takes the two extremal extensions from the classical theorem, clips them to the interval, and averages; the assumption that the subset is closed is exactly what keeps the average strictly inside the interval outside the subset. The same interval-preserving statement is proved for locally pointwise Lipschitz functions and for locally Lipschitz functions, and the locally Lipschitz version needs a new tool: Lipschitz partitions of unity for countable open covers, built directly without appealing to the paracompactness of metrizable spaces. As applications, the paper shows that every locally Lipschitz real-valued function is a locally finite sum of bounded Lipschitz functions, and obtains a locally Lipschitz analogue of a standard continuous selection theorem for open-convex-valued maps.","feed_headline":"Same-constant Lipschitz extensions fit inside any interval","feed_subtitle":"The classical extension theorem now covers interval targets on closed subsets, plus local variants.","key_machinery":"The load-bearing object is the pair of extremal extensions $\\Phi_-(p)=\\sup_{x\\in A}[\\phi(x)-Kd(x,p)]$ and $\\Phi_+(p)=\\inf_{x\\in A}[\\phi(x)+Kd(x,p)]$ from the classical two-sided Lipschitz extension theorem, together with their clipped arithmetic mean $(\\max\\{\\Phi_-,a\\}+\\min\\{\\Phi_+,b\\})/2$. These are $K$-Lipschitz extensions that bracket every other $K$-Lipschitz extension, so averaging their clipped versions keeps values inside any bounded interval; for unbounded intervals one of the two envelopes already works. Closedness of $A$ enters through $d(p,A)>0$ for $p\\notin A$, which forces the average to stay strictly inside $(a,b)$ off $A$. For locally Lipschitz extension and selection, the carrying mechanism is a direct construction of locally finite Lipschitz partitions of unity for countable open covers of a metric space, built from a Lipschitz positive height function and a sequence of Lipschitz cutoffs; this converts local Lipschitz data into globally defined locally finite sums.","core_discovery":"The central claim is Theorem 1.3: for a metric space $(X,d)$, a closed set $A\\subset X$, and any interval $\\Delta\\subset\\mathbb{R}$, every $K$-Lipschitz function $\\phi:A\\to\\Delta$ extends to a $K$-Lipschitz function $f:X\\to\\Delta$. The theorem is proved first for unbounded intervals, where one of the two classical extremal extensions already stays inside $\\Delta$ once $A$ is closed, and then for bounded intervals $(a,b)\\subset\\Delta\\subset[a,b]$, where the average of the clipped upper and lower envelopes is a $K$-Lipschitz extension whose values off $A$ lie strictly inside $(a,b)$. The paper then shows the same interval-preserving extension holds for locally pointwise Lipschitz functions and for locally Lipschitz functions. The locally Lipschitz case is assembled from Lipschitz extensions on local pieces via Lipschitz partitions of unity for countable covers, and the same partition-of-unity machinery yields that locally Lipschitz real-valued functions are locally finite sums of bounded Lipschitz functions and a locally Lipschitz version of a classical convex-valued selection theorem. The paper also notes that closedness cannot be dropped: the identity on an open interval admits no continuous extension into that interval.","pith_inferences":["Extrapolating from the clipped-average proof, the same interval-preserving extension principle should apply to any target set in $\\mathbb{R}$ that is convex, since only the two-envelope comparison and averaging are used; the interval case is the cleanest instance.","The direct partition-of-unity construction avoids the paracompactness of metrizable spaces, so the locally Lipschitz extension and selection results likely need only countable choice rather than the full paracompactness machinery; this is an editorial extrapolation, not a claim in the paper.","Theorem 6.1's representation of locally Lipschitz functions as locally finite sums of bounded Lipschitz pieces gives a practical route to approximating locally Lipschitz functions by finite sums on compact sets, which could support numerical estimates of local Lipschitz constants."],"forward_implications":["Any interval $\\Delta\\subset\\mathbb{R}$ becomes an admissible target for $K$-Lipschitz extension from closed subsets, with the same constant $K$; this includes closed, open, and half-open intervals.","The interval-preserving statement holds for locally pointwise Lipschitz functions (Theorem 4.1) and for locally Lipschitz functions (Theorem 7.1), so the refinement is not a quirk of the global Lipschitz case.","Every locally Lipschitz real-valued function on a metric space is a locally finite sum of bounded Lipschitz functions, and indeed of bounded nonexpansive functions (Theorem 6.1 and Corollary 6.3).","Countable open covers of metric spaces admit locally finite partitions of unity consisting of Lipschitz functions, with a proof that bypasses the usual appeal to the paracompactness of metrizable spaces.","Open-convex-valued lower semi-continuous mappings from a metric space to $\\mathbb{R}$ have locally Lipschitz selections, and prescribed selections on closed subsets extend (Theorem 7.2)."],"supporting_citations":[{"why":"States the original two-sided Lipschitz extension theorem whose formulas (1.1) the whole paper starts from.","marker":"[33]"},{"why":"Provides the paired-extensions argument, invoked in the proof of Proposition 3.2, for the bounded-interval case.","marker":"[10]"},{"why":"Supplies the standard formulation of the two-sided extension construction and the clipping observation used for the compact-interval target.","marker":"[9]"},{"why":"Establishes the Lipschitz partition-of-unity result for open covers of metric spaces whose countable-cover special case Section 5 reproves directly.","marker":"[18]"},{"why":"Supplies earlier locally Lipschitz extension and partition-of-unity lemmas that Sections 6 and 7 build on.","marker":"[21]"},{"why":"Supplies the continuous selection theorem whose locally Lipschitz version is proved as Theorem 7.2.","marker":"[34]"},{"why":"Provides the open-graph and insertion arguments used in the proof of the selection theorem and Corollary 7.5.","marker":"[22]"}],"fun_headline_variants":["Same Lipschitz constant, any interval target","Lipschitz extensions fit inside any interval, globally and locally","Locally Lipschitz functions as sums of Lipschitz functions","New proof for Lipschitz partitions avoids Stone's theorem","McShane-Whitney refined: extensions stay in any interval"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The subset A must be closed; without that, the interval-preserving extension can fail even for the identity map on an open interval.","fun_headline_variants_meta":{"raw":{"variants":["Same Lipschitz constant, any interval target","Lipschitz extensions fit inside any interval, globally and locally","Locally Lipschitz functions as sums of Lipschitz functions","New proof for Lipschitz partitions avoids Stone's theorem","McShane-Whitney refined: extensions stay in any interval"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000933,"raw_usage":{"total_tokens":4011,"prompt_tokens":983,"completion_tokens":3028,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":599,"completion_tokens_details":{"reasoning_tokens":2943}},"tokens_in":599,"tokens_out":3028,"duration_ms":40807,"temperature":1.0,"reasoning_tokens":2943,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T11:48:29.920768+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take $X=\\mathbb{R}$, $A=(0,1)$, $\\Delta=(0,1)$, and $\\phi(t)=t$. This $\\phi$ is 1-Lipschitz, but no continuous $f:\\mathbb{R}\\to(0,1)$ extends it: continuity at 0 and 1 would force $f(0)=0$ and $f(1)=1$, which lie outside $\\Delta$.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"States the original two-sided Lipschitz extension theorem whose formulas (1.1) the whole paper starts from."},{"cited_title":"Czipszer and L","cited_arxiv_id":null,"evidence_quote":"Provides the paired-extensions argument, invoked in the proof of Proposition 3.2, for the bounded-interval case."},{"cited_title":"Cobza¸ s, R","cited_arxiv_id":null,"evidence_quote":"Supplies the standard formulation of the two-sided extension construction and the clipping observation used for the compact-interval target."},{"cited_title":"Frol ´ ık,Existence of ℓ∞ -partitions of unity , Rend","cited_arxiv_id":null,"evidence_quote":"Establishes the Lipschitz partition-of-unity result for open covers of metric spaces whose countable-cover special case Section 5 reproves directly."},{"cited_title":"Gutev, Lipschitz extensions and approximations , J","cited_arxiv_id":null,"evidence_quote":"Supplies earlier locally Lipschitz extension and partition-of-unity lemmas that Sections 6 and 7 build on."},{"cited_title":"Michael, Continuous selections I , Ann","cited_arxiv_id":null,"evidence_quote":"Supplies the continuous selection theorem whose locally Lipschitz version is proved as Theorem 7.2."}],"review_version":1}