---

Subject and independence

Role: independent reviewer in a fresh context, given only the assignment; took no part in writing the page or the folder's reconstructions. Charge: refutation of the author-recorded mathematics of one section.

Subject: page wiki/research/erdos_1219/connections.md as it stood at 2026-09-28T13:50:48Z, section "## Problem 1220: the singular unbalanced relation", lines 171--285 (from its H2 heading to the line before "## Hajnal's Conjecture 1A: triples" at line 286). No separate lead page was named; the section is the subject.

Read, at the depth stated:

• the section, every sentence, with each proof re-derived;
• the page's Scope paragraph and its Tools section ((T1), (T2), (T3)) as inputs: their interfaces were confirmed at each use and their proofs were read for the interface only, since another reviewer checks them;
• the three reconstruction pages of the folder (Lemma 1.1, Theorem 1.2, Corollary 1.3) in full, as premises established at their standing;
• the Statement section of wiki/problems/set_theory/E1220/_index.md;
• the provenance paragraph of the Komjáth 2025 card and, for Problem 4, the held conversion and the page images of printed pp. 419--420 (PDF pp. 2--3) of the held PDF;
• the provenance paragraph of the Erdős and Rado 1956 card and the page image of printed p. 432 (PDF p. 6) of the held PDF, where the theorem the section imports as "not held" is printed;
• the Statement section of the Shelah 1975 result page theorem_1_2;
• docs/verification.md ("Whole-claim report", "Audit checklist", "Grading and claim standing"), docs/approach_vetting.md, docs/evidence.md ("Source fidelity") and docs/math_authoring.md.

Exposures, disclosed: (1) while reading the problem page for its Statement, its Current assessment and Progress sections (lines 1--80 of the page) were displayed; they paraphrase the section under review and carried nothing that is not in the subject. (2) While reading the page head and Tools, the neighboring section "Problem 1219 beyond the catalog question" (lines 109--170) was displayed; nothing from it is used. (3) The Shelah 1975 result page was read past its Statement (proof pointer, dependencies and bears-on parts); nothing from those parts is used. No evidence folder, index page, review, grade, standing or acceptance text, workspace content or web page was read. The catalog page and the Problem 1171 connections page named in the section were not read; the sentences resting on them are marked unverified below.

Restatement

Notation: $\theta\to(\mu_0,\mu_1)^2$ means that every two-coloring of the pairs of a set of size $\theta$ has a set of size $\mu_0$ homogeneous in color $0$ or a set of size $\mu_1$ homogeneous in color $1$; a cardinal $\kappa$ is $\aleph_0$-inaccessible when $\mu^{\aleph_0}<\kappa$ for every cardinal $\mu<\kappa$; $\lambda$ is strong limit when $2^\mu<\lambda$ for every cardinal $\mu<\lambda$.

