---

Subject and independence

Role. Independent reviewer in a fresh context, commissioned for refutation. The reviewer took no part in writing the connections page, the reconstruction pages or their sources, received only the assignment, and read no verification record, grade or standing text of this folder.

Subject. Page wiki/research/erdos_1171/connections.md as it stood on 2026-09-28T13:50:48Z, section ## Problem 1168, lines 597--619 (heading at line 597, last content line 618, next H2 at line 620). The section files no lead page: its outcome line reads "Fails; no lead", so there is no lead page to review.

Read, and at what depth.

• The subject page: frontmatter, scope paragraph and Tools section (lines 1--123) in full, as the page's shared conventions and the interfaces of (T1)--(T5), whose proofs are another reviewer's; the section "Problems 1174, 595 and 596" (lines 549--596) in full, because the subject section consumes its bit coloring by the words "One remark of the previous section applies", and that witness is re-derived below.
• The reconstruction pages Lemma 2.1 and Theorem 3.1 in full, as premises at the standing the assignment states: their statements, definitions, facts 1 and 2 of the lemma page, and the paragraph "An alternative bridge" of the theorem page, the only place where the term "Ramsey bridge" is explained.
• Problem 1168: the statement block under its H1 (lines 13--32) only; the page has no ## Statement heading.
• Garti, Hayut and Shelah: the card's provenance and digest paragraph and its list of results to transcribe; the held PDF (arXiv:2502.16625v2) by text extraction at the abstract, the introduction (pp. 2--3), Theorem 1.1 (p. 5), Corollary 1.2 and Theorem 1.3 (p. 7), Theorem 2.2 (pp. 10--11), the forcing argument at $\lambda=\aleph_\omega$ (pp. 15--18) and Theorem 3.3 (p. 20). No page images were read.
• Erdős and Rado 1956: the card's provenance paragraph and the held PDF by text extraction at the definition of the partition relation (printed pp. 429--430) and Theorem 44 (printed p. 475), the statement the section invokes as the Erdős--Dushnik--Miller theorem without a citation.
• docs/verification.md sections "Whole-claim report", "Audit checklist" and "Grading and claim standing" (the shared and the Erdos-specific versions of the first two), docs/approach_vetting.md, docs/evidence.md section "Source fidelity" and docs/math_authoring.md.

Exposures. Four, none bearing on a verdict. (1) The section "Problem 597: graph targets" of the subject page (lines 318--409) was read while locating where the term "Ramsey bridge" is defined; it holds no review, standing or acceptance text, and nothing from it is used here. (2) The heading list of the Problem 1168 page was seen while locating its statement; the bodies of Current assessment, Progress, Known Results and Linked library material were not read. (3) The digest paragraph of the Garti--Hayut--Shelah card carries one editorial sentence on the problem's openness and one on Problem 597; both were read and neither is used. (4) Each reconstruction page opens with a Standing paragraph (author-recorded reconstruction; unrefereed deposit); it was read as part of the premise pages. No _index.md, nothing under any evidence/ folder, no review or grade, and no web page was read.

Restatement

Conventions. Colorings, homogeneous sets, triangles and the relation $\alpha\to(\beta_0,\ldots,\beta_{n-1})^2_n$ are those of the lemma page: the subscript counts the colors, color $0$ carries the large target, and a homogeneous set of order type $\beta_i$ is meant. A cardinal is its initial ordinal. The catalog and the held sources use the cardinality form, in which the large homogeneous set has cardinality $\kappa$; when the large target is the ambient cardinal the two forms coincide, because a subset of the ordinal $\kappa$ of cardinality $\kappa$ has order type exactly $\kappa$: its order type is an ordinal of cardinality $\kappa$, so at least $\kappa$, and at most $\kappa$ as a subset of $\kappa$. The section's ten statements:

