WP44: Part VII wrapper, figure batch 5 (category ITS panels, LLM co-classification)

modified figures/FIGURES.json +15 −0

		`@@ -135,5 +135,20 @@`
135	135	"sources": [
136	136	"/Users/simon-pierreboucher/wp44_uqo/results/emerging/known_set_monthly.csv"
137	137	]
	138	+ },
	139	+ {
	140	+ "file": "fig_category_its_panels",
	141	+ "caption": "Monthly series and interrupted-time-series fits for every category.",
	142	+ "sources": [
	143	+ "results/econometrics/fit_cat_*.csv"
	144	+ ]
	145	+ },
	146	+ {
	147	+ "file": "fig_llm_cooccurrence",
	148	+ "caption": "Share of each category's multi-label works that also carry the LLM category, by year.",
	149	+ "sources": [
	150	+ "/Users/simon-pierreboucher/wp44_uqo/results/graph/category_cooccurrence_yearly.csv",
	151	+ "/Users/simon-pierreboucher/wp44_uqo/results/graph/category_counts_yearly.csv"
	152	+ ]
138	153	}
139	154	]
		\ No newline at end of file

modified figures/fig_area_flow_matrix.pdf +0 −0

Binary file not shown.

modified figures/fig_area_shares.pdf +0 −0

Binary file not shown.

modified figures/fig_breaks_by_category.pdf +0 −0

Binary file not shown.

modified figures/fig_category_growth.pdf +0 −0

Binary file not shown.

modified figures/fig_category_growth.png +0 −0

Binary file not shown.

added figures/fig_category_its_panels.pdf +0 −0

Binary file not shown.

added figures/fig_category_its_panels.png +0 −0

Binary file not shown.

modified figures/fig_citation_inequality.pdf +0 −0

Binary file not shown.

modified figures/fig_cited_age.pdf +0 −0

Binary file not shown.

modified figures/fig_collaboration.pdf +0 −0

Binary file not shown.

modified figures/fig_country_shares.pdf +0 −0

Binary file not shown.

modified figures/fig_cross_area.pdf +0 −0

Binary file not shown.

modified figures/fig_emerging_metrics.pdf +0 −0

Binary file not shown.

modified figures/fig_field_inflows.pdf +0 −0

Binary file not shown.

modified figures/fig_its_total.pdf +0 −0

Binary file not shown.

modified figures/fig_known_set.pdf +0 −0

Binary file not shown.

added figures/fig_llm_cooccurrence.pdf +0 −0

Binary file not shown.

added figures/fig_llm_cooccurrence.png +0 −0

Binary file not shown.

modified figures/fig_placebo_total.pdf +0 −0

Binary file not shown.

modified figures/fig_prisma_flow.pdf +0 −0

Binary file not shown.

modified figures/fig_types_openness.pdf +0 −0

Binary file not shown.

modified figures/fig_venue_concentration.pdf +0 −0

Binary file not shown.

modified figures/fig_volume_monthly_raw.pdf +0 −0

Binary file not shown.

modified figures/fig_volume_yearly.pdf +0 −0

Binary file not shown.

modified figures/make_figures.py +41 −1

		`@@ -89,7 +89,7 @@ def fig_collab(R, out):`
89	89
90	90	def main():
91	91	ap = argparse.ArgumentParser(); ap.add_argument("--root", required=True); a = ap.parse_args(); out = f"{a.root}/figures"; os.makedirs(out, exist_ok=True)
92		− for f in (fig_prisma, fig_yearly_volume, fig_monthly, fig_types_sources, fig_collab) + _BATCH2 + _BATCH3 + _BATCH4:
	92	+ for f in (fig_prisma, fig_yearly_volume, fig_monthly, fig_types_sources, fig_collab) + _BATCH2 + _BATCH3 + _BATCH4 + _BATCH5:
93	93	try:
94	94	f(a.root, out); print("ok", f.__name__)
95	95	except Exception as e:
		`@@ -333,5 +333,45 @@ def fig_known_set(R, out):`
333	333	_BATCH4 = (fig_category_growth, fig_category_shares, fig_breaks_heatmap, fig_flow_matrix, fig_cross_area, fig_emerging, fig_known_set)
334	334
335	335
	336	+
	337	+
	338	+# ---------------- batch 5: category monthly series with ITS fits; co-classification with LLM-related categories ----------------
	339	+def fig_category_its_panels(R, out):
	340	+ import glob as _g
	341	+ names, areas = _cat_names(R); files = sorted(_g.glob(f"{R}/results/econometrics/fit_cat_*.csv"))
	342	+ if not files:
	343	+ return
	344	+ files = [f for f in files if os.path.basename(f)[8:-4] in names]; n = len(files); cols = 4; rows = int(np.ceil(n / cols))
	345	+ fig, axes = plt.subplots(rows, cols, figsize=(7.6, 1.55 * rows), sharex=True); axes = np.atleast_2d(axes)
	346	+ for ax, f in zip(axes.ravel(), files):
	347	+ d = pd.read_csv(f); x = pd.to_datetime(d.month + "-01"); cid = os.path.basename(f)[8:-4]
	348	+ ax.plot(x, d.y / 1e3, color=style.CATEGORICAL[0], lw=0.8); ax.plot(x, d.fitted / 1e3, color=style.CATEGORICAL[1], lw=1.1)
	349	+ ax.axvline(pd.Timestamp("2022-11-30"), color=style.EVENT, lw=0.6, ls="--"); ax.set_title(names.get(cid, cid), fontsize=6.5, loc="left"); ax.tick_params(labelsize=5.5); ax.grid(True)
	350	+ for ax in axes.ravel()[n:]:
	351	+ ax.axis("off")
	352	+ fig.suptitle("Monthly works per category (thousands; month-precision records) with ITS fits at 2022-12", fontsize=8, x=0.02, ha="left")
	353	+ style.save(fig, f"{out}/fig_category_its_panels.pdf"); reg("fig_category_its_panels", "Monthly series and interrupted-time-series fits for every category.", ["results/econometrics/fit_cat_*.csv"])
	354	+
	355	+
	356	+def fig_llm_cooccurrence(R, out):
	357	+ p = f"{R}/results/graph/category_cooccurrence_yearly.csv"; q = f"{R}/results/graph/category_counts_yearly.csv"
	358	+ if not (os.path.exists(p) and os.path.exists(q)):
	359	+ return
	360	+ names, areas = _cat_names(R); co = pd.read_csv(p); cc = pd.read_csv(q); tot = cc.groupby(["year", "category_id"]).n_multilabel.sum()
	361	+ fig, ax = plt.subplots(figsize=(6.8, 3.6)); shown = 0
	362	+ for target in ("llm", "ai_agents", "generative_ai"):
	363	+ sub = co[(co.cat_a == target) \| (co.cat_b == target)].copy(); sub["other"] = np.where(sub.cat_a == target, sub.cat_b, sub.cat_a)
	364	+ for other in ("cybersecurity", "software_engineering", "hci", "information_retrieval", "robotics", "computational_biology", "computer_vision"):
	365	+ s = sub[sub.other == other].groupby("year").n.sum(); base = tot.xs(other, level=1).reindex(s.index)
	366	+ share = (s / base * 100).dropna(); share = share[(share.index >= 2018) & (share.index <= 2025)]
	367	+ if target == "llm" and len(share):
	368	+ ax.plot(share.index, share.values, marker="o", ms=3, color=style.CATEGORICAL[shown % 8], label=names.get(other, other), lw=1.1); shown += 1
	369	+ ax.set_ylabel("Share of the category's works also classified as LLM (%)"); ax.set_title("Co-classification with Large Language Models, by category", loc="left"); ax.legend(fontsize=6.5, ncol=2)
	370	+ style.save(fig, f"{out}/fig_llm_cooccurrence.pdf"); reg("fig_llm_cooccurrence", "Share of each category's multi-label works that also carry the LLM category, by year.", [p, q])
	371	+
	372	+
	373	+_BATCH5 = (fig_category_its_panels, fig_llm_cooccurrence)
	374	+
	375	+
336	376	if __name__ == "__main__":
337	377	main()

modified literature_synthesis/agent_E/claims.csv +1 −1

		`@@ -9,7 +9,7 @@ E-07,Post-period CAGR (2023–2025) versus baseline (2018–2022): LLM 1.085 vs`
9	9	E-08,The LLM category's share of eligible works rose from 0.33 % in 2018 to 5.06 % in 2025.,established,corpus_statistic,,,"results/bibliometrics/cagr_by_category.csv (share_2018, share_2025); tab_ai_growth",0.95,machine-verified,
10	10	E-09,For the LLM series the ITS at 2022-12 gives a level change of 0.106 (HAC p 0.075) and a slope change of 0.034 per month (p ~1e-60); placebo p 0.938 (level) and 0.200 (slope); Bai–Perron selects one break at 2022-07; PELT one change point at 2023-10.,established,corpus_statistic,,,results/econometrics/breaks.csv row cat_llm; its_results.json series.cat_llm; macros ItsSeriesCatLlm*; tab_ai_breaks,0.95,machine-verified,
11	11	E-10,"Multimodal, generative AI, agents, RL, ML theory and AI safety show significant positive slope changes at 2022-12 with no positive level jump (agents, RL and AI safety have negative level changes); DL architectures shows no significant slope change (−0.002, p 0.136).",established,corpus_statistic,,,results/econometrics/breaks.csv rows cat_*; tab_ai_breaks,0.95,machine-verified,
12		−E-11,"Level effects at the GPT-3 (2020-06) and DALL-E 2 (2022-05) event dates are negative for all nine AI categories, while those at GPT-4 (2023-04) and o1 (2024-09) are positive and significant for LLM, multimodal and generative AI.",established,corpus_statistic,,,"results/econometrics/breaks.csv (gpt3_level, dalle2_level, gpt4_level, o1_level and p columns); tab_ai_altdates",0.9,machine-verified,Each alternative date is a separate ITS fit.
	12	+E-11,"Level effects at the GPT-3 (2020-06) and DALL-E 2 (2022-05) event dates are negative for all nine AI categories, while those at GPT-4 (2023-04) and o1 (2024-10) are positive and significant for LLM, multimodal and generative AI.",established,corpus_statistic,,,"results/econometrics/breaks.csv (gpt3_level, dalle2_level, gpt4_level, o1_level and p columns); tab_ai_altdates",0.9,machine-verified,Each alternative date is a separate ITS fit.
13	13	E-12,The post-2022 acceleration of AI publication is a trend change already under way in mid-2022 that steepened in 2023–2024; the monthly series cannot attribute it to a single product release.,established,corpus_statistic,,,"results/econometrics/breaks.csv; its_results.json (placebo, bai_perron, pelt)",0.85,machine-verified,Interpretive statement about the estimates; no causal claim.
14	14	E-13,"LLM works directed 69.6 % of in-corpus references to NLP in 2019 and 68.6 % in 2022, but 34.2 % in 2025, when 28.4 % went to the LLM category itself, 4.3 % to HCI and 3.4 % to software engineering.",established,corpus_statistic,,,results/graph/category_flows_yearly.csv (citing_cat llm); tab_ai_flows,0.95,machine-verified,
15	15	E-14,Generative-AI references to computer vision fell from 57.9 % (2022) to 33.8 % (2025) while references to LLM rose to 13.7 % and to HCI to 6.6 %; AI-safety references to LLM rose from 0.1 % (2019) to 10.0 % (2025) and to HCI from 9.1 % to 11.5 %.,established,corpus_statistic,,,results/graph/category_flows_yearly.csv; tab_ai_flows,0.95,machine-verified,