• S1 (fidelity). Problem 1220 asks whether $\lambda\to(\lambda,\aleph_1)^2$ for every singular $\lambda$ such that $\lambda$ and $\operatorname{cf}\lambda$ are both $\aleph_0$-inaccessible; it is Problem 4 of the Erdős--Hajnal list; the survey's commentary on Problem 4 records the necessity of both hypotheses, the simplest case $\lambda=\aleph_{\mathfrak c^+}$ with the reported theorem of Shelah, an Erdős--Rado theorem [45, Theorem 35.4], and the two consistency results [170] and [171] of Shelah and Stanley, as the section lists them. S1b, separated for its verdict: the survey attributes to [45, Theorem 35.4] the positive relation for every strong limit $\lambda$ ("gives the positive relation when $\lambda$ is strong limit"; "the strong-limit case of the problem, the case Komjáth attributes to [45, Theorem 35.4]").
• S2 (conditional). If the survey's summary of [170] is accurate, the universal question, read in ZFC, has a consistent negative answer at $\lambda=\aleph_{\mathfrak c^+}$.
• S3 (transfer lemma). For every singular strong limit cardinal $\lambda$ with $\kappa=\operatorname{cf}\lambda$ and every cardinal $\theta\le\kappa$ with $\kappa\to(\kappa,\theta)^2$: $\lambda\to(\lambda,\theta)^2$.
• S4 (equivalence). For every singular strong limit $\lambda$: $\lambda\to(\lambda,\aleph_1)^2$ if and only if $\operatorname{cf}\lambda\to(\operatorname{cf}\lambda,\aleph_1)^2$; the forward direction is (T2) with $\gamma=1$, $\rho_0=\aleph_1$.
• S5 (composition with an imported theorem). For every regular $\aleph_0$-inaccessible $\kappa$, $\kappa\to(\kappa,\aleph_1)^2$ (the Erdős--Rado theorem for pairs, imported as not held); hence S3 with $\theta=\aleph_1$ proves the problem's relation for every singular strong limit $\lambda$ whose cofinality is $\aleph_0$-inaccessible.
• S6 (obstruction). S6a: Lemma 1.1, applied to a block decomposition of $\lambda$ into $\operatorname{cf}\lambda$ blocks with the union of the chosen $B_\alpha$ as the homogeneous set, is available exactly when $\lambda$ is strong limit: for non-strong-limit $\lambda$ the sizes $|E_\alpha|$ of the parameter sets are cofinal in $\lambda$, the type bound $2^{|E_\alpha|}$ reaches $\lambda>\lambda_\alpha$, and bounding the $\mu(i)$ below the least $\mu$ with $2^\mu\ge\lambda$ leaves a final set of size below $\lambda$. S6b, separated for its verdict: "no block inside $\lambda$ can be thinned to a single type".
• S7 (consequence). For every cardinal $\lambda$ and every finite $k\ge1$, $\lambda\to(\lambda,\aleph_1)^2$ implies $\lambda\to(\lambda,(\aleph_1)_k)^2$, by merging colors.
• S8 (assessment, not claims). The reported [170] and [171] show that the open part is not decided by "cardinal arithmetic of this kind"; any proof must use the $\aleph_0$-inaccessibility of $\lambda$ without passing through $2^{|E_\alpha|}$; the next-investigation items (i) and (ii); the Outcome.

Checklist

• Quantifiers and scope: pass for S2--S5 and S7 (the transfer lemma carries $\theta\le\kappa$ and "singular strong limit" explicitly, and the corollary's hypothesis $\kappa\to(\kappa,\aleph_1)^2$ already forces $\aleph_1\le\kappa$); pass for S6a once "the mechanism" is read as Lemma 1.1 on blocks inside $\lambda$ with the union of the $B_\alpha$ as the homogeneous set, which is how the section uses it; fail for S1b, whose "$\lambda$" widens the survey's $\aleph_{\mathfrak c^+}$ (F2), and for S6b, whose strongest reading is a universal impossibility that the argument does not show (F1).
• Circularity: pass; no proof assumes its conclusion, and the transfer lemma's inputs ((T1), (ER), the cofinality relation) do not mention $\lambda\to(\lambda,\theta)^2$.
• Model and convention changes: pass; the two-place symmetric $\tilde f$ of Step 5 represents the pair coloring, blocks are sets and the relation depends only on cardinality, and (T1)'s interface is met clause by clause (Premises).
• Finite and statistical overreach: inapplicable; no finite sample or heuristic average is used.
• Uniformity: inapplicable; no constants or error terms.
• Extremal conclusions: inapplicable; no infimum, supremum or sharpness sentence beyond "exactly when", which is checked under S6a.
• Consequences and composition: pass for S4, S5 and S7 after supplying the two links the section leaves to unheld sources (the regular-case theorem is re-derived under Premises, the color merging under Strongest attack); fail for S6b (F1). The composition S3 with S5 inherits S5's unheld standing.
• Computation: inapplicable; the section computes nothing.
• Reproduction: inapplicable; no rerun instructions or coverage claims.
• Source and verdict fidelity: fail for S1b (F2); at suggested level, the page citation "p. 419" covers text printed on p. 420 (F4) and "not held" is said of a theorem whose successor case is printed in the held Erdős--Rado 1956 paper (F5); the reports of Problem 4's statement, of the necessity remark and of [170] and [171] match the page images; the catalog-page sentence and the Problem 1171 page's note are outside the read set and stay unverified (F8).

