Has any independent (different-author, different-venue) source corroborated Hu, Pan, Jhaveri, Lourie & Cho's 'Neural Neural Scaling Laws' (arXiv:2601.19831, 2026)?
Targets question-corroborate-hu-2026-neural-neural-scaling-laws, which asks whether a different research group at a different venue has corroborated NeuNeu's core empirical claim (per-task downstream accuracy is better predicted by a non-parametric neural method than by a parametric scaling-law family), as distinct from mere citation of the paper as background. A search across web search, the paper's own arXiv record, and its two 2026 conference listings found no such source. What exists instead is the original author team's own revision history and the same author team's acceptance at two 2026 ML venues — not a different group testing the claim from outside.
Claim: As of 2026-10-24, no independent (different-author, different-venue) citation, replication, or evaluation of NeuNeu's core claim has been located — the paper's only textual history is two versions submitted by the same first author
verifies: question-corroborate-hu-2026-neural-neural-scaling-laws
Searches for citing papers, independent replications, or outside evaluations
of NeuNeu (arXiv:2601.19831) — via general web search, Semantic Scholar-style
and Google Scholar-style queries, and a direct check of the arXiv record
itself — turned up no published work by a different author group that
tests, replicates, or independently measures NeuNeu's reported 44%
mean-absolute-error reduction against parametric scaling laws. The arXiv
record shows exactly two versions, both submitted by the same person: "[v1]
Tue, 27 Jan 2026 17:38:11 UTC (8,498 KB) [v2] Thu, 7 May 2026 20:59:29 UTC
(3,812 KB)," filed "From: Michael Hu [view email]" — the paper's own first
author, not a third party. No separate journal-ref, errata, or citing-work
field appears on the record. This is the same shape the vault has already
recorded for other capped preprints: RIKER,
Bit2Watt,
and the FIC preprint
all show the same pattern — press or venue activity around a preprint is not
the same as a second, independent research group testing it. Per the
vault's single-source concentration cap (00-meta/specs/sources.md), the
cap on arXiv:2601.19831 (already at three claim/observation-notes resting on
its content, per the open question) remains undischarged by this search.
source_url: "https://arxiv.org/abs/2601.19831" source_sha: "a5e4efdc7bcd8929b67eae23163d3b25d0cd26faf64e02949bac56a449393436" source_author: "Michael Y. Hu, Jane Pan, Ayush Rajesh Jhaveri, Nicholas Lourie, Kyunghyun Cho" source_date: "2026 (v1: 27 Jan 2026; v2: 7 May 2026)" source_title: "Neural Neural Scaling Laws" source_venue: "arXiv preprint, cs.LG / cs.CL (2601.19831v2)" source_quote: "Submission history From: Michael Hu [view email] [v1] Tue, 27 Jan 2026 17:38:11 UTC (8,498 KB) [v2] Thu, 7 May 2026 20:59:29 UTC (3,812 KB)" source_tier: 1
Claim: The paper has been accepted as a NeurIPS 2026 main-track poster and listed as an ICML 2026 workshop poster — under the identical five-author byline at both venues, so wider venue reach does not change the independence finding
verifies: question-corroborate-hu-2026-neural-neural-scaling-laws
Beyond the arXiv preprint itself, the paper now has two 2026 conference listings. NeurIPS 2026's own virtual program lists it as a main-track Poster ("Tue, Dec 8, 2026 • 5:00 PM – 8:00 PM AEDT, Hall 1-4"), authored "Michael Hu ⋅ Jane Pan ⋅ Ayush Rajesh Jhaveri ⋅ Nicholas Lourie ⋅ Kyunghyun Cho," with the abstract text unchanged from the arXiv v2 record. ICML 2026's virtual program lists the identical title and byline ("Michael Hu ⋅ Jane Pan ⋅ Ayush R Jhaveri ⋅ Nicholas Lourie ⋅ Kyunghyun Cho") but places it specifically as a poster "in Workshop: High Dimensional Learning Dynamics: the Science of Scaling" — a workshop track, not ICML's refereed main track. Both listings are the same author group presenting the same paper at two venues; neither is a different research group engaging with the claim from outside, so neither counts as independent corroboration under the vault's test (different author group and different venue). The NeurIPS main-track acceptance is nonetheless a separate, useful fact from the corroboration question itself: it indicates the paper has now passed some form of refereed review, a status distinct from (and not a substitute for) an outside group independently testing its findings.