modified literature_synthesis/agent_E/review.md +1 −1

		@@ -12,7 +12,7 @@ Date: 2026-11-03. Inputs: `literature_synthesis/landmarks/ai_ml_landmarks.csv` (
12	12
13	13	Volume. The LLM category grows from a primary-category count of a few thousand works in 2018–2022 to tens of thousands in 2025, with a post-period CAGR of 1.085 against 0.189 in the baseline (`cagr_by_category.csv`); generative AI (0.870 vs 0.296), multimodal AI (0.869 vs 0.284), foundation models (0.802 vs 0.661) and AI safety (0.647 vs 0.294) follow; reinforcement learning (0.252 vs 0.208) and ML theory (0.197 vs 0.104) accelerate modestly; DL architectures decelerate slightly (0.155 vs 0.167); AI agents reverse from a negative baseline CAGR (−0.045) to 0.329 [E-07, corpus_statistic]. The LLM share of eligible works rises from 0.33 % (2018) to 5.06 % (2025) [E-08]. Multi-label counts (`tab_ai_counts`) are three to ten times larger than primary counts because the AI categories overlap heavily; the ordering of growth rates is the same under both definitions.
14	14
15		−Structural change. The interrupted time series at 2022-12 (`breaks.csv`) shows, for the LLM series, a level change of 0.106 log points (HAC p = 0.075) and a slope change of 0.034 log points per month (p ≈ 10⁻⁶⁰), with a placebo p-value of 0.938 for the level and 0.200 for the slope; the Bai–Perron procedure selects a single break at 2022-07 and PELT a change point at 2023-10 [E-09]. The same pattern — no significant level jump at the reference month, a strongly significant slope change, placebo p-values that do not single out December 2022, data-driven breaks before or well after it — holds for multimodal (slope 0.024; break 2022-04) and generative AI (slope 0.019; break 2021-11). AI agents (level −0.210, p = 0.028; slope 0.040), reinforcement learning (level −0.164, p = 0.013) and AI safety (level −0.206, p = 0.002) show negative level changes with positive slope changes, i.e. a dip followed by acceleration; DL architectures is the only AI category with no significant slope change (−0.002, p = 0.136) [E-10]. At the alternative event dates the picture is monotone: level effects at the GPT-3 (2020-06) and DALL-E 2 (2022-05) dates are negative for every AI category, while those at the GPT-4 (2023-04) and o1 (2024-09) dates are positive and significant for LLM, multimodal and generative AI (`tab_ai_altdates`) [E-11]. The conservative reading is that the AI literature's acceleration is a trend change that was already under way in mid-2022 and steepened through 2023–2024; the monthly series cannot attribute it to a single product release, and the manuscript must not do so [E-12, established as a statement about the estimates].
	15	+Structural change. The interrupted time series at 2022-12 (`breaks.csv`) shows, for the LLM series, a level change of 0.106 log points (HAC p = 0.075) and a slope change of 0.034 log points per month (p ≈ 10⁻⁶⁰), with a placebo p-value of 0.938 for the level and 0.200 for the slope; the Bai–Perron procedure selects a single break at 2022-07 and PELT a change point at 2023-10 [E-09]. The same pattern — no significant level jump at the reference month, a strongly significant slope change, placebo p-values that do not single out December 2022, data-driven breaks before or well after it — holds for multimodal (slope 0.024; break 2022-04) and generative AI (slope 0.019; break 2021-11). AI agents (level −0.210, p = 0.028; slope 0.040), reinforcement learning (level −0.164, p = 0.013) and AI safety (level −0.206, p = 0.002) show negative level changes with positive slope changes, i.e. a dip followed by acceleration; DL architectures is the only AI category with no significant slope change (−0.002, p = 0.136) [E-10]. At the alternative event dates the picture is monotone: level effects at the GPT-3 (2020-06) and DALL-E 2 (2022-05) dates are negative for every AI category, while those at the GPT-4 (2023-04) and o1 (2024-10) dates are positive and significant for LLM, multimodal and generative AI (`tab_ai_altdates`) [E-11]. The conservative reading is that the AI literature's acceleration is a trend change that was already under way in mid-2022 and steepened through 2023–2024; the monthly series cannot attribute it to a single product release, and the manuscript must not do so [E-12, established as a statement about the estimates].
16	16
17	17	Knowledge flows. In-corpus references of LLM works went to NLP for 69.6 % in 2019 and 68.6 % in 2022, but only 34.2 % in 2025, when 28.4 % go to the LLM category itself, 4.3 % to HCI and 3.4 % to software engineering (`category_flows_yearly.csv`) [E-13]. Generative-AI works cited computer vision for 53.6 % (2019) and 57.9 % (2022) of their in-corpus references and 33.8 % in 2025, when 13.7 % go to LLM and 6.6 % to HCI; AI-safety works cite LLM for 10.0 % and HCI for 11.5 % in 2025 against 0.1 % and 9.1 % in 2019 [E-14]. Reinforcement learning moves the other way — its share of references to robotics rises from 14.4 % to 23.1 % and to LLM stays below 1 % — and ML theory and DL architectures are essentially unchanged [E-15]. The LLM category has thus become a citation hub for generative AI, safety and agents while remaining peripheral to RL, theory and architectures; this is the corpus-level answer to RQ5 at the level of categories, and it is consistent with but does not establish the "foundation model" convergence thesis (Bommasani et al. 2021) [E-16, contested].
18	18