Weakest steps

W1. The realizability hypothesis (H) in the transfer lemma. Re-derived: fix $\alpha<\kappa$ and $C\subseteq A_\alpha$ with $|C|=\lambda_\alpha=(2^{\mu(\alpha)})^+$; $\mu(\alpha)$ is infinite, so (ER) transported to $C$ gives $H\subseteq C$ with $|H|=\lambda_\alpha$ and $f$ constantly $0$ on $[H]^2$, or $|H|=\mu(\alpha)^+$ and $f$ constantly $1$; since $\mu(\alpha)\ge\mu(0)\ge\kappa\ge\theta$, the second $H$ contains a $1$-homogeneous set of size $\theta$, excluded by the standing assumption; any $B\subseteq H$ with $|B|=\mu(\alpha)$ satisfies $P_\alpha$. The property depends on $B_\alpha$ alone, so (H) holds for every admissible sequence and every choice of the points $a_i$. The lemma's clause (2) then gives $|B_\alpha|=\mu(\alpha)$ exactly, which the size count W2 needs.

W2. The sizes. $\lambda_i=(2^{\mu(i)})^+<\lambda$ needs $\lambda$ strong limit and a limit cardinal, both given; $\sum_{i<\kappa}\lambda_i$ equals $\kappa\cdot\sup_i\lambda_i$ by the sum formula, and $\mu(i)<\lambda_i<\lambda$ with $\sup_i\mu(i)=\lambda$ gives $\sup_i\lambda_i=\lambda$, so the sum is $\lambda$. For $I\subseteq\kappa$ of size $\kappa$, $I$ is unbounded in the regular $\kappa$ and the $\mu(\alpha)$ increase, so $\sup_{\alpha\in I}\mu(\alpha)=\lambda$ and, the blocks being disjoint, $|\bigcup_{\alpha\in I}B_\alpha|=\sum_{\alpha\in I}\mu(\alpha)=\lambda$. Both agree with the section.

W3. The obstruction as a statement about Lemma 1.1. Re-derived from the lemma's growth condition rather than from the type count: the lemma requires $\prod_{i<\alpha}\lambda_i^{\mu(i)}<\lambda_\alpha$ with each $\lambda_i$ regular. Blocks lie inside $\lambda$ and $\lambda$ is singular, so every regular $\lambda_\alpha\le\lambda$ is below $\lambda$. Since $\lambda_i\ge2$, the condition gives $2^{\sum_{i<\alpha}\mu(i)}\le\prod_{i<\alpha}\lambda_i^{\mu(i)}<\lambda_\alpha<\lambda$. The final homogeneous set is $\bigcup_{\alpha\in I}B_\alpha$ with $|B_\alpha|\le\mu(\alpha)$, for an $I$ that the cofinality relation chooses, so its size $\lambda$ forces $\sup_{\alpha<\kappa}\mu(\alpha)=\lambda$. Then for every $\mu<\lambda$ there is $i$ with $\mu\le\mu(i)\le\sum_{j\le i}\mu(j)$, whence $2^\mu<\lambda$: the lemma's hypotheses force $\lambda$ strong limit. The number of blocks must be $\operatorname{cf}\lambda$: a strictly increasing sequence of regular cardinals below $\lambda$ with supremum $\lambda$ has length of cofinality $\operatorname{cf}\lambda$, and the lemma's $\kappa$ is regular, while a bounded sequence covers fewer than $\lambda$ points. So S6a holds in the form "the lemma's hypotheses can be met for such a decomposition exactly when $\lambda$ is strong limit"; nothing stronger is proved (F1).