• S1 (target). Problem 1168, carried at status open, asks for a proof without GCH, that is in ZFC, of $\aleph_{\omega+1}\not\to(\aleph_{\omega+1},3,\ldots,3)^2_{\aleph_0}$: a coloring $c:[\aleph_{\omega+1}]^2\to\omega$ with no set of cardinality $\aleph_{\omega+1}$ homogeneous in color $0$ and no triangle homogeneous in any color $\gamma\ge1$.
• S2 (source characterization). The library card of Garti, Hayut and Shelah records (a) the classical case: for singular $\lambda$ with $2^\lambda=\lambda^+$, $\lambda^+\not\to(\lambda^+,(3)_{\mathrm{cf}(\lambda)})^2$; and (b) the consistency of that negative relation with $2^\lambda>\lambda^+$ under pcf and guessing hypotheses.
• S3 (imported theorem). For every infinite cardinal $\kappa$, $\kappa\to(\kappa,\omega)^2$ (Erdős--Dushnik--Miller), read in the order-type convention: every two-coloring of $[\kappa]^2$ has a subset of order type $\kappa$ homogeneous in color $0$ or a subset of order type $\omega$ homogeneous in color $1$.
• S4. For every infinite cardinal $\kappa$, $\kappa\to(\kappa,3)^2$.
• S5. For every infinite cardinal $\kappa$ and every finite $k$, $\kappa\to(\kappa,3,\ldots,3)^2_{k+1}$ with $k$ triangle targets, by Lemma 2.1 at $\alpha=\kappa$.
• S6. The conclusion of S5 also follows, by the Ramsey bridge, from $\kappa\to(\kappa,n)^2$ for finite $n$.
• S7 (obstruction). The problem's relation has $\aleph_0$ colors; the color-merging induction proves the relation for each finite $k$ and has no limit step, so it does not reach the $\aleph_0$-color relation.
• S8 (conditional witness). If $\aleph_{\omega+1}\le2^{\aleph_0}$, the bit coloring of the previous section witnesses $\aleph_{\omega+1}\not\to(\aleph_{\omega+1},3,3,\ldots)^2_{\aleph_0}$, which is the problem's relation.
• S9 (reduction). Hence the content of the problem is the case $2^{\aleph_0}<\aleph_{\omega+1}$, and the cited work operates in that case.
• S10 (outcome). The mechanism of the folder gives no lead on Problem 1168: only the finite-color positive relations transfer, and they are trivial.

Checklist

• Quantifiers and scope. Pass. S3--S5 quantify over every infinite cardinal, singular ones included, and every finite $k$; S5 includes $k=0$, the one-color relation, which holds outright since $\kappa$ itself is homogeneous under the only coloring. S8 states its hypothesis. S6 leaves $n$ unquantified (F3, a note). S9's case split is a trichotomy of cardinals, so no case is dropped.
• Circularity. Pass. The only inputs are an external theorem (S3), the reviewed Lemma 2.1, the finite Ramsey theorem and the bit coloring; none is equivalent to a statement of the section, and no statement of the section is the problem's relation or its negation.
• Model and convention changes. Pass with a note. The held sources state S3 and the problem in the cardinality convention and the section states them in the order-type convention; the bridge is the sentence under Conventions above together with the fact that an infinite set of ordinals has a subset of order type $\omega$. The section leaves it implicit (F2). The countable-color relation is read with the subscript counting colors, as on the lemma page and in the source's definition of $\lambda\to(\kappa,(3)_\theta)^2$.
• Finite and statistical overreach. Pass. The finite-color positives are not used to say anything about the $\aleph_0$-color relation; the section says the opposite (S7).
• Uniformity. Pass. The only parameter-dependent quantity is the Ramsey number in S6, which depends on $k$ and is used for one $k$ at a time.
• Extremal conclusions. Inapplicable; the section states no extremum.
• Consequences and composition. Pass with a note. Each "so" is checked separately below: S3 to S4 by (T2); S4 to S5 within Lemma 2.1's hypotheses; S3 to S6 by (T2) and the finite Ramsey theorem; S8 to S9 by the trichotomy of cardinals. S9 is true as written and can be sharper (F4).
• Computation. Inapplicable; no computation is claimed.
• Reproduction. Inapplicable; no rerun command or coverage line is claimed.
• Source and verdict fidelity. One suggested correction. S1 matches the problem page's statement block in every clause. S2 is faithful to the card but, checked against the held PDF, reads more than the source proves on the guessing route and omits the large-cardinal hypotheses of the consistency results (F1). S3 is invoked by name with no citation; the held Erdős--Rado paper states it as Theorem 44 (F2). The clause "which is where the cited work operates" is confirmed by the source's standing assumption $2^{\mathrm{cf}(\lambda)}<\lambda$ (p. 2). The section quotes no verdict of another reviewer.