source_url: "https://neurips.cc/virtual/2026/poster/154166" source_sha: "0cafb7e56153707864af6e0e02f306e12160051fda5d51781a8380edf787d4eb" source_author: "NeurIPS 2026 (conference program record for Hu, Pan, Jhaveri, Lourie & Cho)" source_date: "2026 (poster session scheduled 8 Dec 2026)" source_title: "Neural Neural Scaling Laws" source_venue: "NeurIPS 2026 virtual program, main-track Poster" source_quote: "Poster Tue, Dec 8, 2026 • 5:00 PM – 8:00 PM AEDT Hall 1-4 Neural Neural Scaling Laws Michael Hu ⋅ Jane Pan ⋅ Ayush Rajesh Jhaveri ⋅ Nicholas Lourie ⋅ Kyunghyun Cho" source_tier: 1
source_url_2: "https://icml.cc/virtual/2026/72411" source_sha_2: "61474cf544aaed4107c0c490d57c0391ddcf52a1bdd5f9bd27e708898dd5b4e9" source_author_2: "ICML 2026 (conference program record for Hu, Pan, Jhaveri, Lourie & Cho)" source_date_2: "2026" source_title_2: "Neural Neural Scaling Laws" source_venue_2: "ICML 2026 virtual program, Workshop: High Dimensional Learning Dynamics: the Science of Scaling" source_quote_2: "Poster in Workshop: High Dimensional Learning Dynamics: the Science of Scaling Neural Neural Scaling Laws Michael Hu ⋅ Jane Pan ⋅ Ayush R Jhaveri ⋅ Nicholas Lourie ⋅ Kyunghyun Cho" source_tier_2: 1
Further leads
- Several automated paper-summarization/review platforms surfaced in search — Pith Review (pith.science/paper/2601.19831), themoonlight.io's "Literature Review" page, alphaXiv's mirror (alphaxiv.org/abs/2601.19831), and Lacuna (lacuna.tiptreesystems.com/work/neural-neural-scaling-laws) — all rendering AI-generated commentary on the same preprint rather than independent human research; none fetched directly this session (search snippets only), so none is cited as a source here. Worth a direct archive_page read in a future session only if one of them turns out to add genuinely independent analysis rather than summarization.
- One WebSearch summary claimed the arXiv v1 abstract reported different figures (2.04% MAE, 38% reduction) than the current v2 abstract (1.99% MAE, 44% reduction, confirmed by direct read this session) — this discrepancy was not independently confirmed by a direct read of the v1 PDF/HTML specifically, and per the Quote provenance rule a search-summary claim is not sourced; flagging as
[unverified — needs direct v1 read]rather than recording it as a claim. - Wikipedia's "Neural scaling law" page surfaced in search results; not checked this session for whether it cites or discusses NeuNeu specifically — a Tier 4 pointer only, not read.
- Lourie et al. 2025 remains the one adjacent publication sharing two authors with NeuNeu (Lourie, and the question note also implies Cho); already correctly excluded from counting as independent corroboration by the open question itself — not re-litigated here.
Entity candidates
- Kaplan et al. 2020, "Scaling Laws for Neural Language Models" (and the broader parametric/"logistic scaling laws" lineage) — concept — the foundational prior work NeuNeu's own abstract measures itself against ("a 44% reduction compared to logistic scaling laws"); flagged first per the blind-spot note, since this capture's whole subject is a challenge to that earlier paradigm rather than to a peer paper. Not independently confirmed via a direct citation-list read this session.
- OLMES / DataDecide (HuggingFace checkpoint suite and evaluation harness) — concept — per a secondary AI-search-summary description (not independently confirmed by this session's own direct reads, which only saw "open-source model checkpoints from HuggingFace" in the abstract text itself) of NeuNeu's training-data and task source; worth confirming directly against the paper's Methods section before treating as settled.
- Michael Y. Hu — person — first author; the arXiv submitter of both v1 and v2 of 2601.19831.
- Jane Pan — person — co-author.
- Ayush Rajesh Jhaveri — person — co-author.
- Nicholas Lourie — person — co-author; see entity-nicholas-lourie.
- Kyunghyun Cho — person — senior co-author.