added literature_synthesis/agent_G/review.md +256 −0

		`@@ -0,0 +1,256 @@`
	1	+# Agent G — Theory Specialist review (Phase 5), WP44
	2	+
	3	+Scope: algorithms and data structures, computational complexity, programming languages, compilers, formal verification,
	4	+software engineering (areas `theory` and `software_pl` of `taxonomy/ontology.yaml`). Date: 2026-11-03.
	5	+
	6	+Inputs read: `literature_synthesis/landmarks/theory_landmarks.csv` (50 candidates) and `software_pl_landmarks.csv`
	7	+(100 candidates) with their `*_abstracts.jsonl`; `results/graph/category_counts_yearly.csv`, `category_flows_yearly.csv`,
	8	+`category_cooccurrence_yearly.csv`; `results/bibliometrics/cagr_by_category.csv`; `results/econometrics/breaks.csv`
	9	+(rows `cat_algorithms`, `cat_computational_complexity`, `cat_programming_languages`, `cat_compilers`,
	10	+`cat_formal_verification`, `cat_software_engineering`); `manuscript/references.bib`. No knowledge-graph extraction file
	11	+exists for these areas (`data/processed/graph/extraction/` is absent), so no extracted claim or method node was used.
	12	+Evidence levels are stated per statement: metadata (title, venue, type, year, identifiers), abstract (the abstract
	13	+recorded in OpenAlex), corpus statistic (a result file). No full text was read. Companion artefacts:
	14	+`claims.csv` (G-01…G-24), `contributions.csv` (150 rows, one per candidate, with an audit verdict), `open_problems.csv`
	15	+(G-P1…G-P10), `verification_lookups_raw.txt` (raw Crossref/DataCite/arXiv/OpenAlex responses), and the generated tables
	16	+`manuscript/tables/tab_theory_*.tex`.
	17	+
	18	+## 1. The candidate landmark set: an audit before a synthesis
	19	+
	20	+The protocol selects landmarks as the works with the highest field-normalised 36-month citation score (NCS36) within
	21	+each primary category, with a floor of fifteen per category, plus the 200 most cited works overall
	22	+(`literature_synthesis/select_landmarks.py`). For the six categories reviewed here this produced 25 candidates each,
	23	+150 in total. Before any of them can be discussed as a landmark, the list has to be audited, because its composition is
	24	+not what the selection rule intends (claim G-01; `contributions.csv`, column `verdict`):
	25	+
	26	+\| verdict \| n \| meaning \|
	27	+\|---\|---\|---\|
	28	+\| in_scope_research \| 29 \| a research contribution inside the category's definition \|
	29	+\| in_scope_survey \| 8 \| survey, handbook or competition report inside the category \|
	30	+\| misclassified \| 33 \| outside the category and outside the theory/software scope \|
	31	+\| dissertation \| 27 \| thesis (in scope), selected through the normalisation of small category×year cells \|
	32	+\| reprint_misdated \| 24 \| a classic re-deposited (repository, arXiv, anthology) under a recent record date \|
	33	+\| textbook \| 22 \| textbook, chapter or monograph \|
	34	+\| metaheuristic \| 6 \| metaphor-based metaheuristic optimiser (in scope, contested genre) \|
	35	+\| duplicate \| 1 \| second record of the Handbook of Satisfiability \|
	36	+
	37	+Four mechanisms explain the composition, and each is a finding about the instrument rather than about the field.
	38	+
	39	+Identifier collisions (claim G-02, machine-verified). Seven records carry OpenAlex identifiers that, resolved live on
	40	+2026-11-03, return a different and far more cited work. The record titled "Exploiting Generative AI to Scale up
	41	+Intelligent Tutoring Systems" (DOI 10.4230/lipics.itp.2023.19, `c36` = 72,130) resolves to Attention Is All You Need;
	42	+its abstract in the corpus is that of MizAR 60 for Mizar 50 (ITP 2023), which is what the DOI actually denotes. The
	43	+ECOOP 2025 paper on functionality-specific vulnerabilities (`c36` = 15,245) resolves to the graph-convolutional-network
	44	+preprint; the SEA 2024 paper on GNN branching for maximum independent set (`c36` = 5,183) resolves to GraphSAGE; the WADS
	45	+2025 Kd-tree paper (`c36` = 2,355) resolves to the variational-autoencoder preprint; the SAIA 2024 robustness framework
	46	+resolves to the Madry et al. adversarial-training preprint; the ESEM 2026 record resolves to the Codex paper; the DX
	47	+2024 answer-set-programming paper resolves to an "Uncertainty-aware RSS" preprint. In every case the recorded `c36`
	48	+matches the citation count of the other work, so the citation windows attached to these records are not theirs. The
	49	+mechanism (identifier merge or reuse between the 2026-10-14 snapshot and the live database) is inferred, not observed;
	50	+the consequence is observed: the four most "cited" candidates of the software_pl area and the first candidate of the
	51	+theory area owe their rank to another work's citations.
	52	+
	53	+Generic-title chapters. Book chapters titled "Introduction to Algorithms" (2020, 2021, 2023), "Particle Swarm
	54	+Optimization" (2019), "Introduction to Software Testing" (2021), "Introduction to Graph Theory" (2023) and Royce's 1970
	55	+paper in a 2021 anthology carry hundreds to thousands of citations in their 36-month window. A live lookup confirms the
	56	+counts (the PSO chapter has 6,321 citations; the 2023 "Introduction to Algorithms" chapter 1,645). The plausible reading
	57	+is that citations intended for the canonical work (the Cormen et al. textbook, the Kennedy–Eberhart PSO paper, Royce
	58	+1970) are attributed by the indexer to a same-titled chapter; this is an interpretation, recorded as such (`claims.csv`
	59	+G-01 notes).
	60	+
	61	+Re-deposited classics (24 records). Dijkstra's 1959 note appears as a 2022 ACM Books chapter; Turing's 1936 paper as a
	62	+2021 MIT Press chapter; the 1994 survey of decidability for Petri nets as a 2024 arXiv posting; Lynch and Tuttle's
	63	+introduction to I/O automata as a 2026 journal article; Cormode's 2003 thesis "Sequence distance embeddings" and the SODA
	64	+2000 "Communication complexity of document exchange" as 2026 Warwick repository records; three UNSW theses of the L4/seL4
	65	+verification lineage as 2022 records; the 2018 NeurIPS paper on online algorithms with predictions as a 2024 arXiv
	66	+posting. The year field of these records is the deposit year, so they enter the "post" period and inflate its apparent
	67	+landmark count (`\ThyLandTheoryDatedTwentysix` = 10 of 50 theory candidates are dated 2026).
	68	+
	69	+Lexical misclassification (33 records). The computational_complexity list contains fifteen works from psychology,
	70	+management, education and metrology (UTAUT adoption models, PLS-SEM studies of SMEs and students, "Complexity Theory and
	71	+the Social Sciences", fringe-projection profilometry), captured by the phrase "complexity" (claim G-03). The compilers
	72	+list contains ten works captured by "just-in-time" (adaptive interventions in mental health), "compiling" (a guide to
	73	+literature reviews), "polyhedral" (power-system dispatch) and generic education studies (claim G-20). The
	74	+programming_languages list includes a Nature paper on Calabi–Yau manifolds and a debris-flow hazard study. The
	75	+dissertation and misclassification problems interact: the NCS36 normalisation divides by the expected citations of a
	76	+category×year×type cell, and small cells of theses or of non-CS articles produce large normalised scores.
	77	+
	78	+What remains after the audit is a set of 29 research works and 8 surveys or reports, listed in
	79	+`tab_theory_landmarks_discussed.tex`. Sixteen of the 29 are post-period works. They are discussed below by category,
	80	+together with the verified references of the pre-2022 lineage from `references.bib` and the Part VII-a bibliography
	81	+(`manuscript/references_part07a.bib`, 65 entries, each resolved by a single live lookup). The major theoretical results of
	82	+the window are absent from the candidate list (claim G-14) and are brought in through verified records, not through the
	83	+corpus ranking.
	84	+
	85	+## 2. What the corpus statistics say about the six categories
	86	+
	87	+Volume. Multi-label counts grew in every category after 2022 (Table `tab_theory_counts`): the 2023→2025 compound
	88	+annual growth is 25.5% for software engineering, 22.7% for computational complexity, 22.4% for compilers, 14.2% for
	89	+algorithms, 13.1% for formal verification and 11.5% for programming languages, against 0.9%, 1.9%, −0.3%, 1.6%, −5.1%
	90	+and −2.4% over 2018→2022. The comparison with the corpus-wide multi-label growth (6.1% → 26.0%) shows that this is mostly
	91	+the corpus-wide acceleration (claim G-05), and the primary-label shares of all six categories fell between 2018 and 2025
	92	+(algorithms 5.83% → 3.79%, software engineering 5.81% → 4.07%, formal verification 1.06% → 0.52%, programming languages
	93	+0.88% → 0.47%, computational complexity 0.77% → 0.47%, compilers 0.15% → 0.09%; claim G-06). The theory and software
	94	+categories are growing in absolute terms and shrinking in relative terms. Two cautions: the post-period CAGR spans two
	95	+years and is exposed to indexing lag (Part IV, truncation section); and `results/bibliometrics/cagr_by_category.csv`
	96	+reports primary counts about three times smaller than `category_counts_yearly.csv` for these categories (algorithms
	97	+2018: 9,716 against 29,148), so it predates the current classification and its CAGR values are not used here (claim
	98	+G-07; the generator `bibliometrics/04_diversity_cagr.py` reads the counts file, so re-running it resolves the issue).
	99	+
	100	+Structural change. The interrupted-time-series slope change at 2022-12 is positive and HAC-significant for all six
	101	+series (1.0% to 1.8% per month), but no series passes the placebo test at the 5% level: the placebo slope p-values are
	102	+0.062 (algorithms), 0.062 (computational complexity), 0.154 (software engineering), 0.215 (compilers), 0.246 (formal
	103	+verification) and 0.431 (programming languages). The 2022-12 slope change is therefore not distinguishable from
	104	+pseudo-events placed in the baseline (claim G-08). Level changes at 2022-12 are negative or null in five categories
	105	+(formal verification −10.9%, p = 0.029; computational complexity −6.8%, p < 0.001; algorithms −4.5%, p = 0.039) and
	106	+positive only in software engineering (+4.8%, p = 0.124), which is also the only category with a significant positive
	107	+level change at the GPT-4 date (0.102, p = 0.003; claim G-09). Bai–Perron dates a single break at 2023-04 for
	108	+algorithms, computational complexity and compilers, two breaks for programming languages (2022-04, 2024-06) and formal
	109	+verification (2022-02, 2024-07), and 2022-07 for software engineering. None of this supports a causal reading, and the
	110	+negative level changes are at least partly the seasonality of proceedings dates.
	111	+
	112	+Composition. Co-classification with the LLM category rose in every category between 2022 and 2025: software
	113	+engineering 4.0% → 21.2% of its works, programming languages 4.2% → 18.3%, compilers 3.1% → 14.9%, formal verification
	114	+0.7% → 7.9%, algorithms and computational complexity 1.3% → 5.3% (claim G-10; `tab_theory_llm_cooccurrence`). The
	115	+gradient runs from the applied, code-facing categories to the mathematical ones. Reference flows tell a more conservative
	116	+story: in the post period, software-engineering works direct 2.5% of their categorised references to LLM works,
	117	+computational complexity 1.8%, compilers 1.2%, programming languages 0.7%, formal verification 0.4% and algorithms 0.3%,
	118	+while self-citation shares are nearly unchanged (algorithms 65.1% → 65.0%, software engineering 64.6% → 62.4%, formal
	119	+verification 49.9% → 47.0%; claim G-11). Two flow changes deserve notice. Computational complexity's self-citation share
	120	+fell from 30.0% to 20.6% while its references to HCI doubled (3.0% → 6.7%); given the candidate list, this is more
	121	+plausibly the social-science contamination of the category than a change in theoretical research (claim G-12,
	122	+contested). Formal verification's inbound citations from computer vision doubled (6.7% → 13.8%), consistent with the
	123	+neural-network verification literature being cited by robustness work (claim G-13, contested; no paper-level breakdown).
	124	+
	125	+## 3. Algorithms and data structures
	126	+
	127	+Established. The window contains results that the field itself regards as landmarks and that the verified bibliography
	128	+records: maximum flow and minimum-cost flow in almost-linear time (Chen et al., FOCS 2022), negative-weight single-source
	129	+shortest paths in near-linear time (Bernstein, Nanongkai and Wulff-Nilsen, FOCS 2022; JACM and CACM republications
	130	+2025), the first directed SSSP algorithm below the sorting barrier (Duan et al., STOC 2025, best paper; JACM 2026), and
	131	+optimal bounds for open addressing without reordering (Farach-Colton, Krapivin and Kuszmaul, FOCS 2024), which overturned
	132	+a conjecture dating from the 1980s. None of them is in the candidate list (claim G-14). The corpus records the programme
	133	+of algorithms with predictions through a 2024 arXiv re-posting of the 2018 NeurIPS paper on ski rental and
	134	+non-clairvoyant scheduling (abstract: algorithms "oblivious to the performance of the predictor, improve with better
	135	+predictions, but do not degrade much if the predictions are poor"); the programme's canonical references are the JACM
	136	+competitive-caching paper and the CACM overview (claim G-22). The fine-grained-complexity programme that organised
	137	+baseline-period algorithms research (Part II) is represented by the ICM 2018 survey.
	138	+
	139	+Contested. Six of the 25 candidates are metaphor-based metaheuristics (Arithmetic Optimization Algorithm 2021, Aquila
	140	+Optimizer 2021, artificial lemming 2025, glider snake 2026, gray langurs 2026, starfish 2026), published in engineering
	141	+and AI-review journals and benchmarked on CEC function suites. Their abstracts promise balanced "exploration and
	142	+exploitation" and superiority over "13 state-of-the-art algorithms". The operational-research literature has argued since
	143	+Sörensen's "Metaheuristics—the metaphor exposed" that the genre re-describes known methods under new metaphors, a
	144	+position restated in the 2021 "call for action" and in the 2022 analysis of six such algorithms. The corpus shows the
	145	+genre's weight in the highly cited stratum of the category (claim G-04); it does not evaluate the algorithms. The two
	146	+post-period research candidates that survive the audit are modest: Kd-trees extended to Bregman divergences (WADS 2025,
	147	+with a C++ implementation) and GNN-selected branching inside branch-and-reduce for maximum independent set (SEA 2024);
	148	+both carry citation counts that belong to other works.
	149	+
	150	+AI-discovered algorithms. AlphaTensor (matrix multiplication, 2022), AlphaDev (sorting routines, 2023), FunSearch
	151	+(cap sets, 2023) and AlphaEvolve (2025, preprint) report improvements on specific instances; no independent replication
	152	+appears in the verified bibliography (claim G-23, contested). Whether these results change how algorithms are designed is
	153	+an open problem (G-P9), as is the diffusion of the 2022–2025 breakthroughs into practice (G-P6).
	154	+
	155	+## 4. Computational complexity
	156	+
	157	+The category is the least precise of the six (fifteen misclassified candidates). The two in-scope research candidates
	158	+are the STOC 2019 non-elementary lower bound for Petri-net reachability, whose abstract describes the problem as open
	159	+since the 1960s and the previous best lower bound as Lipton's 1976 exponential space, and a 2026 TCS paper on SAT in
	160	+spiking neural P systems (no abstract). The reachability story was completed in the baseline period: Czerwiński and
	161	+Orlikowski (FOCS 2021) and Leroux (FOCS 2021) established Ackermann-completeness, matching the Leroux–Schmitz upper bound
	162	+(claim G-15). In the window, the result the field regards as the main complexity-theoretic event is Williams's STOC 2025
	163	+simulation of time t in space O(√(t log t)) (JACM 2026), which improves the Hopcroft–Paul–Valiant bound after fifty
	164	+years; Kelley and Meka's strong bounds for 3-progressions (FOCS 2023) are the main additive-combinatorics result.
	165	+Neither appears in the candidate list; both are recorded through verified DOIs.
	166	+
	167	+The boundary with language models is where the category's post-2022 growth is most visible in the co-classification
	168	+statistics, yet the relevant works are mostly classified elsewhere: the proof that log-precision transformers lie in
	169	+uniform TC⁰ (TACL 2023) and the survey of formal-language results for transformers (TACL 2024) bound what a single forward
	170	+pass can compute and motivate chain-of-thought as a route to greater expressive power (claim G-21, speculative as to its
	171	+place in the corpus). Textbooks on Kolmogorov complexity (4th edition, 2019) and on parameterized complexity dominate the
	172	+rest of the list.
	173	+
	174	+## 5. Programming languages
	175	+
	176	+The baseline lineage (Part II) is Rust and its semantic foundation (RustBelt, 2017; the CACM account of safe systems
	177	+programming, 2021), the Lean 4 prover-as-language and mathlib, and the verified artefacts CompCert and seL4. The
	178	+candidate list records this lineage only through re-deposited theses of the L4/seL4 school (separation logic for virtual
	179	+memory, formal memory models for C systems code, the clustered multikernel), all dated 2022 by their deposit year. In the
	180	+window, Verus (SOSP 2024) extends the lineage with a Rust-based verification framework for systems code (claim G-24).
	181	+
	182	+The two post-period research candidates mark the entry of language models into the community's top venue (claim G-16):
	183	+LLift (OOPSLA 2024) combines static analysis with an LLM to find use-before-initialisation bugs in the Linux kernel, and
	184	+its abstract reports four previously unknown bugs acknowledged by the kernel community; WhiteFox (OOPSLA 2024) uses an
	185	+LLM that reads optimisation source code to generate test programs for compiler optimisations. Both report bugs found,
	186	+neither reports guarantees. The contrast with translation validation (Alive2, PLDI 2021, bounded) and with
	187	+Csmith-style differential testing (PLDI 2011) is the natural axis of comparison and an open problem (G-P4). ZEUS (NDSS
	188	+2018), which combines abstract interpretation and constrained Horn clauses to check smart contracts, is the baseline
	189	+representative of program analysis for a new target. Co-classification of programming-languages works with LLM reached
	190	+18.3% in 2025, the second highest of the six categories; references to LLM works are 0.7% of the category's references.
	191	+
	192	+## 6. Compilers
	193	+
	194	+The category is the smallest (0.09% of primary labels in 2025) and the most lexically contaminated after complexity. Its
	195	+in-scope candidates trace a coherent line. Baseline: MLIR (CGO 2021), the multi-level infrastructure that reduced the
	196	+cost of domain-specific compilers; Tiramisu (CGO 2019), a polyhedral compiler with an explicit scheduling language;
	197	+ExaStencils (2020), a layered DSL for multigrid solvers; and two surveys of machine learning in compiler optimisation
	198	+(Proc. IEEE 2018; CSUR 2018 on autotuning and phase ordering), which show that learned compiler heuristics are a
	199	+pre-2018 programme. Post-period: LLM4VV (FGCS 2024), an LLM-generated validation suite for OpenACC compilers, and LLM
	200	+Compiler (CC 2025), foundation models trained on 546 billion tokens of LLVM-IR and assembly, whose abstract reports 77%
	201	+of the optimising potential of an autotuning search for code size. The shift is from learning a heuristic inside a
	202	+compiler to pre-training a model on compiler artefacts; what it leaves open is correctness, which the compiler community
	203	+has historically handled by differential testing and translation validation. Co-classification with LLM reached 14.9%
	204	+in 2025; references to LLM works are 1.2%. The deep-learning compiler stack recorded in `references.bib` (TVM, Triton,
	205	+PyTorch 2) is the other post-2018 development and is reviewed in Part VI.
	206	+
	207	+## 7. Formal verification
	208	+
	209	+This is the category with the most in-scope research candidates (eight) and the clearest internal structure. Tools and
	210	+competitions: cvc5 (TACAS 2022), the SV-COMP 2024 report (76 tools, 30,300 C tasks), Marabou (CAV 2019), the Handbook of
	211	+Satisfiability (2nd ed., 2021). Neural-network verification: the 2018 symbolic-interval preprint, DeepPoly (POPL 2019),
	212	+Marabou, with Reluplex (CAV 2017) and VNN-COMP (2023 report) as the lineage; the doubling of formal verification's
	213	+inbound citations from computer vision is consistent with this literature (claim G-13). Smart contracts: VerX (S&P
	214	+2020), "the first automated verifier able to prove functional properties of Ethereum smart contracts" per its abstract.
	215	+Learned theorem proving: MizAR 60 for Mizar 50 (ITP 2023), which reports automatic proofs of about 60% of Mizar theorems
	216	+in the hammer setting and 75% with human-selected premises, through the E and Vampire provers with ENIGMA and Deepire
	217	+learning and an incremental loop over millions of ATP proofs (claim G-17). The LLM-based line (Baldur, 2023; LeanDojo and
	218	+DeepSeek-Prover, preprints; the 2024 position paper on formal mathematical reasoning) reports whole-proof generation on
	219	+Lean and Isabelle benchmarks whose figures are not comparable across systems and are exposed to training-set overlap
	220	+(open problem G-P5). Equality saturation (egg, POPL 2021; CACM research highlight 2026) and quantitative interface
	221	+automata (CSL 2026) are the theory-of-verification entries. The category's primary share halved (1.06% → 0.52%) while
	222	+its multi-label count grew, which suggests that works carrying both a verification and an AI or security label lose the
	223	+primary label to the larger category (open problem G-P10).
	224	+
	225	+## 8. Software engineering
	226	+
	227	+Software engineering is where the window's change is largest and best documented. Surveys: machine-learning testing
	228	+(TSE 2020, 144 papers) at the baseline, software testing with LLMs (TSE 2024, 102 studies) in the window; the latter's
	229	+abstract names test-case preparation and program repair as the most common uses. Program repair and testing with
	230	+pre-trained models (ICSE 2023) and universal fuzzing with LLMs (ICSE 2024) are the method entries from `references.bib`
	231	+and the Part VII-a bibliography; SWE-bench and SWE-agent define the agentic evaluation setting. Empirical: the
	232	+Management Science field experiments (Microsoft, Accenture, a Fortune 100 firm; 4,867 developers) report a 26.08%
	233	+increase in completed tasks (s.e. 10.3%) with larger gains for less experienced developers, while the 2025 METR
	234	+randomised trial preprint with experienced open-source developers reports slower completion. The two are not
	235	+contradictory once population and outcome are distinguished (enterprise task completion versus expert time per issue in
	236	+mature repositories), but the reconciliation is the specialist's reading of two abstracts (claim G-18, contested).
	237	+Security: Copilot's vulnerable suggestions (2022), the user study on insecure code with assistants (CCS 2023) and the
	238	+ICPEC 2025 CWE-classification study agree that high detection or completion rates coexist with poor fine-grained security
	239	+reasoning (claim G-19). Co-classification with LLM reached 21.2% of the category in 2025, the highest of the six, while
	240	+references to LLM works are 2.5%; whether the vocabulary shift is a method shift is open problem G-P2. The category
	241	+also contains the only positive 2022-12 level change and the only significant GPT-4-date level change (claim G-09).
	242	+
	243	+## 9. Synthesis
	244	+
	245	+Three conclusions follow, in decreasing order of evidential strength. First, the landmark instrument fails in these
	246	+areas, for identifiable and partly correctable reasons (identifier collisions, generic-title chapters, re-deposited
	247	+classics, lexical misclassification, thesis cells); the correction is a selection-stage problem (G-P1), not a synthesis
	248	+problem, and the specialist's audit table should accompany any use of the list. Second, the six categories grew in
	249	+absolute terms and declined in relative terms after 2022; their ITS slope changes are positive but not distinguishable
	250	+from baseline pseudo-events; their vocabulary moved toward language models in proportion to their distance from
	251	+mathematics (software engineering and programming languages most, algorithms and complexity least); and their citation
	252	+behaviour moved little. Third, the substantive developments of the window are of two kinds that the corpus treats very
	253	+differently: theoretical results (flow, shortest paths, hashing, time–space simulation, 3-progressions) that the
	254	+citation-normalised ranking does not see, and the entry of language models into verification, program analysis,
	255	+compiler testing and software engineering, which the co-classification statistics see clearly but whose guarantees,
	256	+comparability and productivity effects remain unresolved (G-P3, G-P4, G-P5).

