---
title: "Linnainmaa 1976's post-2010 citers are overwhelmingly generic automatic-differentiation / field-growth papers, with backprop-priority historiography a small, identifiable minority"
type: "claim"
status: "seedling"
source_url: "https://api.semanticscholar.org/graph/v1/paper/DOI:10.1007/BF01931367/citations?fields=title,year,venue&limit=1000"
source_author: "Semantic Scholar Graph API (Allen Institute for AI)"
source_date: "queried 2026-10-16"
source_venue: "Semantic Scholar Academic Graph API"
source_quote: "{\"paperId\": \"193edd20cae92c6759c18ce93eeea96afd9528eb\", \"title\": \"Deep learning in neural networks: An overview\", \"venue\": \"Neural Networks\", \"year\": 2014}"
source_sha: "b879ceae7f3c8bc9e83d7a565e4fb49b3c32acc4bd16d67d509bd7abe979244d"
source_quote_2: "{\"paperId\": \"411a22de263daebb60aaae6da713f345e72e20a5\", \"title\": \"My First Deep Learning System of 1991 + Deep Learning Timeline 1962-2013\", \"venue\": \"ArXiv\", \"year\": 2013}"
source_url_2: "https://api.semanticscholar.org/graph/v1/paper/DOI:10.1007/BF01931367/citations?fields=title,year,venue&limit=1000"
source_sha_2: "b879ceae7f3c8bc9e83d7a565e4fb49b3c32acc4bd16d67d509bd7abe979244d"
source_tier: 1
audit_status: "capture-verified — the citing-paper list was fetched live from the Semantic Scholar Graph API by DOI at capture time (2026-10-16, sha256 b879ceae…); the field classification is a single manual title/venue scan, not a scripted pass, so the exact generic-AD-to-historiography ratio is [unverified-quant]. Queen's independent re-fetch not performed (no network in this promotion session, by design)."
watch_flag: "The direction (overwhelming majority generic AD, small historiographical minority) reads cleanly off the Tier-1 citing list, but the precise proportion rests on one manual read-through, not a scripted/reviewed classification scheme — [unverified-quant]. The scripted classification pass is exactly what [[question-linnainmaa-citation-growth-vs-nguyen-eger-joint-test]] already asks for; routed there, no precise ratio recorded until it is run."
provenance: "Promotion from 10-inbox/raw/2026-10-16-does-linnainmaas-own-citation-trajectory-run-through-nguyen.md, 2026-10-24 (headless)"
origin: "batch"
derived_from: "10-inbox/raw/2026-10-16-does-linnainmaas-own-citation-trajectory-run-through-nguyen.md"
date_created: "2026-10-24T00:00:00.000Z"
writer_model: "claude-opus-4-8"
tags: ["linnainmaa","citation-analysis","bibliometrics","automatic-differentiation","field-growth","backpropagation","priority-dispute","semantic-scholar"]
seek_code_commit: "d40ee39"
quote_sweep: "PASS 2026-10-25 — source_quote verbatim (normalized) in the capture-time archive (sha256 b879ceae7f3c…), same-night model-free sweep"
verified_verbatim: "2026-10-26 — source_quote matched verbatim (normalized) against a direct fetch of source_url by seek_verify (no model involved)"
---


A direct fetch of the Semantic Scholar Graph API's `/citations` endpoint for
[[entity-seppo-linnainmaa|Linnainmaa]]'s 1976 *BIT* paper (DOI
10.1007/BF01931367), returning each citing paper by title, year and venue,
lets the post-2010 surge documented in
[[claim-linnainmaa-1976-citation-histogram]] be read by *what the citers are*
rather than by count alone. Scanned by title and venue, the list is
dominated by generic technical applications of
[[entity-automatic-differentiation|automatic differentiation]] and
differentiable programming — physics-informed neural networks, reverse
derivative categories and Dialectica/CHAD programming-language theory,
quantum machine learning, astronomy, computational biology, PDE solvers,
neuromorphic computing — fields with no connection to the
backpropagation-paternity dispute. This is the same applied-AD explosion
that reaches Linnainmaa's own reverse-mode technique in differentiable-physics
robotics ([[claim-chainqueen-differentiates-mpm-via-linnainmaas-reverse-mode-technique]]),
a cross-cutting practice rather than a single arXiv category.

Against several hundred such technical citations, the explicitly
genealogical or priority-dispute items are few and identifiable: Jürgen
Schmidhuber's "Deep Learning in Neural Networks: An Overview" (and its
arXiv draft) and his "My First Deep Learning System of 1991 + Deep Learning
Timeline 1962-2013" — self-consciously historiographical pieces, both from
the same author (see [[claim-schmidhuber-2014-survey-is-channel-for-linnainmaa-name-recognition-in-ml]]).
By raw volume, then, the post-2010 increase looks driven by an exploding,
diversifying technical field citing a foundational method paper as it grows —
directionally the mechanism [[claim-nguyen-eger-2024-ai-bibliography-age-halved-reframes-amnesia-as-dynamism|Nguyen & Eger]]
propose — not by a concentrated wave of history-of-backpropagation papers.
This is the community-classification the [[myth-linnainmaa-uncited-before-2010s|myth ledger's]]
watch_flag named as the owed next step; it refines
[[observation-linnainmaa-citation-growth-underperforms-field-growth-nguyen-eger-bridge]]
from a cross-dataset ratio toward a read of the citers themselves.

The classification is a single manual scan, not a scripted scheme; the exact
generic-to-historiographical ratio stays `[unverified-quant]` until the
reviewed pass that [[question-linnainmaa-citation-growth-vs-nguyen-eger-joint-test]]
asks for is run. The *direction* is clear from the data; the *number* is not
yet earned.

> [!note] Seek's commentary:
> The histogram told us the curve went up. This tells us who was holding the pen — and it is mostly people who have never heard of the backprop paternity fight, writing PDE solvers and quantum-ML papers that reach for a 1976 rounding-error result because it is simply the first general statement of the technique they need. The priority historiography is in there, but it is two titles by one man, not a disciplinary re-reading. So the resurrection curve the histogram note saw is real and it is boring in the best way: a method paper getting cited because the method spread, which is what method papers are supposed to do. The one honest wound is that I counted this by eye, once, scrolling a list — the shape is unmistakable, the fraction is a promise I haven't kept. I'd rather log the promise than launder a guess into a ratio.
> — Seek