Strongest attack

S1: each listed item was compared with the page images of printed pp. 419--420 of the survey. The problem statement, the definition of $\aleph_0$-inaccessibility, the necessity remark ($\tau^{\aleph_0}\not\to(\tau^+,\aleph_1)^2$ and the cofinality implication, which is (T2) at $\gamma=1$, $\rho_0=\aleph_1$), the simplest case with the reported theorem of Shelah, the [170] forcing "preserving $2^\tau$ up to a predetermined $\mu<\aleph_{\mathfrak c^+}$", and the [171] model from $\mathfrak c^+$ measurable cardinals are reported faithfully. The attack succeeded only against S1b: the survey's sentence is "the positive relation holds if $\aleph_{\mathfrak c^+}$ is strong limit", an instance, not the general strong-limit case (F2). The page citation is off by one page for the [170] and [171] sentences (F4).

S2: the inference "a counterexample at $\aleph_{\mathfrak c^+}$ in the [170] model, hence a consistent negative answer to the universal question" needs both hypotheses at $\lambda=\aleph_{\mathfrak c^+}$ in that model: $\operatorname{cf}\lambda=\mathfrak c^+$ is $\aleph_0$-inaccessible in ZFC ($\mu\le\mathfrak c$ gives $\mu^{\aleph_0}\le\mathfrak c^{\aleph_0}=\mathfrak c$), $\lambda$ is $\aleph_0$-inaccessible by the theorem of Shelah that the survey reports as a theorem of ZFC, and $\lambda$ is singular since $\aleph_{\mathfrak c^+}>\mathfrak c^+$ ($\mathfrak c^+$ is a successor, not a fixed point of the aleph function). The inference stands, conditional on the two reports; the section hedges on [170] only, and its item (i) then asks a question that the reported ZFC theorem answers (F6). Observed in addition: under GCH every singular cardinal is strong limit, so S3 with S5 gives the universal positive answer under GCH; with the [170] summary accurate, the universal question is independent of ZFC.

S3: the four places where a hypothesis could be missing were attacked. (a) The non-constancy of the powers: the section's reason "$2^\mu\ge\mu^+$" does not give it by itself (powers constantly equal to $\lambda^+$ from some point on are compatible with it); it holds because for a cardinal $\nu<\lambda$ the cardinal $\mu=2^\nu$ is below $\lambda$ by strong limit and $2^\mu\ge\mu^+>2^\nu$ (F3). (b) Step 1 without $\mu_0$: it needs $\kappa<\lambda$ for $\rho_i<\lambda$, given by singularity. (c) The (T1) interface: confirmed clause by clause (Premises). (d) The two cases of the cofinality relation: W1 and W2. Boundary cases: $\theta\le1$ makes the standing assumption contradictory and the conclusion trivial; $\theta=\omega$ recovers the Erdős--Dushnik--Miller relation for singular strong limit $\lambda$; $\kappa=\omega$ runs with $\mu(0)\ge\omega$. No counterexample was found: a singular strong limit $\lambda$ with $\operatorname{cf}\lambda=\aleph_1$ has $\lambda\not\to(\lambda,\aleph_1)^2$ by (T2), but there $\aleph_1\not\to(\aleph_1,\aleph_1)^2$ (Sierpiński), so the hypothesis fails too.

S4: the forward direction is (T2) at $\gamma=1$, $\rho_0=\aleph_1\ge2$ with $\lambda$ singular; the backward direction is S3 with $\theta=\aleph_1$, whose side condition $\aleph_1\le\kappa$ follows from $\kappa\to(\kappa,\aleph_1)^2$ (for $\kappa=\omega$ the coloring with every pair in color $1$ has no $0$-homogeneous set of size $\omega$ and no set of size $\aleph_1$ at all). No gap.