modified manuscript/appendices/appF_supplementary_tables.tex +0 −16

		`@@ -25,22 +25,6 @@ This appendix collects the generated tables not shown in the main text. Every ta`
25	25	\clearpage
26	26	\input{tables/tab_known_set}
27	27	\clearpage
28		−\input{tables/tab_net_area_flows}
29		−\clearpage
30		−\input{tables/tab_net_modularity}
31		−\clearpage
32		−\input{tables/tab_net_category_to_ai}
33		−\clearpage
34		−\input{tables/tab_net_category_actors}
35		−\clearpage
36		−\input{tables/tab_net_ai_vs_rest}
37		−\clearpage
38		−\input{tables/tab_net_ai_vs_rest_yearly}
39		−\clearpage
40		−\input{tables/tab_net_category_gini}
41		−\clearpage
42		−\input{tables/tab_net_gini_ai_vs_rest}
43		−\clearpage
44	28	\input{tables/tab_ai_growth}
45	29	\clearpage
46	30	\input{tables/tab_ai_counts}

modified manuscript/appendices/appG_figure_documentation.tex +2 −0

		`@@ -24,5 +24,7 @@ All figures are generated by \texttt{figures/make\_figures.py} (matplotlib, vect`
24	24	\texttt{fig\_cross\_area.pdf} & Cross-area reference shares by citing area and modularity of the category citation network by year. & cross\_area\_share\_yearly.csv; modularity\_yearly.csv \\
25	25	\texttt{fig\_emerging\_metrics.pdf} & Out-of-time detection performance of the emerging-topic models by horizon. & metrics\_h12.json; metrics\_h24.json \\
26	26	\texttt{fig\_known\_set.pdf} & Monthly keyword series for the twelve preregistered concepts with the detector's first alarm. & known\_set\_monthly.csv \\
	27	+\texttt{fig\_category\_its\_panels.pdf} & Monthly series and interrupted-time-series fits for every category. & fit\_cat\_*.csv \\
	28	+\texttt{fig\_llm\_cooccurrence.pdf} & Share of each category's multi-label works that also carry the LLM category, by year. & category\_cooccurrence\_yearly.csv; category\_counts\_yearly.csv \\
27	29	\bottomrule
28	30	\end{longtable}

modified manuscript/figures/fig_area_flow_matrix.pdf +0 −0

Binary file not shown.

modified manuscript/figures/fig_area_shares.pdf +0 −0

Binary file not shown.

modified manuscript/figures/fig_breaks_by_category.pdf +0 −0

Binary file not shown.

modified manuscript/figures/fig_category_growth.pdf +0 −0

Binary file not shown.

added manuscript/figures/fig_category_its_panels.pdf +0 −0

Binary file not shown.

modified manuscript/figures/fig_citation_inequality.pdf +0 −0

Binary file not shown.

modified manuscript/figures/fig_cited_age.pdf +0 −0

Binary file not shown.

modified manuscript/figures/fig_collaboration.pdf +0 −0

Binary file not shown.

modified manuscript/figures/fig_country_shares.pdf +0 −0

Binary file not shown.

modified manuscript/figures/fig_cross_area.pdf +0 −0

Binary file not shown.

modified manuscript/figures/fig_emerging_metrics.pdf +0 −0

Binary file not shown.

modified manuscript/figures/fig_field_inflows.pdf +0 −0

Binary file not shown.

modified manuscript/figures/fig_its_total.pdf +0 −0

Binary file not shown.

modified manuscript/figures/fig_known_set.pdf +0 −0

Binary file not shown.

added manuscript/figures/fig_llm_cooccurrence.pdf +0 −0

Binary file not shown.

modified manuscript/figures/fig_placebo_total.pdf +0 −0

Binary file not shown.

modified manuscript/figures/fig_prisma_flow.pdf +0 −0

Binary file not shown.

modified manuscript/figures/fig_types_openness.pdf +0 −0

Binary file not shown.

modified manuscript/figures/fig_venue_concentration.pdf +0 −0

Binary file not shown.

modified manuscript/figures/fig_volume_monthly_raw.pdf +0 −0

Binary file not shown.

modified manuscript/figures/fig_volume_yearly.pdf +0 −0

Binary file not shown.

deleted manuscript/scratch_p6.pdf +0 −0

Binary file not shown.

deleted manuscript/scratch_p6.tex +0 −52

		`@@ -1,52 +0,0 @@`
1		−% ============================================================================
2		−% UQO Working Paper 44 — The Post-ChatGPT Computer Science Revolution
3		−% Preamble follows the UQO working-paper template (_modele-working-paper/main.tex)
4		−% ============================================================================
5		−\documentclass[12pt,letterpaper]{article}
6		−\usepackage[utf8]{inputenc}
7		−\usepackage[T1]{fontenc}
8		−\usepackage[english]{babel}
9		−\usepackage[letterpaper,top=1in,bottom=1in,left=1in,right=1in]{geometry}
10		−\usepackage{mathptmx}
11		−\usepackage{microtype}
12		−\usepackage{setspace}
13		−\onehalfspacing
14		−\usepackage{amsmath,amssymb,amsthm}
15		−\usepackage{booktabs,tabularx,multirow,threeparttable,longtable,array}
16		−\usepackage{siunitx}
17		−\sisetup{input-symbols = {()}, table-format = -1.3, parse-numbers = false}
18		−\usepackage{graphicx,subcaption,float}
19		−\usepackage[dvipsnames]{xcolor}
20		−\usepackage[colorlinks=true,linkcolor=NavyBlue,citecolor=NavyBlue,urlcolor=NavyBlue,breaklinks=true]{hyperref}
21		−\usepackage[authoryear,round,semicolon]{natbib}
22		−\bibliographystyle{aer}
23		−\usepackage[font=small,labelfont=bf,labelsep=period,justification=justified,singlelinecheck=false]{caption}
24		−\usepackage{fancyhdr}
25		−\pagestyle{fancy}\fancyhf{}\renewcommand{\headrulewidth}{0pt}\fancyfoot[C]{\thepage}
26		−\usepackage{titlesec}
27		−\titleformat{\section}{\large\bfseries}{\thesection.}{0.5em}{}
28		−\titleformat{\subsection}{\normalsize\bfseries}{\thesubsection.}{0.5em}{}
29		−\titleformat{\subsubsection}{\normalsize\itshape}{\thesubsubsection.}{0.5em}{}
30		−\usepackage[toc,page]{appendix}
31		−\usepackage{enumitem}
32		−\setlist{nosep,leftmargin=*}
33		−\newtheorem{definition}{Definition}
34		−\newtheorem{proposition}{Proposition}
35		−\newtheorem{hypothesis}{Hypothesis}
36		−\newcommand{\E}{\mathbb{E}}
37		−\newcommand{\R}{\mathbb{R}}
38		−\newcommand{\pending}[1]{\textcolor{BrickRed}{[pending human review: #1]}}
39		−\graphicspath{{figures/}}
40		−
41		−% ---- metadata ----
42		−\newcommand{\WPnumber}{44}
43		−\newcommand{\WPtitle}{The Post-ChatGPT Computer Science Revolution}
44		−\newcommand{\WPsubtitle}{A Comprehensive Systematic Mapping, Bibliometric Analysis, and Knowledge Graph of Computer Science Research, 2022--2026}
45		−\newcommand{\WPdate}{November 2026}
46		−\newcommand{\WPversion}{0.1}
47		−\input{tables/numbers} % machine-generated macros (counts, estimates) — never edited by hand
48		−\begin{document}
49		−\input{tables/tab_systems_categories}
50		−\input{sections/part06_systems}
51		−\bibliography{references_all}
52		−\end{document}

modified manuscript/sections/part07_theory_security.tex +3 −1

		`@@ -1 +1,3 @@`
1		−% part07_theory_security — to be written (Phase 7)
	1	+% Part VII — Theory, Algorithms, and Security: theory half (Agent G) then security half (Agent H)
	2	+\input{sections/part07a_theory}
	3	+\input{sections/part07b_security}