Weakest steps

W1: the convention bridge behind S3--S5. Theorem 44 of Erdős and Rado (printed p. 475) states, for every cardinal $a\ge\aleph_0$, $a\to(\aleph_0,a)^2$ in the cardinality form of their Section 1: whenever $|S|=a$ and $[S]^2=K_0\cup K_1$ with $K_0\cap K_1=\emptyset$, some $B\subseteq S$ has $|B|=\aleph_0$ and $[B]^2\subseteq K_0$, or has $|B|=a$ and $[B]^2\subseteq K_1$. Swapping the two classes gives $a\to(a,\aleph_0)^2$. Take $S=\kappa$, the initial ordinal, and a coloring $c:[\kappa]^2\to\{0,1\}$; put $K_0=c^{-1}(0)$ and $K_1=c^{-1}(1)$. Either some $B$ with $|B|=\kappa$ is homogeneous in color $0$, and then $\operatorname{otp}(B)=\kappa$ by the sentence under Conventions; or some infinite $B$ is homogeneous in color $1$, and its first $\omega$ elements form a subset of order type $\omega$ homogeneous in color $1$. This is S3 in the order-type form. S4 is (T2) with $3\le\omega$. S5: Lemma 2.1 asks for an ordinal $\alpha$ with $\alpha\to(\alpha,3)^2$ and returns, for every finite $k\ge1$, $\alpha\to(\alpha,3,\ldots,3)^2_{k+1}$; $\kappa$ is an ordinal and S4 is the hypothesis, so S5 holds for $k\ge1$, and for $k=0$ outright. The proof of Theorem 44 treats singular $a$ in its Case 2, so nothing restricts S3--S5 to regular cardinals.

W2: the witness of S8 in the problem's own convention. Assume $\aleph_{\omega+1}\le2^{\aleph_0}$ and fix an injection $\iota:\aleph_{\omega+1}\to2^\omega$. Color $\{x,y\}$ by $1+n$, where $n$ is the least position at which $\iota(x)$ and $\iota(y)$ differ; $n$ exists because $\iota$ is injective. The coloring takes values in $\{1,2,\ldots\}\subseteq\omega$, so it is a coloring with $\aleph_0$ colors in which color $0$ is unused. No set with two or more elements is homogeneous in color $0$, since it contains a pair and no pair has color $0$; in particular no set of cardinality $\aleph_{\omega+1}$, and no set of order type $\aleph_{\omega+1}$, is. If $\{x,y,z\}$ were homogeneous in color $1+n$, the three sequences $\iota(x)$, $\iota(y)$, $\iota(z)$ would agree below $n$ and differ pairwise at $n$, which needs three distinct bits. So the coloring witnesses $\aleph_{\omega+1}\not\to(\aleph_{\omega+1},(3)_{\aleph_0})^2$ in the source's definition (a $0$-homogeneous set of full cardinality or a $\gamma$-homogeneous triangle with $0<\gamma<\aleph_0$) and equally in the order-type form. The source records the same fact on p. 2: forcing $2^{\mathrm{cf}(\lambda)}\ge\lambda^+$ gives the negative relation, as the monograph of Erdős, Hajnal, Máté and Rado notes.

W3: the reduction S9. By the trichotomy of cardinals, either $\aleph_{\omega+1}\le2^{\aleph_0}$, where S8 settles the problem negatively, or $2^{\aleph_0}<\aleph_{\omega+1}$. Since $\operatorname{cf}(2^{\aleph_0})>\aleph_0=\operatorname{cf}(\aleph_\omega)$ by König's theorem, $2^{\aleph_0}\ne\aleph_\omega$, so the second case is $2^{\aleph_0}<\aleph_\omega$, that is, $2^{\aleph_0}=\aleph_n$ for some finite $n$. That is the source's standing assumption $2^{\mathrm{cf}(\lambda)}<\lambda$ at $\lambda=\aleph_\omega$ (p. 2), and every consistency result of the source has $\lambda$ a strong limit, which is stronger. So the cited work operates inside the case S9 names. The case is not all open: its sub-case $2^{\aleph_\omega}=\aleph_{\omega+1}$ is the classical theorem the section itself records (Theorem 0.1 of the source at the singular cardinal $\lambda=\aleph_\omega$), so what remains is $2^{\aleph_0}<\aleph_\omega$ together with $2^{\aleph_\omega}>\aleph_{\omega+1}$ (F4).