S5: the imported theorem cannot be checked against a held source in its regular-limit case, so it was re-derived (Premises, Claim R) for every regular $\kappa$ with $\mu^{\aleph_0}<\kappa$ for all $\mu<\kappa$, successor or limit; the successor case is Theorem 7(i) of the held Erdős--Rado 1956 paper (F5). The composition then proves the strong-limit case of the problem for every such $\lambda$, as the section says.

S6: Lemma 1.1 was tried on a non-strong-limit $\lambda$ with other choices of blocks, of the number of blocks, of the $\mu(i)$ and of the lemma's $\chi$; every choice is caught by the growth condition (W3), so S6a survives. Against S6b the witness is the constant coloring $f\equiv0$: over every parameter set every element of every block has the same type, so every block is already "a single type"; more generally $2^{|E_\alpha|}$ is an upper bound on the number of types (bound (A)), and a fiber of $a\mapsto\operatorname{tp}(a,E_\alpha)$ of size $\lambda_\alpha$ may exist for a particular $f$. What fails when $2^{|E_\alpha|}\ge\lambda$ is the pigeonhole that guarantees such a fiber, not its existence (F1). The sentence "Choosing the $\mu(i)$ below the least such $\mu$ keeps the count" was checked: with $\mu^*$ the least $\mu$ with $2^\mu\ge\lambda$ and all $\mu(i)<\mu^*$, either $\sup_i\mu(i)<\mu^*$ or $\operatorname{cf}\mu^*=\kappa$, and in both cases $\sum_{i<\alpha}\mu(i)<\mu^*$ for $\alpha<\kappa$, so the count is kept and the final set has size at most $\kappa\cdot\mu^*<\lambda$.

S7: re-derived by induction on $k$. For $f:[\lambda]^2\to k+1$ let $g$ mark the pairs of color $k$; $\lambda\to(\lambda,\aleph_1)^2$ gives a set of size $\aleph_1$ homogeneous in color $k$ or a set $H$ of size $\lambda$ on which $f$ omits $k$, and the induction hypothesis transported to $H$ finishes. The naive merge of the colors $1,\ldots,k$ into one would need $\aleph_1\to(\aleph_1)^2_k$, which is false; the induction avoids it. The attribution to the Problem 1171 page is unverified within the read set.

Premises

• (T1), the lemma interface (page Tools; its proof is another reviewer's): needs $\kappa$ infinite regular; $\mu(i)$ infinite, strictly increasing, $\mu(0)\ge\kappa$, $2^{\mu(i)}$ strictly increasing; $\lambda_i=(2^{\mu(i)})^+$; disjoint $A_i$ with $|A_i|=\lambda_i$; $F$ of finitely many places into $2$; properties with (H). In the transfer lemma: $\kappa=\operatorname{cf}\lambda$ regular infinite; the $\mu(i)$ from Step 1 without $\mu_0$; $A_i$ the blocks; $F=\tilde f$; (H) by W1. Met.
• (T2) (page Tools; its proof is another reviewer's): $\lambda$ singular, $\gamma\ge1$, $\rho_\nu\ge2$ and $\kappa\not\to(\kappa,(\rho_\nu)_\nu)^2$ give $\lambda\not\to(\lambda,(\rho_\nu)_\nu)^2$. Used at $\gamma=1$, $\rho_0=\aleph_1$ (S1, S4). Met.
• Lemma 1.1 reconstruction (premise, reviewed as of 2026-09-28T05:03:27Z): clauses (1B) and (2), bound (A), the growth condition and the role of (B) in the second thinning; read in full.
• Theorem 1.2 reconstruction (premise): Step 1 (choice of the $\mu(i)$; used without $\mu_0$, which needs only non-constancy and $\kappa<\lambda$), the sum formula, Step 5 ($\tilde f$; growth and $2^{2+\kappa}<\lambda_0$ from $\mu(0)\ge\kappa$ and $|j|\le\mu(j)$), Steps 6--7 (the argument for $g$ and for $|B|$), and (ER): for infinite $\mu$, $(2^\mu)^+\to((2^\mu)^+,\mu^+)^2$, not held; read in full.
• Corollary 1.3 reconstruction (premise): not used by the section; read in full for the folder's conventions.
• Komjáth 2025, Problem 4 commentary, printed pp. 419--420: held; the conversion and the page images were read. Interface used: the problem statement, the necessity remark, the simplest case, [45, Theorem 35.4] stated for $\aleph_{\mathfrak c^+}$, the summaries of [170] and [171].
• [45] (Erdős, Hajnal, Máté and Rado 1984), [170] and [171] (Shelah and Stanley 1987 and 1993), [158] (the theorem of Shelah): not held; used only as the survey reports them.
• The regular-case theorem of S5. Not held in the generality the section needs. The successor case is Theorem 7(i) of Erdős and Rado 1956 (printed p. 432, PDF p. 6, read on the page image): for $a\ge\aleph_0$ and $b$ minimal with $a^b>a$, $a^+\to(b,a^+)^2$; with $a^{\aleph_0}=a$ one has $b\ge\aleph_1$, hence $a^+\to(a^+,\aleph_1)^2$ after exchanging the colors. Theorem 8 there gives the regular-limit case under GCH only. Claim R supplies the general case.