added manuscript/sections/part07a_theory.tex +1 −0

		`@@ -0,0 +1 @@`
	1	+% part07a_theory — being written by Agent G

modified manuscript/sections/part08_networks.tex +246 −44

		`@@ -203,13 +203,36 @@ if anything slower in 2024 and 2025 than in 2021 and 2022.`
203	203
204	204	The China part of H6 asks whether the share of works with China-affiliated authors rises in the post period
205	205	\emph{in the AI categories}. The aggregate rise is established by Table~\ref{tab:country_concentration},
206		−whose column of works with a China-affiliated author is the quantity the hypothesis names; the category-specific version requires the
207		−classification and is \pending{China share by category after classification}. The aggregate result is,
208		−however, weaker evidence than it looks, because the rise is a continuation of a trend that predates the
209		−window and is documented in every recent comparison of national output
210		−\citep{bornmann2021growth}; a test of H6 that would be informative must compare the post-period
211		−growth of China's share in AI categories with its growth in the rest of computer science, which is what the
212		−pending analysis will do.
	206	+whose column of works with a China-affiliated author is the quantity the hypothesis names. The
	207	+category-specific version uses the primary category of Part~IX: Table~\ref{tab:net_category_actors} gives,
	208	+for each of the forty categories, the share of works with a China-affiliated author in the baseline and in
	209	+the post period, Table~\ref{tab:net_ai_vs_rest} pools the nine categories of the AI/ML area against the
	210	+thirty-one others, and Table~\ref{tab:net_ai_vs_rest_yearly} gives the two pooled series by year. Three
	211	+things must be said before reading them. The denominator of these shares is every work of the category,
	212	+including works without affiliation data, so the level is below that of Table~\ref{tab:country_concentration}
	213	+and is depressed wherever affiliation coverage is low, which is the case for the most recent cohorts and for
	214	+preprint-heavy categories; the within-category comparison across periods is affected to the extent that
	215	+coverage changed. The grouping by primary category assigns a work to one category only, so a vision paper
	216	+that uses a language model counts under computer vision, not under large language models. And the AI/ML
	217	+area of the ontology is wider than the four categories that define the post period, since it also contains
	218	+reinforcement learning, machine-learning theory, deep-learning architectures, agents and AI safety.
	219	+
	220	+With these qualifications, the China share rises from the baseline to the post period in the AI/ML area, as
	221	+H6 predicts, and it rises in the pooled other areas as well, by a larger margin; the note of
	222	+Table~\ref{tab:net_category_actors} counts the categories in which the share rises, and the rise is close to
	223	+universal outside the AI/ML area. Within the AI/ML area the rise is carried by the older categories,
	224	+reinforcement learning, deep-learning architectures and agents, while the categories that define the post
	225	+period behave differently: the share is nearly flat for large language models, falls for foundation models and
	226	+generative AI, and is highest of all forty categories, in both periods, for multimodal AI. The United States
	227	+share falls in both groups and in most categories, with large language models the one AI/ML category in
	228	+which it rises. The yearly series of Table~\ref{tab:net_ai_vs_rest_yearly} show the pooled shares of the two
	229	+groups moving together: both rise through 2023 and fall in 2024 and 2025, and the two series are closer to
	230	+each other in every post-period year than in any baseline year. The fall after 2023 coincides with the lower
	231	+affiliation coverage of recent cohorts (Section~\ref{sec:net-teams}) and cannot be read as a reversal; what
	232	+it does establish is that the AI/ML area is not where the China share grows fastest. The verdict on this
	233	+component of H6 is therefore a supported direction with an unsupported specificity: the share rises in AI
	234	+categories, but less than elsewhere, and the rise is the continuation of a trend that predates the window and
	235	+is documented in every recent comparison of national output \citep{bornmann2021growth}.
213	236
214	237	Concentration across countries, measured by the Herfindahl--Hirschman index of fractional shares and by the
215	238	shares of the five and ten largest countries, rises through 2025 (Table~\ref{tab:country_concentration}).
		`@@ -234,6 +257,9 @@ reported with the composition caveat rather than as a change in behaviour.`
234	257	\begingroup\renewcommand{\small}{\footnotesize}\setlength{\tabcolsep}{3.5pt}\input{tables/tab_countries}\endgroup
235	258	\begingroup\renewcommand{\small}{\footnotesize}\setlength{\tabcolsep}{3.5pt}\input{tables/tab_country_shares}\endgroup
236	259	\begingroup\renewcommand{\small}{\scriptsize}\setlength{\tabcolsep}{3pt}\input{tables/tab_country_concentration}\endgroup
	260	+\begingroup\scriptsize\setlength{\tabcolsep}{3pt}\input{tables/tab_net_category_actors}\endgroup
	261	+\begingroup\renewcommand{\small}{\scriptsize}\setlength{\tabcolsep}{3pt}\input{tables/tab_net_ai_vs_rest}\endgroup
	262	+\begingroup\renewcommand{\small}{\scriptsize}\setlength{\tabcolsep}{3pt}\input{tables/tab_net_ai_vs_rest_yearly}\endgroup
237	263
238	264	\section{Institutions and sectors}
239	265	\label{sec:net-institutions}
		`@@ -264,8 +290,23 @@ academic. Table~\ref{tab:institution_types} shows the same thing from the other`
264	290	share of affiliations rising through the window and the facility, government and company shares all giving
265	291	way. The aggregate part of H6 is therefore not supported; the sign is opposite to the one predicted. Whether
266	292	the category-specific version holds, that is, whether the company share rises within the AI categories while
267		−falling in the aggregate, is exactly the kind of composition question that the aggregate cannot settle, and it
268		−is \pending{industry share by category after classification}.
	293	+falling in the aggregate, is exactly the kind of composition question that the aggregate cannot settle, and
	294	+Tables~\ref{tab:net_category_actors}, \ref{tab:net_ai_vs_rest} and~\ref{tab:net_ai_vs_rest_yearly} settle it
	295	+in the negative. The company share of the AI/ML area is higher than that of the other areas in every year,
	296	+which is the familiar result that industry is more present in machine learning than in the rest of computer
	297	+science; but it peaks in 2021 and falls in every subsequent year, so that the baseline-to-post decline of the
	298	+AI/ML area is larger than that of the other areas, not smaller. The note of
	299	+Table~\ref{tab:net_category_actors} counts the categories whose company share rises: one of the nine AI/ML
	300	+categories (agents), against a sizeable minority of the others, among them cloud computing, information
	301	+retrieval, distributed computing, database systems, operating systems and quantum computing. The four
	302	+categories with the highest baseline company shares, multimodal AI, foundation models, reinforcement
	303	+learning and large language models, all lose share in the post period, and generative AI loses more than
	304	+half of its share. The denominator caveat of the previous paragraph applies with more force here than for the
	305	+China share, because the AI/ML area has the highest and fastest-growing share of arXiv-linked works
	306	+(Table~\ref{tab:net_ai_vs_rest}) and preprints are the records least often enriched with affiliations; part of
	307	+the measured decline is therefore coverage. Nothing in the category tables, however, suggests the opposite
	308	+sign for the AI categories, and the frontier-share results of \citet{ahmed2023industry} remain the only
	309	+sense in which an industry share has risen.
269	310
270	311	The result is less at odds with the cited literature than it seems, for three reasons that matter for its
271	312	interpretation. \citet{ahmed2023industry} measure industry's presence at the frontier---the largest models,
		`@@ -417,8 +458,24 @@ rather than a change in the shape of the tail, and the 2023 uptick is the trunca`
417	458	H5 predicts that citation inequality is higher in the post period than in the baseline \emph{for AI
418	459	categories}. The aggregate series can test only the aggregate statement, and for the one post-period
419	460	publication year with a complete window (2023) the aggregate Gini is above the 2022 value and below every
420		−baseline value before 2021, so the aggregate evidence is mixed and dominated by truncation; the category
421		−version of the test, which is the one the hypothesis states, is \pending{Gini by category after classification}.
	461	+baseline value before 2021, so the aggregate evidence is mixed and dominated by truncation. The category
	462	+version of the test, which is the one the hypothesis states, is in Table~\ref{tab:net_category_gini}, which
	463	+gives the Gini of $c^{(36)}$ by primary category for every publication year with a complete window, and in
	464	+Table~\ref{tab:net_gini_ai_vs_rest}, which pools the AI/ML area against the other areas as work-weighted means
	465	+of the category statistics. Two patterns are visible. Across the baseline the within-category Gini falls in
	466	+almost every category, as the aggregate does, and for the same reason, the falling share of uncited works.
	467	+For the 2023 cohort the pooled AI/ML Gini is above its baseline mean while the pooled Gini of the other
	468	+areas is below its own; the note of Table~\ref{tab:net_category_gini} counts the categories whose 2023 value
	469	+exceeds their baseline mean, and the proportion is far higher among the AI/ML categories than among the
	470	+others. Large language models have the highest Gini of any category in every year, and theirs rises in every
	471	+year of the window, baseline included, which is the one category-level series that moves against the
	472	+baseline decline. This is evidence in the direction of H5, and it is weak evidence: it rests on a single
	473	+post-period cohort whose window is affected by truncation, the uncited share of that cohort rises by about
	474	+the same amount in both groups, so the difference between them is a tail effect rather than a zeros effect,
	475	+and the small categories, together with the implausible 2023 value for formal verification, show the
	476	+sensitivity of a within-category Gini to a single misattributed record
	477	+(Section~\ref{sec:net-topcited}). The 2024 cohort, whose window closes at the end of 2026, will be the first
	478	+test with two post-period cohorts.
422	479	The reader should also weigh the finding of \citet{nielsen2021inequality} that global citation inequality has
423	480	risen over several decades when measured on long windows across all fields; the three-year windows used here
424	481	measure a different quantity, early attention, whose inequality is driven by zeros, and the two results are
		`@@ -444,8 +501,11 @@ the methodology's reservations about FWCI (Section~\ref{sec:m-bibliometrics}) ar`
444	501	indicator is not used further.
445	502
446	503	\begingroup\renewcommand{\small}{\scriptsize}\setlength{\tabcolsep}{3pt}\input{tables/tab_citation_windows}\endgroup
	504	+\begingroup\scriptsize\setlength{\tabcolsep}{3pt}\input{tables/tab_net_category_gini}\endgroup
	505	+\begingroup\renewcommand{\small}{\scriptsize}\setlength{\tabcolsep}{2.5pt}\input{tables/tab_net_gini_ai_vs_rest}\endgroup
447	506
448	507	\subsection{The most cited works and what they reveal about the database}
	508	+\label{sec:net-topcited}
449	509
450	510	Table~\ref{tab:top_cited} lists the ten works with the largest $c^{(36)}$ in each publication year from 2018
451	511	to 2025, with their total citations to date. The list is reported because research question RQ3 asks for the
		`@@ -487,12 +547,15 @@ cannot rely on a ranked list from the database without a record-level audit, and`
487	547	consistency of window and total counts, the uniqueness of the work across records, and the agreement of the
488	548	record's date with the work's first appearance. That audit is part of the human-review checkpoint on RQ3 and
489	549	is \pending{record-level audit of the top-cited lists}; until it is done, the titles above are quoted as what
490		−the database returns, not as a ranking endorsed by this study. The lesson for the field-normalised scores
	550	+the database returns, not as a ranking endorsed by this study. (This item does not depend on the
	551	+classification and remains open; it belongs to the RQ3 human-review checkpoint, not to the corpus pipeline.) The lesson for the field-normalised scores
491	552	of Section~\ref{sec:m-citations} is the same as the one \citet{waltman2016review} draw for all mean-based
492	553	indicators: a reference class containing one record with tens of thousands of misattributed citations has a
493	554	mean that is wrong for every work in the class, and the normalised scores of Part~IX should be computed
494	555	after the exclusion of records failing the consistency check; whether the current pipeline applies that
495		−exclusion is \pending{NCS computed after exclusion of inconsistent records}.
	556	+exclusion is \pending{NCS computed after exclusion of inconsistent records}. (Also left open: the
	557	+category-level mean normalised scores of \texttt{category\_impact\_yearly.csv} are computed on the unaudited
	558	+reference classes and are for that reason not reported in this part.)
496	559
497	560	\begingroup\footnotesize\setlength{\tabcolsep}{4pt}\input{tables/tab_top_cited}\endgroup
498	561
		`@@ -636,16 +699,133 @@ references, not as corrected for the composition of the unresolved ones, which c`
636	699
637	700	The flow matrices $F_t$ between the categories of the ontology, the cross-area shares $X_{\alpha,t}$ of
638	701	equation~\eqref{eq:cross-area} and the yearly modularities $Q_{\mathrm{cat}}$ and $Q_{\mathrm{area}}$ of
639		−equation~\eqref{eq:modularity} are the statistics on which H4 turns and the ones that would show whether
640		−systems, security and human--computer interaction cite the AI categories more after 2022 than before. They
641		−require the primary category of every work and are produced by
642		−\texttt{knowledge\_graph/02\_category\_flows.py} once the classification of Part~IX is complete. The result
643		−files \texttt{category\_flows\_yearly.csv}, \texttt{cross\_area\_share\_yearly.csv} and
644		−\texttt{modularity\_yearly.csv} do not exist at the time of writing, and this subsection is
645		−\pending{category flows after classification}. When it is filled, it should report, for each citing area, the
646		−cross-area share by year with the composition-adjusted variant of Section~\ref{sec:m-networks}, the two
647		−modularities with the self-share, and a comparison of the baseline and post means with the caveat that the
648		−post-period citing years have shorter reference lists in the incomplete 2026 row.
	702	+equation~\eqref{eq:modularity} are the statistics on which H4 turns. They require the primary category of
	703	+every work, which the classification of Part~IX supplies, and they are computed by
	704	+\texttt{knowledge\_graph/02\_category\_flows.py} on the internal edges of Section~\ref{sec:m-citgraph},
	705	+that is, on references from eligible works to eligible works with a primary category at both ends.
	706	+Two properties of this edge set govern everything that follows. First, the internal network is small in
	707	+2018 and grows with every citing year, because the corpus begins in 2018 and a citing year can only cite
	708	+the corpus years that precede it; the 2018 row of every table below describes a network of references to
	709	+works of the same year and is not comparable with the later rows, so the series effectively begin in 2019.
	710	+Second, the restriction to internal references means that an area's flows are measured over its references
	711	+to \emph{recent} computer science only; references to older work and to other fields, which
	712	+Sections~\ref{sec:net-flows-fields} and~\ref{sec:net-age} showed to be a declining share, are outside the
	713	+matrix. Table~\ref{tab:net_area_flows} and Figure~\ref{fig:area_flow_matrix} give the area flow matrices
	714	+by period, Table~\ref{tab:net_category_to_ai} the share of each category's internal references that go to
	715	+the nine categories of the AI/ML area, and Table~\ref{tab:net_modularity} and
	716	+Figure~\ref{fig:cross_area} the cross-area shares and the modularities by year.
	717	+
	718	+\begin{figure}[htbp]
	719	+\centering
	720	+\includegraphics[width=0.95\textwidth]{fig_area_flow_matrix.pdf}
	721	+\caption{Citation-flow matrices between the eight areas of the ontology, baseline (citing years
	722	+2018--2022) and post (2023--2026) periods: row shares of internal references from the citing area (rows)
	723	+to the cited area (columns), per cent. Source: \texttt{results/graph/area\_flows\_yearly.csv}.}
	724	+\label{fig:area_flow_matrix}
	725	+\end{figure}
	726	+
	727	+The area matrices are diagonal-dominant in both periods: every area sends the majority of its internal
	728	+references to itself, from about two thirds for the security and software areas to a much higher share for
	729	+the perception area, which is the largest area and the one whose categories (vision, language, speech,
