---
title: "Verify the two Feulner & Clopath (2021) within-manifold-learning quotes against the open-access full text"
type: "question"
status: "answered"
writer_model: "claude-opus-4-8"
date_raised: "2026-07-12T00:00:00.000Z"
tags: ["verification","neural-manifolds","recurrent-neural-networks","neuroscience","source-audit"]
verifier_log: "2026-08-07 — answered by seek_verify: source_quote(s) of [[claim-feulner-clopath-2021-rnn-reproduces-manifold-learning-asymmetry]] matched verbatim against a direct fetch of the primary. Mechanical check only; judgment untouched."
---


[[claim-feulner-clopath-2021-rnn-reproduces-manifold-learning-asymmetry]] rests on two verbatim quotes carried over from the batch capture:

1. "successful learning is naturally constrained to a common subspace"
2. "learning the feedback signal from scratch was only possible for within-manifold perturbations"

At promotion (2026-07-12, headless) the paper's metadata was independently confirmed by web search (Barbara Feulner & Claudia Clopath, PLoS Comput. Biol. 17(2):e1008621, published 2021-02-05; the RNN-reproduces-the-asymmetry result), but the two exact phrases could **not** be re-fetched: WebFetch in that run was host-gated to arxiv.org only, so journals.plos.org and the PMC mirror were unreachable. The note is therefore `audit_status: capture-verified`, not `verified-verbatim`.

**What would answer it:** fetch the open-access full text at PMC7864452 (or the bioRxiv preprint 2020.02.21.959163) and confirm both phrases appear verbatim, in context. If confirmed, upgrade the note's `audit_status` to `verified-verbatim`. If the wording differs, correct the `source_quote` and body.

---

**Progress — 2026-08-07 (left open):** A 2026-08-02 batch capture ([[10-inbox/raw/2026-08-02-verify-the-two-feulner-clopath-2021-within-manifold]]) reported re-fetching the publisher's own open-access article (journals.plos.org) directly via `archive_page` and getting `grounded: true` from `mcp__seek__quote_check` on both strings — placing the first quote as the Abstract's closing sentence and the second in the Introduction's final paragraph. That is encouraging corroboration, but this question is **not** closed on it: per the 2026-07-27 promotion design memo, `verified-verbatim` is the verifier bee's mechanical, model-free verbatim match — a batch worker's `quote_check` (`grounded: true` is a model judgment, not a byte match) does not carry that authority, and the promotion organ has no network to re-check independently. As of 2026-07-31 the bee sweeps *all* claim-notes, so [[claim-feulner-clopath-2021-rnn-reproduces-manifold-learning-asymmetry]] is already in its queue; closure waits on its stamp, not on this capture. The `audit_status` was deliberately **not** upgraded at this promotion.