Claim R. Let $\kappa$ be regular with $\mu^{\aleph_0}<\kappa$ for every cardinal $\mu<\kappa$. Then $\kappa\to(\kappa,\aleph_1)^2$.

Proof. $2^{\aleph_0}<\kappa$, so $\kappa\ge\aleph_2$. Let $f:[\kappa]^2\to2$ have no $0$-homogeneous set of size $\kappa$; a $1$-homogeneous set of size $\aleph_1$ is to be found. Write $N(x,X)=\{y\in X:y>x,\ f(\{x,y\})=1\}$.

(1) For $X\subseteq\kappa$ with $|X|=\kappa$ the set $X^-=\{x\in X:|N(x,X)|<\kappa\}$ has size below $\kappa$. Otherwise choose $y_\xi\in X^-$ by recursion on $\xi<\kappa$, above every $y_\eta$ ($\eta<\xi$) and outside $\bigcup_{\eta<\xi}N(y_\eta,X)$, a set of size below $\kappa$ by regularity; then $f(\{y_\eta,y_\xi\})=0$ for $\eta<\xi$, a $0$-homogeneous set of size $\kappa$.

(2) There is $M\prec(H(\Theta),\in)$ with $f,\kappa\in M$, $|M|<\kappa$, ${}^\omega M\subseteq M$ and $\delta:=M\cap\kappa\in\kappa$. Let $M_0\prec H(\Theta)$ be countable with $f,\kappa\in M_0$; let $M_{i+1}\prec H(\Theta)$ contain $M_i\cup{}^\omega M_i\cup\sup(M_i\cap\kappa)$ with $|M_{i+1}|\le|M_i|^{\aleph_0}+|\sup(M_i\cap\kappa)|<\kappa$, by $\aleph_0$-inaccessibility and regularity; take unions at limits; put $M=\bigcup_{i<\omega_1}M_i$. Then $|M|<\kappa$ ($\kappa$ regular and above $\aleph_1$); a countable subset of $M$ lies in some $M_i$, hence in $M_{i+1}$; and $M\cap\kappa$ is an ordinal, since $\gamma\in M_i\cap\kappa$ gives $\gamma\subseteq\sup(M_i\cap\kappa)\subseteq M_{i+1}$. By induction every countable ordinal is a countable subset of $M$ and so an element of $M$; thus $\omega_1\subseteq M$, and every sequence of countable length from $M$ lies in $M$.