	730	+robotics, interaction, graphics) cite each other most. The largest off-diagonal flows are structural and do
	731	+not change between periods: the AI/ML area sends about a third of its references to the perception area,
	732	+which is where the application work on vision and language that the machine-learning categories build on
	733	+is classified; the security area sends about a sixth of its references to the data and distributed area, the
	734	+networks and cloud work that security builds on; and computational science splits its outside references
	735	+between AI/ML and perception. What changes between periods is the AI/ML column. The self-share of every
	736	+area except perception falls, and the share of references each of these areas sends to the AI/ML area rises,
	737	+the increments being largest for the software area, computational science, security and the data and
	738	+distributed area (Table~\ref{tab:net_area_flows}; the post-minus-baseline differences are the subject of
	739	+Part~IX). The AI/ML area's own self-share falls as well, in favour of the perception area: the
	740	+machine-learning categories cite vision and language work more in the post period than before, which is the
	741	+flow one expects when the models of the post period are evaluated on the tasks of those categories. The
	742	+perception area is the exception on both counts: its self-share is the same in the two periods and its
	743	+share of references to the AI/ML area changes little, so the convergence that the AI/ML column shows is a
	744	+convergence of the smaller areas toward AI, not of the largest.
	745	+
	746	+The category-level table (Table~\ref{tab:net_category_to_ai}) tests H4 at the resolution at which it is
	747	+stated, since the hypothesis names systems, security and human--computer interaction as the citing fields
	748	+that should cite the AI categories more. All three do. The share of references going to the AI/ML area
	749	+rises in human--computer interaction, software engineering, cybersecurity, database systems, distributed
	750	+computing, computer networks and computer architecture, in each case from the beginning of the window,
	751	+and in each case the post-period values are above every baseline value. The timing differs across these
	752	+series in a way the hypothesis does not predict. For human--computer interaction and software engineering
	753	+the series is nearly flat through 2023 and rises steeply in 2024, 2025 and the incomplete 2026, so the
	754	+change is concentrated in the second and third post-period years rather than at the boundary, consistent
	755	+with the lag between the appearance of a model and the appearance of empirical work that cites it; for
	756	+cybersecurity, database systems and distributed computing the post-period increments continue the baseline
	757	+pace; for computer networks and computer architecture the share is the lowest among the systems categories
	758	+and its rise is slight. Within the perception area, natural language processing is the category with the
	759	+largest post-period change of any in the table: its share of references to the AI/ML area falls from 2018
	760	+to 2020, which is the period in which the pre-trained language model became the standard tool of the
	761	+category and was cited as NLP work rather than as machine learning, and rises steeply from 2023, as the
	762	+large-language-model category (which the ontology places in the AI/ML area) absorbs the work the category
	763	+now builds on. Computer vision shows no such rise, which again reflects the classifier's boundary: vision
	764	+work citing vision models is a within-category flow. The theory categories rise modestly, and
	765	+computational biology and robotics return in the post period to the shares they had in 2018. These are
	766	+flows of references between classified works; they say that the systems, security and interaction
	767	+categories build on AI/ML-classified work more than they did, which is what H4 asserts, and not that the
	768	+work itself has become AI research, which is a question about classification rather than citation.
	769	+
	770	+\begin{figure}[htbp]
	771	+\centering
	772	+\includegraphics[width=0.95\textwidth]{fig_cross_area.pdf}
	773	+\caption{Left: share of internal references leaving the citing area, $X_{\alpha,t}$ of
	774	+equation~\eqref{eq:cross-area}, by citing area and year (per cent). Right: modularity of the yearly
	775	+category citation-flow network at the area and category partitions (equation~\eqref{eq:modularity}) and the
	776	+within-category self-share. Sources: \texttt{results/graph/cross\_area\_share\_yearly.csv},
	777	+\texttt{results/graph/modularity\_yearly.csv}. 2026 is incomplete.}
	778	+\label{fig:cross_area}
	779	+\end{figure}
	780	+
	781	+The cross-area shares (Figure~\ref{fig:cross_area}, left; the yearly values by area are in Part~IX) order
	782	+the areas by how much of their citation leaves them. The AI/ML area has the highest boundary-crossing rate
	783	+in every year and the rate rises through 2022 and is flat thereafter; its position at the top is a
	784	+consequence of where the ontology draws its boundaries, since the area's two largest partners, vision and
	785	+language, are placed in a different area, and the statistic would be lower under any ontology that grouped
	786	+machine learning with its application fields. This dependence on the partition is the reason
	787	+Section~\ref{sec:m-networks} describes $X_{\alpha,t}$ as a boundary-crossing rate and not as a measure of
	788	+interdisciplinarity in the sense of \citet{stirling2007diversity} or \citet{rafols2010diversity}, whose
	789	+indicators weight pairs of categories by their distance. The security, software, theory and data areas
	790	+rise through the whole window at a steady pace, the hardware area is flat, computational science rises and
	791	+falls back in the incomplete year, and the perception area, after a fall in 2019 that belongs to the small
	792	+2018 network, is flat at the lowest level. Table~\ref{tab:net_modularity} aggregates these shares over areas
	793	+in two ways. The raw aggregate is flat through the baseline and rises in every post-period year; the
	794	+composition-adjusted aggregate, which holds the weights of the citing areas at their baseline values, rises
	795	+by less, so part of the post-period rise of the raw series is the growing weight among citing works of the
	796	+AI/ML area, which is the most boundary-crossing area, and the remainder, which is positive in every
	797	+post-period year, is a within-area rise of boundary crossing. H4's first component, that the share of
	798	+citations crossing area boundaries rises in the post period, is supported on both the raw and the adjusted
	799	+series, with the composition qualification stated.
	800	+
	801	+H4's second component is the modularity. $Q_{\mathrm{cat}}$ peaks in 2019 and falls in every subsequent
	802	+year; $Q_{\mathrm{area}}$ falls in every year from 2018 to 2025 and is flat, with a slight uptick in the
	803	+incomplete year, from 2023 onward; the self-share peaks in 2019 and falls thereafter, fastest after 2023
	804	+(Table~\ref{tab:net_modularity}, Figure~\ref{fig:cross_area}, right). The post-period means are below the
	805	+baseline means at both partitions, as the note of the table records, so the verdict on this component is
	806	+that the hypothesis is supported in direction at both levels. The caveats are substantive. The decline is
	807	+present throughout the baseline; the 2022 boundary is not a break in either series, and the yearly
	808	+decrements of $Q_{\mathrm{cat}}$ in the post period are of the same order as the larger baseline
	809	+decrements, while $Q_{\mathrm{area}}$ stops falling in the post period, so that at the area partition the
	810	+post period is where the decline ends rather than where it begins. Because the partition is fixed, a fall in
	811	+$Q$ can arise from a change in the mix of citing works as well as from a change in citing behaviour; the
	812	+adjusted cross-area series shows that both are present. The modularity of equation~\eqref{eq:modularity}
	813	+inherits every property of the classifier, and in particular the placement of large language models in the
	814	+AI/ML area and of natural language processing outside it converts the post-period citation of LLM work by
	815	+NLP work into a cross-area flow by construction. The internal network grows by an order of magnitude over
	816	+the window, so the 2018 and 2019 values describe sparse networks; and the 2026 row is a partial year whose
	817	+works carry fewer parsed references. The co-classification matrix of
	818	+\texttt{category\_cooccurrence\_yearly.csv}, which counts works carrying two categories in their full label
	819	+set, is reported in Part~IX and is not used here, because, as Section~\ref{sec:m-classification} notes, it
	820	+responds to abstract length and vocabulary in a way the primary-label flows do not. What the flows and the
	821	+modularity together establish is modest and definite: the citation network of recent computer science is
	822	+becoming less assortative by category, the change is gradual and began before 2022, and the areas that
	823	+have moved are the smaller ones, which cite the machine-learning categories more, while the largest area
	824	+cites itself as much as it ever did.
	825	+
	826	+\begingroup\renewcommand{\small}{\scriptsize}\setlength{\tabcolsep}{3pt}\input{tables/tab_net_area_flows}\endgroup
	827	+\begingroup\renewcommand{\small}{\scriptsize}\setlength{\tabcolsep}{3pt}\input{tables/tab_net_modularity}\endgroup
	828	+\begingroup\scriptsize\setlength{\tabcolsep}{3pt}\input{tables/tab_net_category_to_ai}\endgroup
649	829
650	830	\section{The boundary of the database}
651	831	\label{sec:net-boundaries}
		`@@ -702,17 +882,31 @@ The protocol states its hypotheses with a direction so that null and opposite re`
702	882	such. On the evidence of this part, the verdicts are the following, each limited to the component of the
703	883	hypothesis that the available tables test.
704	884
705		−H4 (convergence) is not testable in its stated form, which concerns citation flows between areas of the
706		−ontology and the modularity of the area network; those statistics await the classification. Its field-level
707		−counterpart, the share of computer-science references that cross into other OpenAlex fields, falls through
708		−the window with no change of pace at the boundary (Table~\ref{tab:field_flows},
709		−Section~\ref{sec:net-flows-fields}), part of the fall being a mechanical consequence of the corpus
710		−boundary; it neither supports nor contradicts the within-field convergence the hypothesis describes.
711		−
712		−H5 (concentration) has two components. The citation-inequality component is stated for AI categories and is
713		−pending; its aggregate counterpart gives a Gini of three-year citations that falls through the baseline and
714		−rises for the one complete post-period cohort, a movement dominated by the share of uncited works and by
715		−right truncation, and therefore not interpretable as a change in concentration. The venue-concentration
	885	+H4 (convergence) is supported in direction on both of its components and qualified on both. The share of
	886	+internal references crossing area boundaries rises in the post period, on the raw aggregate and on the
	887	+composition-adjusted aggregate alike, and the systems, security and interaction categories the hypothesis
	888	+names all send a growing share of their references to the AI/ML categories, with the steepest rises for
	889	+human--computer interaction and software engineering in 2024--2026 rather than at the boundary
	890	+(Tables~\ref{tab:net_category_to_ai} and~\ref{tab:net_modularity}). The modularity falls at both the
	891	+category and the area partition, with post-period means below baseline means, but the fall is a
	892	+baseline trend that continues at the category level and stops at the area level after 2023; part of it is
	893	+the growing weight of the most boundary-crossing area among citing works, and all of it depends on where the
	894	+ontology draws the boundary between machine learning and its application fields
	895	+(Section~\ref{sec:net-flows-categories}). The field-level counterpart, the share of computer-science
	896	+references that cross into other OpenAlex fields, falls through the window with no change of pace at the
	897	+boundary (Table~\ref{tab:field_flows}, Section~\ref{sec:net-flows-fields}), part of the fall being a
	898	+mechanical consequence of the corpus boundary; convergence within computer science and a narrowing of its
	899	+references outside computer science are therefore observed together.
	900	+
	901	+H5 (concentration) has two components. The citation-inequality component is stated for AI categories, and
	902	+for the one post-period cohort with a complete window the pooled Gini of the AI/ML categories is above its
	903	+baseline mean while that of the other categories is below, with most AI/ML categories and a minority of the
	904	+others above their baseline means and large language models the most unequal category in every year
	905	+(Tables~\ref{tab:net_category_gini} and~\ref{tab:net_gini_ai_vs_rest}); this is the predicted direction, on
	906	+one cohort affected by right truncation, with the uncited share rising equally in both groups, so the
	907	+evidence is weak and provisional until the 2024 window closes. The aggregate counterpart gives a Gini of
	908	+three-year citations that falls through the baseline and rises for that same cohort, a movement dominated by
	909	+the share of uncited works and by truncation, and not interpretable as a change in concentration. The venue-concentration
716	910	component predicted a stable or lower index once preprints are treated separately; the index with arXiv
717	911	included rises, for the mechanical reason the protocol anticipated, and the index without the preprint type
718	912	is pending. The non-arXiv venue landscape widens.
		`@@ -720,16 +914,24 @@ is pending. The non-arXiv venue landscape widens.`
720	914	H6 (geography and collaboration) has three components. The international co-authorship share does not
721	915	rise, as predicted, and in fact declines slightly, with a composition explanation that the country shares
722	916	support. The China share rises in the aggregate under both counting conventions, as predicted, but as the
723		−continuation of a baseline trend, and the category-specific version is pending. The industry share does
724		−\emph{not} rise in the aggregate: it peaks in 2021 and declines in every post-period year, which is the
725		−opposite of the predicted direction, with the category-specific version pending and with the frontier-level
726		−measures of the literature it was drawn from measuring a different object.
	917	+continuation of a baseline trend; it rises in the AI/ML categories too, by less than in the other categories,
	918	+so the specificity the hypothesis asserts is not observed, and the categories that define the post period
	919	+(large language models, foundation models, generative AI) are flat or falling
	920	+(Tables~\ref{tab:net_category_actors} and~\ref{tab:net_ai_vs_rest}). The industry share does \emph{not}
	921	+rise in the aggregate: it peaks in 2021 and declines in every post-period year, which is the opposite of the
	922	+predicted direction; the category-specific version is also opposite, with the AI/ML area losing more company
	923	+share than the other areas and all of its highest-share categories losing share, part of the decline being
	924	+the lower affiliation coverage of preprint-heavy categories, and with the frontier-level measures of the
	925	+literature it was drawn from measuring a different object.
727	926
728	927	What the part adds beyond the hypotheses is a set of structural observations that were not predicted and
729		−that the pending analyses will have to accommodate: the fall in absolute joint output between China and the
730		−United States during a period in which both grew; the stability of the institutional hierarchy; the rise of
731		−generalist repositories among the largest sources of computer-science records; the steady shift of
732		−references toward younger work without acceleration at the boundary; and the record-level artefacts in the
733		−most-cited lists, which make any ranking of influential works conditional on an audit that this part has
734		−specified but not performed.
	928	+that the interpretation of Part~IX will have to accommodate: the fall in absolute joint output between China
	929	+and the United States during a period in which both grew; the stability of the institutional hierarchy; the
	930	+rise of generalist repositories among the largest sources of computer-science records; the steady shift of
	931	+references toward younger work without acceleration at the boundary; the asymmetry of the category flows,
	932	+in which the smaller areas move toward the machine-learning categories while the largest area, perception
	933	+and interaction, cites itself as much as before; the late timing of the rise in citation of AI work by
	934	+human--computer interaction and software engineering, two to three years after the boundary; and the
	935	+record-level artefacts in the most-cited lists, which make any ranking of influential works conditional on
	936	+an audit that this part has specified but not performed.
735	937