Strongest attack

• S1. Compared clause by clause with the problem page's statement block: the relation, the subscript, the "without GCH" qualification and the status agree. No attack is available.
• S2. Attacked as a source characterization. Against the card: the card lists Theorem 0.1 as the classical case and describes Theorem 1.1, Corollary 1.2 and Theorem 1.3 as giving the negative relation with $2^\lambda>\lambda^+$ under pcf-style scales and a guessing principle, so the sentence reports the card correctly and survives as stated. Against the held PDF the attack lands in part: the hypotheses (a)--(h) of Theorem 1.1 are cardinal-arithmetic and pcf conditions with no guessing principle; the guessing principle enters only in Theorem 3.3 (from stick), and the source says on p. 3 that it does not know whether stick at $\lambda$ is consistent with $2^\lambda>\lambda^+$ for strong limit singular $\lambda$; and the consistency results assume a supercompact cardinal (Corollary 1.2, and the argument at $\aleph_\omega$) or a strong cardinal (Theorem 1.3). The sentence is therefore true of the card and imprecise about the mathematics; filed as F1, suggested, since no deduction of the section consumes it.
• S3. Attacked at singular $\kappa$, where the original argument of Dushnik and Miller needed Erdős's addition: the held Theorem 44 covers every $a\ge\aleph_0$ and its proof treats $a'<a$ separately. Attacked at the convention: the bridge of W1 closes it. Failed.
• S4, S5. Attacked by looking for a hypothesis of Lemma 2.1 that a cardinal might miss: the lemma needs only an ordinal with $\alpha\to(\alpha,3)^2$, used on subsets of order type exactly $\alpha$, which fact 1 of the lemma page transports; the cardinal $\kappa$ is an ordinal. Failed.
• S6. Attacked by asking which $n$: the bridge needs $\kappa\to(\kappa,n)^2$ for $n$ the least integer such that every $k$-coloring of the pairs of an $n$-element set has a monochromatic triangle, which exists by the finite Ramsey theorem, and S3 with (T2) gives $\kappa\to(\kappa,n)^2$ for every finite $n$. Re-derived: given $c:[\kappa]^2\to\{0,\ldots,k\}$, merge the colors $1,\ldots,k$ into $1$; a color-$0$ set of type $\kappa$ is what is wanted, and an $n$-element set all of whose pairs have merged color $1$ carries a $k$-coloring by $c$, hence a triangle homogeneous in some color $j\in\{1,\ldots,k\}$. The statement survives; the unquantified $n$ is F3.
• S7. Attacked by trying to build a limit step. Given $c:[\aleph_{\omega+1}]^2\to\omega$, merging every color above $k$ into $0$ gives a coloring with the colors $0,\ldots,k$, and S5 gives for each $k$ either a triangle homogeneous under $c$ in a color in $\{1,\ldots,k\}$ or a set $Y_k$ of full type on which $c$ takes values in $\{0\}\cup(k,\omega)$; the sets $Y_k$ need not have a common subset of full type, and no diagonal argument produces one. The attack cannot succeed in ZFC at all: the $\aleph_0$-color positive relation at $\aleph_{\omega+1}$ fails under $2^{\aleph_\omega}=\aleph_{\omega+1}$ (Theorem 0.1 of the source) and under $\aleph_{\omega+1}\le2^{\aleph_0}$ (S8), so no ZFC argument, with or without a limit step, reaches it. The obstruction stands; that it is an obstruction of the statement rather than of the proof is F5.
• S8. Attacked in the problem's convention (W2), at the hypothesis (the injection exists exactly when $\aleph_{\omega+1}\le2^{\aleph_0}$) and at the color count (the colors $1+n$ for $n<\omega$ are $\aleph_0$ many, and the unused color $0$ is allowed by the definition on the lemma page and by the source's, both of which quantify over all colorings into the color set). Failed.
• S9. Attacked at the case split (W3) and at the clause about the cited work, which p. 2 of the source confirms. Failed; sharpened in F4.
• S10. Attacked by looking for a transfer the section misses. The mechanism produces positive relations from positive relations ((T3), (T4)), and its one negative-direction tool, (T5), turns a finite-color witness into a two-color witness, which cannot exist at $\aleph_{\omega+1}$ by S3 and transport; a countable analog of (T5) is false, since under $2^{\aleph_\omega}=\aleph_{\omega+1}$ the $\aleph_0$-color relation fails while every finite-color relation holds by S5. Nothing in the folder bears on the case S9 leaves open. Failed.