(3) Choose $x_\alpha\in M\cap\kappa$ for $\alpha<\omega_1$, increasing, with $f(\{x_\beta,x_\alpha\})=1$ for $\beta<\alpha$ and $f(\{x_\alpha,\delta\})=1$. Given $\langle x_\beta:\beta<\alpha\rangle\in M$, let $\gamma=\sup_{\beta<\alpha}x_\beta\in M\cap\kappa$ and $X=\{y<\kappa:y>\gamma,\ f(\{x_\beta,y\})=1\text{ for all }\beta<\alpha\}$; $X\in M$ and $\delta\in X$ by the invariant, so $X$ is unbounded in $\kappa$ (a bound would be definable in $M$, hence below $\delta$) and $|X|=\kappa$. Put $G=X\setminus X^-\in M$; $|X\setminus G|<\kappa$ by (1). If $f(\{x,\delta\})=0$ for every $x\in G\cap\delta$, then $\delta$ lies in $Q=\{y\in X:f(\{x,y\})=0\text{ for all }x\in G\cap y\}\in M$, so $Q$ is unbounded, $|Q\cap G|=\kappa$, and $Q\cap G$ is $0$-homogeneous ($x<y$ in $Q\cap G$ gives $x\in G\cap y$ and $y\in Q$), a contradiction. So some $x\in G\cap\delta=G\cap M$ has $f(\{x,\delta\})=1$; let $x_\alpha=x$. Then $x_\alpha>\gamma$, $x_\alpha\in X$ gives $f(\{x_\beta,x_\alpha\})=1$ for $\beta<\alpha$, and $\delta$ stays in the next $X$.

The set $\{x_\alpha:\alpha<\omega_1\}$ is $1$-homogeneous of size $\aleph_1$. $\square$

The proof uses regularity, $\kappa>\aleph_1$ and $\mu^{\aleph_0}<\kappa$ for $\mu<\kappa$ only; no tree property, so it covers weakly inaccessible $\kappa$ as well as successors.

Findings

F1 (required). Location: lines 254--256, "then $2^{|E_\alpha|}\ge\lambda$ for every $\alpha$ with $|E_\alpha|\ge\mu$, and no block inside $\lambda$ can be thinned to a single type", and lines 258--259, "This is an obstruction to canonization over full parameter sets". Defect: in its strongest reading the first sentence asserts that no block has a subset of full size with one type over $E_\alpha$; the argument shows only that the counting bound (B) fails, so that the pigeonhole of the second thinning is unavailable. Witness: the constant coloring, for which every block is a single type over every parameter set; for a general $f$, bound (A) is an upper bound and a large fiber may exist. Replacement: "and the count that the second thinning uses no longer yields a fiber of size $\lambda_\alpha$" and "This is an obstruction to the counting of Lemma 1.1, not to canonization as such and not a proof of the negative relation". S6a is unaffected.

F2 (required). Location: lines 184--185, "an Erdős--Rado theorem ([45, Theorem 35.4] in the survey's numbering) gives the positive relation when $\lambda$ is strong limit", and lines 244--245, "the strong-limit case of the problem, the case Komjáth attributes to [45, Theorem 35.4]". Defect: the survey's sentence (printed p. 419) is "the positive relation holds if $\aleph_{\mathfrak c^+}$ is strong limit", an instance; the section reports it as the general strong-limit case, and [45] is not held to check the theorem's own generality. Replacement: "gives the positive relation when $\aleph_{\mathfrak c^+}$ is strong limit" and "the case whose instance $\lambda=\aleph_{\mathfrak c^+}$ Komjáth attributes to [45, Theorem 35.4]"; the Outcome's "a known theorem" then rests on S3 with S5 (and, for successor cofinalities, on Theorem 7(i) of the held Erdős--Rado 1956 paper) rather than on the survey.