modified manuscript/sections/part09_results.tex +84 −0

		`@@ -941,3 +941,87 @@ and three others (methodological adoption, benchmark proliferation, open-source`
941	941	feature set (Section~\ref{sec:m-emerging}); the result is therefore a lower bound on what the full framework
942	942	could achieve, and an upper bound on what growth and citation signals alone achieve. \pending{H-4: whether
943	943	to describe the detector as a negative result for RQ10 or as a baseline for the semantic-cluster extension}.
	944	+
	945	+\section{Validation of the classification}
	946	+\label{sec:r-validation}
	947	+
	948	+Every category-level result of this part is conditional on the rule-based classifier of Part~IV, and the
	949	+protocol requires its validation on a stratified sample of two thousand works rated independently by two
	950	+raters, with Cohen's $\kappa$ between the raters, and the precision, recall and F1 of the assignment against
	951	+each rater and against their consensus, by category (Section~\ref{sec:m-validation}; checkpoint H-2). The
	952	+ratings of the two raters exist in \texttt{taxonomy/validation/}, but the metrics file
	953	+\texttt{taxonomy/validation/metrics.json} and the disagreement list had not been produced when this part was
	954	+written, so the validation cannot be reported here. \pending{taxonomy validation: insert the inter-rater
	955	+$\kappa$, the overall precision, recall and F1 of the assignment against the consensus, the number of
	956	+disagreements, and the categories with the lowest F1, from taxonomy/validation/metrics.json and
	957	+disagreements.csv; the generated Table~\ref{tab:validation} will exist once the file does}. Until then, the
	958	+reader should weight the category results by the known failure modes of the classifier (Section~\ref{sec:m-classification}):
	959	+works without an abstract are classified from venue and arXiv category alone, which favours the venue-defined
	960	+categories; the keyword-defined categories (the AI area, Sustainable Computing, Autonomous Scientific
	961	+Discovery) are sensitive to vocabulary drift; and the primary category of a multi-label work is decided by a
	962	+score whose ties are not random. The record-linkage precision estimated on the manually reviewed sample
	963	+(\LinkagesampleDoiAutoPrecision{} for DOI links, \LinkagesampleTitleYearAutoPrecision{} for title-and-year
	964	+links; Section~\ref{sec:m-linkprecision}) is the one validation statistic of the corpus that is available,
	965	+and it bounds the error introduced by linkage rather than by classification.
	966	+
	967	+\section{Summary of the hypotheses and of what remains uncertain}
	968	+\label{sec:r-summary}
	969	+
	970	+Table~\ref{tab:p9-hypotheses} collects the verdicts. The table is a reading aid and not a result file; each
	971	+verdict is argued in the section cited and carries the qualifications stated there.
	972	+
	973	+\begin{table}[htbp]\centering
	974	+\caption{Verdicts on the pre-specified hypotheses on the evidence of this part.}
	975	+\label{tab:p9-hypotheses}
	976	+\small
	977	+\begin{tabular}{p{0.07\textwidth}p{0.36\textwidth}p{0.17\textwidth}p{0.3\textwidth}}
	978	+\toprule
	979	+\textbf{H} & \textbf{Clause} & \textbf{Verdict} & \textbf{Where and why} \\
	980	+\midrule
	981	+H1 & Post-period monthly growth rate higher than baseline & supported & Sec.~\ref{sec:r-h1}; slope change ranks first among all candidate dates \\
	982	+ & Level shift at the boundary not distinguishable from placebo breaks & supported & placebo rank of the level near the bottom; acceleration dated to spring 2022 by Bai--Perron \\
	983	+ & Growth differences concentrated in a few categories & not supported & Sec.~\ref{sec:r-its-categories}; positive slope change in 37 of 40 categories \\
	984	+H2 & AI-category share rises in the post period & supported & Sec.~\ref{sec:r-h2}; driven by the language and generative categories; Deep Learning Architectures falls \\
	985	+ & A non-AI area declines relatively while growing absolutely & supported & three areas (data and distributed, theory, software) \\
	986	+H3 & Bai--Perron break in Oct.\ 2022--June 2023 for LLM, generative AI, agents, AI safety & not supported & Sec.~\ref{sec:r-h3}; all four break before the window (2021--2022) \\
	987	+ & No break in the window for formal verification, complexity, OS, architecture & opposite (two of four) & complexity and architecture break inside the window; no $p$-value attached to the dates \\
	988	+H4 & Cross-area citation shares rise & supported, small & Sec.~\ref{sec:r-h4}; about half composition; Perception area does not rise \\
	989	+ & Area modularity decreases & supported & monotone through the window, not concentrated at the boundary \\
	990	+H5 & Citation Gini higher in post period for AI categories & inconclusive & Sec.~\ref{sec:r-h5}; per-category windows not produced; aggregate movement is a truncation effect \\
	991	+ & Venue HHI stable or lower & opposite as computed, inconclusive as intended & index without repositories not produced \\
	992	+H6 & Industry share rises (AI categories) & opposite (aggregate), inconclusive (AI) & Sec.~\ref{sec:r-h6}; company share peaks in 2021 \\
	993	+ & China share rises (AI categories) & supported (aggregate), inconclusive (AI) & continuation of a baseline trend \\
	994	+ & International co-authorship share does not rise & supported & declines from 2020 \\
	995	+H7 & arXiv-version share rises & supported & Sec.~\ref{sec:r-h7}; inflated by unmatched recent preprints \\
	996	+ & Linked-code share rises & inconclusive & not measured \\
	997	+H8 & Trained detector AUC above 0.70 and P@50 above 0.40 versus the baseline & not supported & Sec.~\ref{sec:r-h8}; AUC at the threshold, P@50 an order of magnitude below; baseline wins at 24 months \\
	998	+ & Known-emergence set detected with median lead at least 6 months & supported with reservations & substitute keyword test, baseline rule, ex-post peak \\
	999	+\bottomrule
	1000	+\end{tabular}
	1001	+\end{table}
	1002	+
	1003	+Four sources of uncertainty run through every verdict and are stated here once. The first is right
	1004	+truncation: the post period ends in an incomplete year, the reference-indexing lag removes a growing share
	1005	+of references from the most recent cohorts, and every citation-based statistic of 2024 and later is
	1006	+provisional; the structural-change analysis handles this by truncating at \ItsLastComplete{} and varying the
	1007	+truncation month, and the citation analysis by reporting only complete windows, but neither device recovers
	1008	+what the database has not yet indexed. The second is the boundary of the database: the growth of the corpus
	1009	+in the post period is concentrated in records outside DBLP's curation, and the study cannot separate growth
	1010	+in computer-science output from growth in OpenAlex's coverage of it \citep{culbert2025openalexcoverage,
	1011	+alperin2024openalex}; the DBLP-only records of Table~\ref{tab:dblp_only} and the strict core of
	1012	+\PrismaStrictCoreDblpAndCsPrimary{} works are the instruments for a sensitivity analysis that the current
	1013	+result files do not yet contain (\pending{macro: structural-change estimates on the strict core, as the
	1014	+protocol's sensitivity analysis}). The third is the classifier, whose validation is pending and whose
	1015	+boundaries are fixed in 2026 vocabulary and applied to 2018 works; every category-level and every flow
	1016	+statistic inherits it, and the semantic measurements that do not depend on it are the ones still to come.
	1017	+The fourth is the inferential status of the time-series results: the reference date is a marker, the
	1018	+estimates at it are descriptions of a kink in a series that was already bending, and the placebo and
	1019	+alternative-date analyses show that the kink is not tied to the month. What the evidence supports is a
	1020	+multi-year acceleration of the indexed output of computer science, concentrated in share terms in the
	1021	+language and generative-model categories but present in nearly all of them, accompanied by a modest increase
	1022	+in the citation traffic from the other areas toward AI and by a reorganisation of the co-classification
	1023	+structure around the foundation-model categories, and a continuation, without inflection, of the baseline
	1024	+trends in collaboration, geography and openness. What it does not support is a discrete change at the
	1025	+reference date in any series examined, a concentration of the acceleration in a few categories, or the
	1026	+ability of simple growth and citation signals to pick the emerging topics of 2023--2024 with useful
	1027	+precision. \pending{H-4: the summary reading of this paragraph}.

