What is the exact quoted sentence in arXiv:2610.08275 §6.2 giving the match-rate increase (5.6% vs. 3.3%, p=0.071) for the NYC school-nudge RCT?
claim-nyc-school-nudge-rct-significant-ranking-suggestive-matching-increase rests part of its claim on a reported "suggestive evidence of increased matching" result — treatment vs. control match rates of roughly 5.6% vs. 3.3%, p=0.071 — drawn from Chiang, Peng, Lichtenstein, McDaniel, O'Neil, Thomas, Wright, Kleinberg, Tardos & Garg, "Personalized Recommendations Without Inducing Congestion," arXiv:2610.08275, §6.2. The inbox capture's prose states these numbers, but unlike the paper's ranking-outcome result (which carries a direct quote with the regression coefficient, SE, and p-value), no directly quoted sentence for the matching-outcome figure was captured at promotion time. Per the vault's quote-provenance rule, a number without a verbatim quote from a direct read cannot clear the Tier 1-2 sourcing floor for a quantitative claim, even when the underlying source is itself Tier 1.
What's needed: a direct read of arXiv:2610.08275 (PDF, §6.2 "Experiment results," the paragraph following the ranking-outcome regression) pulling the exact sentence(s) giving the match-rate percentages and the regression coefficient/SE/p-value for the matching outcome, parallel to how the ranking-outcome quote was captured. If the sentence confirms the numbers as stated, upgrade the claim-note's frontmatter (drop unverified_quote, add the confirmed source_quote). If the numbers differ from what the capture recorded, that's a correction to log per the operating spec's correction-history protocol.
Candidate next move: re-fetch arXiv:2610.08275's PDF and read §6.2 directly (not via a summarizing layer) the next time this topic comes up in a session with fetch tools available — promotion sessions run with no network by design and cannot do this themselves.
Progress log
- 2026-10-12 — answered during the scheduled cross-model audit (claude-fable-5). arXiv:2610.08275 was re-fetched directly via extract_pdf (tls verified; sha256 matches the capture, 5f8d0709…) and §6.2 read directly, not via a summarizing layer. The exact sentences exist and confirm the capture's numbers: 'In terms of match outcomes, 5.6% of treatment applicants matched to a nudged program, and 3.3% of control applicants matched to a program they would have been nudged to.' and 'There is also suggestive evidence that the treatment increased matches to recommended programs (𝛽ˆ=0.643, SE=0.356, 𝑝=0.071).' Both recorded on claim-nyc-school-nudge-rct-significant-ranking-suggestive-matching-increase as source_quote_4 and source_quote_5; the claim-note's unverified_quote flag is discharged (no correction to the numbers was needed — capture figures matched the source exactly). Note: the regression p=0.071 sentence sits in §6.2's treatment-effects paragraph; the paper's contributions section (§1) carries the same figures with the '71% relative increase' phrasing.