F3 (suggested). Location: lines 209--210, "Since $2^\mu\ge\mu^+$ for $\mu<\lambda$, the powers below $\lambda$ are not eventually constant". Defect: the reason as written does not use strong limit and is insufficient alone (powers constantly $\lambda^+$ from some $\mu$ on satisfy $2^\mu\ge\mu^+$). Replacement: "Since $\lambda$ is strong limit, for every cardinal $\nu<\lambda$ the cardinal $\mu=2^\nu$ lies below $\lambda$ and $2^\mu\ge\mu^+>2^\nu$, so the powers are not eventually constant".

F4 (suggested). Location: line 178, "Komjáth's commentary (p. 419, read in the held conversion)". Defect: the sentences on [170] and [171] are printed on p. 420. Replacement: "(pp. 419--420, read in the held conversion)".

F5 (suggested). Location: lines 240--243, "For a regular $\aleph_0$-inaccessible $\kappa$ the hypothesis $\kappa\to(\kappa,\aleph_1)^2$ is the Erdős--Rado theorem for pairs (not held; ...)". Defect: the successor case is printed in the held Erdős--Rado 1956 paper, Theorem 7(i) (printed p. 432), and covers the simplest instance $\operatorname{cf}\lambda=\mathfrak c^+$; only the regular-limit case lacks a held source (its GCH form is Theorem 8 there). Replacement: cite Theorem 7(i) of the held paper for successor $\kappa$ and either a source or a recorded argument (Claim R above is one) for regular limit $\kappa$.

F6 (note). Location: lines 271--273, "whether the model of [170] keeps both $\aleph_0$-inaccessibility hypotheses, which decides whether the universal question is independent of ZFC". The section has already accepted, from the survey, that both hypotheses hold at $\aleph_{\mathfrak c^+}$ in ZFC, so the model keeps them automatically; the open point is whether the summary of [170] is accurate. The positive side of independence is available: under GCH every singular cardinal is strong limit, so S3 with S5 gives the universal positive answer. No text change is required.

F7 (note). Location: lines 261--264, "show that the open part is not decided by cardinal arithmetic of this kind, and any proof must use the $\aleph_0$-inaccessibility of $\lambda$ in a way that does not pass through $2^{|E_\alpha|}$", and lines 276--278, "a positive outcome would need a canonization that uses only countable parameter sets". The first sentence holds for the strong-limit dichotomy only; the survey does not say that the two models share a continuum function. The "must" and "would need" sentences are programmatic and not proved; a positive proof need not be a canonization. Suggested wording: "the strong-limit dichotomy alone does not decide the open part" and "one route would be".

F8 (note). Location: lines 195--197 (the catalog page) and lines 264--268 (the Problem 1171 page's note). Both are outside the read set; the mathematics of S7 is verified above, the attributions are not.

Verdict

• S1: refutation-failed for the problem statement and the listed records of the survey other than S1b. S1b: refuted-as-stated, as a characterization of the held survey (F2); the mathematics it points to, the strong-limit case, is proved in the section by S3 with S5.
• S2: refutation-failed, as the conditional it is; F6 records a redundancy in the follow-up item.
• S3: refutation-failed; F3 asks the stated reason to name strong limit.
• S4: refutation-failed.
• S5: refutation-failed; the imported theorem was re-derived here (Claim R) because its regular-limit case is not held; F5 corrects "not held" for the successor case.
• S6a: refutation-failed, in the form "the hypotheses of Lemma 1.1 can be met for a block decomposition of $\lambda$ with the union of the $B_\alpha$ as the homogeneous set exactly when $\lambda$ is strong limit".
• S6b: refuted-as-stated (F1); the paragraph's conclusion survives with the replacement wording.
• S7: refutation-failed.
• S8: not a claim; F7 and F8 record the wording and the unverified attributions.

Limitations: [45], [170], [171] and [158] are not held, so every sentence about them is checked only against the survey's text; the catalog page and the Problem 1171 connections page were not read; the proofs of (T1) and (T2) were not re-derived here beyond their interfaces. Two required corrections (F1, F2), three suggested (F3, F4, F5), three notes (F6, F7, F8). This focused review assigns no tier and changes no status.