Premises

• Lemma 2.1: for an ordinal $\alpha$ with $\alpha\to(\alpha,3)^2$ and every finite $k\ge1$, $\alpha\to(\alpha,3,\ldots,3)^2_{k+1}$ with $k$ triangle targets; with fact 1 (transport along order isomorphisms) and fact 2 (relabeling of colors). Read in full; treated as established at the standing the assignment states; used at $\alpha=\kappa$ within its hypotheses.
• Theorem 3.1: read in full; its theorem is not used by the section. Its paragraph "An alternative bridge" describes the Ramsey bridge as an author-recorded step with no written source; the section uses that bridge at $\kappa$, and the S6 attack re-derives it here.
• Tool (T2) of the subject page (restriction and weakening; a target may be replaced by a smaller one, a color left unused): interface confirmed; proof another reviewer's. (T3)--(T5) are not used by the section.
• The bit coloring of the section "Problems 1174, 595 and 596" (subject page, lines 549--596): for every order type $\alpha$ with $2\le|\alpha|\le2^{\aleph_0}$, $\alpha\not\to(\alpha,3,3,\ldots)^2_{\aleph_0}$, author-recorded; re-derived in W2 at $\alpha=\aleph_{\omega+1}$.
• Erdős and Rado 1956, Theorem 44 (held PDF, printed p. 475; text extraction): for every cardinal $a\ge\aleph_0$, $a\to(\aleph_0,a)^2$ in the cardinality convention of the paper's Section 1 (printed pp. 429--430). Held. Not cited by the section (F2).
• Garti, Hayut and Shelah, arXiv:2502.16625v2 (held PDF; text extraction): Theorem 0.1 (p. 2; $\lambda$ singular and $2^\lambda=\lambda^+$ imply $\lambda^+\not\to(\lambda^+,(3)_{\mathrm{cf}(\lambda)})^2$); the remark on p. 2 that $2^{\mathrm{cf}(\lambda)}\ge\lambda^+$ gives the negative relation, and the standing assumption $2^{\mathrm{cf}(\lambda)}<\lambda$; Question 0.2 (p. 2); Theorem 1.1 (p. 5, hypotheses (a)--(h): $\mu$ singular strong limit with $2^\mu>\mu^+$, an increasing sequence of singular strong limit cardinals $\mu_i$ of the same cofinality with $2^{\mu_i}=\mu_i^+$ and union $\mu$, and true cofinality $\mu^+$ of the product of the $\mu_i^+$ modulo the bounded ideal); Corollary 1.2 (p. 7, from a supercompact cardinal); Theorem 1.3 (p. 7, from a strong cardinal, at $\mu=\aleph_{\omega^2}$); the argument of the second section forcing the negative relation at $\lambda=\aleph_\omega$ with $2^{\aleph_\omega}=\aleph_{\omega+2}$ (pp. 15--18, from a supercompact cardinal); Theorem 3.3 (p. 20, from stick at $\lambda$) and the remark on p. 3 that the consistency of stick with $2^\lambda>\lambda^+$ at strong limit singular $\lambda$ is not known. Held.
• The finite Ramsey theorem for triangles in $k$ colors and König's theorem ($\operatorname{cf}(2^{\aleph_0})>\aleph_0$): standard, used in W3 and the S6 attack; no library source consulted.
• Problem 1168: statement block only.

Findings

• F1 (suggested). Location: lines 603--605, "records the classical case $2^\lambda=\lambda^+$ and the consistency of the negative relation with $2^\lambda>\lambda^+$ under pcf and guessing hypotheses." Defect: read against the held PDF, the consistency of the negative relation with $2^\lambda>\lambda^+$ is obtained on the pcf route only (Corollary 1.2, Theorem 1.3 and the second section at $\aleph_\omega$), each from a supercompact or a strong cardinal; the guessing route (Theorem 3.3, from stick) yields the negative relation, and the source states on p. 3 that whether stick at $\lambda$ is consistent with $2^\lambda>\lambda^+$ for strong limit singular $\lambda$ is not known. The sentence attributes the consistency to both kinds of hypothesis and omits the large-cardinal hypotheses. It follows the card's digest, which places a guessing principle among the hypotheses of Theorem 1.1; the theorem's list (a)--(h) contains none. Witness: the PDF, pp. 3, 5, 7, 18 and 20. Replacement: "records the classical case $2^\lambda=\lambda^+$ (Theorem 0.1), the negative relation from pcf hypotheses (Theorem 1.1) and from the guessing principle stick (Theorem 3.3), and, from a supercompact or a strong cardinal, the consistency of the negative relation with $2^\lambda>\lambda^+$ along the pcf route (Corollary 1.2, Theorem 1.3 and the second section at $\lambda=\aleph_\omega$)." The card's own digest is outside this review's subject.
• F2 (note). Location: lines 607--609, "by the Erdős--Dushnik--Miller theorem $\kappa\to(\kappa,\omega)^2$ for every infinite cardinal $\kappa$". Defect: no source is cited, although the page's scope paragraph says that sources are cited at the standing their library pages record, and the theorem is stated in the order-type convention while the held source states it in the cardinality convention. Witness: Theorem 44 of the held Erdős--Rado paper. Replacement: cite Erdős and Rado 1956, Theorem 44, and add "a subset of $\kappa$ of cardinality $\kappa$ has order type $\kappa$, so the cardinality form gives the order-type form".
• F3 (note). Location: lines 610--611, "the Ramsey bridge gives the same from $\kappa\to(\kappa,n)^2$". Defect: $n$ is unquantified; the bridge for $k$ triangle colors needs $n$ equal to the $k$-color Ramsey number of the triangle. Replacement: "from $\kappa\to(\kappa,n)^2$ for every finite $n$, which (T2) gives from the same theorem".
• F4 (note). Location: lines 614--616, "so the content of the problem is the case $2^{\aleph_0}<\aleph_{\omega+1}$, which is where the cited work operates." Defect: true, but the sub-case $2^{\aleph_\omega}=\aleph_{\omega+1}$ is the classical theorem the Target paragraph records, and by König's theorem the case equals $2^{\aleph_0}<\aleph_\omega$, the source's own standing assumption. Replacement: "so the content of the problem is the case $2^{\aleph_0}<\aleph_\omega$ (the same as $2^{\aleph_0}<\aleph_{\omega+1}$ by König's theorem) with $2^{\aleph_\omega}>\aleph_{\omega+1}$, the classical case being settled; the cited work operates there, with $\aleph_\omega$ a strong limit."
• F5 (note). Location: line 612, "where the induction has no limit step." Defect: none in the sentence; the record can say, as the previous section does for its targets, that the obstruction is of the statement: the $\aleph_0$-color positive relation at $\aleph_{\omega+1}$ fails under $2^{\aleph_\omega}=\aleph_{\omega+1}$ and under $\aleph_{\omega+1}\le2^{\aleph_0}$, so no ZFC argument reaches it, and the mechanism produces positive relations while the problem asks for a negative one.

Verdict

• S1: refutation-failed.
• S2: refutation-failed as a report of what the card records; as a characterization of the held source it needs the precision of F1.
• S3: refutation-failed.
• S4: refutation-failed.
• S5: refutation-failed.
• S6: refutation-failed.
• S7: refutation-failed.
• S8: refutation-failed.
• S9: refutation-failed.
• S10: refutation-failed.

Required corrections: none. Suggested: F1. Notes: F2--F5.

Limitations. The held PDFs were read by text extraction and not on page images; the statements relied on (Theorem 0.1, the remarks on p. 2, the hypothesis list of Theorem 1.1, Corollary 1.2, Theorem 1.3, Theorem 3.3 and Theorem 44 of Erdős and Rado) extracted legibly. The proofs of (T1)--(T5) and of the reconstruction pages were not re-verified here; the section uses only (T2) and Lemma 2.1, both within their stated hypotheses. The bit coloring was re-derived because the section consumes it as a witness. This review does not assess the proofs of the cited sources.

This focused review assigns no tier and changes no status.