modified manuscript/tables/TABLES.json +56 −56

		`@@ -103,7 +103,7 @@`
103	103	"caption": "Fractional output of the fifteen largest countries by year.",
104	104	"label": "tab:countries",
105	105	"sources": [
106		− "./results/bibliometrics/countries_yearly.csv"
	106	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/countries_yearly.csv"
107	107	],
108	108	"rows": 15
109	109	},
		`@@ -112,7 +112,7 @@`
112	112	"caption": "Structural-change estimates by series: ITS level and slope changes at 2022-12 (HAC p-values), placebo p-values, Bai\u2013Perron (BIC) and PELT break dates.",
113	113	"label": "tab:breaks",
114	114	"sources": [
115		− "./results/econometrics/breaks.csv"
	115	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/econometrics/breaks.csv"
116	116	],
117	117	"rows": 41
118	118	},
		`@@ -121,7 +121,7 @@`
121	121	"caption": "Out-of-time performance of the emerging-topic detectors at horizon 12 months (test cutoffs 2023-01\u20132024-10).",
122	122	"label": "tab:emerging_h12",
123	123	"sources": [
124		− "./results/emerging/metrics_h12.json"
	124	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/emerging/metrics_h12.json"
125	125	],
126	126	"rows": 3
127	127	},
		`@@ -130,7 +130,7 @@`
130	130	"caption": "Out-of-time performance of the emerging-topic detectors at horizon 24 months (test cutoffs 2023-01\u20132024-04).",
131	131	"label": "tab:emerging_h24",
132	132	"sources": [
133		− "./results/emerging/metrics_h24.json"
	133	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/emerging/metrics_h24.json"
134	134	],
135	135	"rows": 3
136	136	},
		`@@ -139,7 +139,7 @@`
139	139	"caption": "Preregistered known-emergence set: detection month, peak-growth month and lead time.",
140	140	"label": "tab:known_set",
141	141	"sources": [
142		− "./results/emerging/known_set_lead_times.csv"
	142	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/emerging/known_set_lead_times.csv"
143	143	],
144	144	"rows": 12
145	145	},
		`@@ -148,8 +148,8 @@`
148	148	"caption": "Team size and collaboration indicators by publication year (all eligible works; periods of the source file summed).",
149	149	"label": "tab:collaboration",
150	150	"sources": [
151		− "./results/bibliometrics/team_size_yearly.csv",
152		− "./results/bibliometrics/international_collab_yearly.csv"
	151	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/team_size_yearly.csv",
	152	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/international_collab_yearly.csv"
153	153	],
154	154	"rows": 9
155	155	},
		`@@ -158,7 +158,7 @@`
158	158	"caption": "Share of typed affiliations by OpenAlex institution type and year (per cent of affiliation-work pairs).",
159	159	"label": "tab:institution_types",
160	160	"sources": [
161		− "./results/bibliometrics/institution_types_yearly.csv"
	161	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/institution_types_yearly.csv"
162	162	],
163	163	"rows": 9
164	164	},
		`@@ -167,7 +167,7 @@`
167	167	"caption": "Share of fractional output of the fifteen largest countries by year (per cent).",
168	168	"label": "tab:country_shares",
169	169	"sources": [
170		− "./results/bibliometrics/countries_yearly.csv"
	170	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/countries_yearly.csv"
171	171	],
172	172	"rows": 15
173	173	},
		`@@ -176,7 +176,7 @@`
176	176	"caption": "Concentration of output across countries by year, and the two largest producers under full counting.",
177	177	"label": "tab:country_concentration",
178	178	"sources": [
179		− "./results/bibliometrics/countries_yearly.csv"
	179	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/countries_yearly.csv"
180	180	],
181	181	"rows": 9
182	182	},
		`@@ -185,7 +185,7 @@`
185	185	"caption": "The country co-authorship network by period: edges with at least fifty joint works.",
186	186	"label": "tab:country_network",
187	187	"sources": [
188		− "./results/bibliometrics/country_pairs_by_period.csv"
	188	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/country_pairs_by_period.csv"
189	189	],
190	190	"rows": 2
191	191	},
		`@@ -194,7 +194,7 @@`
194	194	"caption": "The twenty strongest country pairs by period (number of joint works).",
195	195	"label": "tab:country_pairs",
196	196	"sources": [
197		− "./results/bibliometrics/country_pairs_by_period.csv"
	197	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/country_pairs_by_period.csv"
198	198	],
199	199	"rows": 20
200	200	},
		`@@ -203,7 +203,7 @@`
203	203	"caption": "Institutions with at least two hundred works per period: coverage and concentration (fractional counting).",
204	204	"label": "tab:institution_concentration",
205	205	"sources": [
206		− "./results/bibliometrics/top_institutions_by_period.csv"
	206	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/top_institutions_by_period.csv"
207	207	],
208	208	"rows": 3
209	209	},
		`@@ -212,7 +212,7 @@`
212	212	"caption": "The twenty-five institutions with the largest fractional output in the post period, with their baseline output.",
213	213	"label": "tab:top_institutions",
214	214	"sources": [
215		− "./results/bibliometrics/top_institutions_by_period.csv"
	215	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/top_institutions_by_period.csv"
216	216	],
217	217	"rows": 25
218	218	},
		`@@ -221,7 +221,7 @@`
221	221	"caption": "Distribution of works over source types among the listed sources, by period (per cent).",
222	222	"label": "tab:source_types",
223	223	"sources": [
224		− "./results/bibliometrics/top_sources_by_period.csv"
	224	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/top_sources_by_period.csv"
225	225	],
226	226	"rows": 2
227	227	},
		`@@ -230,7 +230,7 @@`
230	230	"caption": "The twenty largest sources (venues) by period, with OpenAlex source type.",
231	231	"label": "tab:top_sources",
232	232	"sources": [
233		− "./results/bibliometrics/top_sources_by_period.csv"
	233	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/top_sources_by_period.csv"
234	234	],
235	235	"rows": 20
236	236	},
		`@@ -239,7 +239,7 @@`
239	239	"caption": "Age of cited works by citing year (resolved references of eligible works; corpus version of the result file at generation time).",
240	240	"label": "tab:cited_age",
241	241	"sources": [
242		− "./results/graph/cited_age_yearly.csv"
	242	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/cited_age_yearly.csv"
243	243	],
244	244	"rows": 9
245	245	},
		`@@ -248,7 +248,7 @@`
248	248	"caption": "The ten most cited works of each publication year by 36-month window citations (c36), with total citations to date.",
249	249	"label": "tab:top_cited",
250	250	"sources": [
251		− "./results/bibliometrics/top_cited_by_year.csv"
	251	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/top_cited_by_year.csv"
252	252	],
253	253	"rows": 80
254	254	},
		`@@ -257,7 +257,7 @@`
257	257	"caption": "DBLP records of 2018--2026 without an OpenAlex counterpart, by DBLP record type and year.",
258	258	"label": "tab:dblp_only",
259	259	"sources": [
260		− "./results/bibliometrics/dblp_only_yearly.csv"
	260	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/dblp_only_yearly.csv"
261	261	],
262	262	"rows": 9
263	263	},
		`@@ -266,7 +266,7 @@`
266	266	"caption": "The OpenAlex \\texttt{created\\_date} relative to the publication date, by publication year.",
267	267	"label": "tab:indexing_lag",
268	268	"sources": [
269		− "./results/bibliometrics/indexing_lag.csv"
	269	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/indexing_lag.csv"
270	270	],
271	271	"rows": 9
272	272	},
		`@@ -275,7 +275,7 @@`
275	275	"caption": "Reference coverage by citing year: declared, resolved, internal and external references of eligible works.",
276	276	"label": "tab:reference_coverage",
277	277	"sources": [
278		− "./results/graph/reference_coverage_yearly.csv"
	278	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/reference_coverage_yearly.csv"
279	279	],
280	280	"rows": 9
281	281	},
		`@@ -284,7 +284,7 @@`
284	284	"caption": "Destination of the resolved references of eligible works by citing year: inside the corpus, computer science outside it, and the seven largest other OpenAlex fields (per cent of resolved references).",
285	285	"label": "tab:field_flows",
286	286	"sources": [
287		− "./results/graph/field_flows_yearly.csv"
	287	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/field_flows_yearly.csv"
288	288	],
289	289	"rows": 9
290	290	},
		`@@ -293,7 +293,7 @@`
293	293	"caption": "Citation-flow matrices between the eight areas of the ontology by period: share of each citing area's internal references received by each cited area (per cent).",
294	294	"label": "tab:net_area_flows",
295	295	"sources": [
296		− "./results/graph/area_flows_yearly.csv"
	296	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/area_flows_yearly.csv"
297	297	],
298	298	"rows": 16
299	299	},
		`@@ -302,8 +302,8 @@`
302	302	"caption": "Modularity of the yearly category citation-flow network at the category and area partitions, the within-category self-share, and the aggregate cross-area share of internal references, raw and composition-adjusted.",
303	303	"label": "tab:net_modularity",
304	304	"sources": [
305		− "./results/graph/modularity_yearly.csv",
306		− "./results/graph/cross_area_share_yearly.csv"
	305	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/modularity_yearly.csv",
	306	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/cross_area_share_yearly.csv"
307	307	],
308	308	"rows": 9
309	309	},
		`@@ -312,7 +312,7 @@`
312	312	"caption": "Share of each category's internal references received by the nine categories of the AI/ML area, by citing year (per cent).",
313	313	"label": "tab:net_category_to_ai",
314	314	"sources": [
315		− "./results/graph/category_flows_yearly.csv"
	315	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/category_flows_yearly.csv"
316	316	],
317	317	"rows": 40
318	318	},
		`@@ -321,7 +321,7 @@`
321	321	"caption": "Shares of works with a China-affiliated author, a United States-affiliated author, a company affiliation and authors in at least two countries, by primary category and period (per cent).",
322	322	"label": "tab:net_category_actors",
323	323	"sources": [
324		− "./results/bibliometrics/category_actors_yearly.csv"
	324	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/category_actors_yearly.csv"
325	325	],
326	326	"rows": 40
327	327	},
		`@@ -330,7 +330,7 @@`
330	330	"caption": "Actor shares of works whose primary category is in the AI/ML area against all other areas, by period (per cent).",
331	331	"label": "tab:net_ai_vs_rest",
332	332	"sources": [
333		− "./results/bibliometrics/category_actors_yearly.csv"
	333	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/category_actors_yearly.csv"
334	334	],
335	335	"rows": 4
336	336	},
		`@@ -339,7 +339,7 @@`
339	339	"caption": "China-affiliated, company-affiliated and international shares of works by year, AI/ML area against all other areas (per cent).",
340	340	"label": "tab:net_ai_vs_rest_yearly",
341	341	"sources": [
342		− "./results/bibliometrics/category_actors_yearly.csv"
	342	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/category_actors_yearly.csv"
343	343	],
344	344	"rows": 9
345	345	},
		`@@ -348,7 +348,7 @@`
348	348	"caption": "Gini coefficient of 36-month citations $c^{(36)}$ by primary category and publication year, for years with a complete window; baseline mean, change of the last complete year against it, and share uncited at 36 months in that year.",
349	349	"label": "tab:net_category_gini",
350	350	"sources": [
351		− "./results/bibliometrics/category_impact_yearly.csv"
	351	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/category_impact_yearly.csv"
352	352	],
353	353	"rows": 40
354	354	},
		`@@ -357,7 +357,7 @@`
357	357	"caption": "Citation inequality and levels at 36 months, AI/ML area against all other areas, by publication year with a complete window.",
358	358	"label": "tab:net_gini_ai_vs_rest",
359	359	"sources": [
360		− "./results/bibliometrics/category_impact_yearly.csv"
	360	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/category_impact_yearly.csv"
361	361	],
362	362	"rows": 6
363	363	},
		`@@ -366,7 +366,7 @@`
366	366	"caption": "Volume, compound annual growth and corpus share of the nine AI categories, baseline (2018--2022) and post (2023--2025) periods.",
367	367	"label": "tab:ai_growth",
368	368	"sources": [
369		− "./results/bibliometrics/cagr_by_category.csv"
	369	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/cagr_by_category.csv"
370	370	],
371	371	"rows": 9
372	372	},
		`@@ -375,7 +375,7 @@`
375	375	"caption": "Yearly multi-label counts of the nine AI categories (all period rows summed; 2026 incomplete).",
376	376	"label": "tab:ai_counts",
377	377	"sources": [
378		− "./results/graph/category_counts_yearly.csv"
	378	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/category_counts_yearly.csv"
379	379	],
380	380	"rows": 9
381	381	},
		`@@ -384,7 +384,7 @@`
384	384	"caption": "Structural change at 30 November 2022 in the nine AI categories: interrupted-time-series level and slope changes, placebo p-values, Bai--Perron and PELT break dates.",
385	385	"label": "tab:ai_breaks",
386	386	"sources": [
387		− "./results/econometrics/breaks.csv"
	387	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/econometrics/breaks.csv"
388	388	],
389	389	"rows": 9
390	390	},
		`@@ -393,7 +393,7 @@`
393	393	"caption": "Interrupted-time-series level changes of the nine AI categories at the alternative event dates (GPT-3, DALL-E 2, GPT-4, o1).",
394	394	"label": "tab:ai_altdates",
395	395	"sources": [
396		− "./results/econometrics/breaks.csv"
	396	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/econometrics/breaks.csv"
397	397	],
398	398	"rows": 9
399	399	},
		`@@ -402,7 +402,7 @@`
402	402	"caption": "Destination of the in-corpus references of the nine AI categories in 2019, 2022 and 2025 (per cent of in-corpus references of the citing category).",
403	403	"label": "tab:ai_flows",
404	404	"sources": [
405		− "./results/graph/category_flows_yearly.csv"
	405	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/category_flows_yearly.csv"
406	406	],
407	407	"rows": 27
408	408	},
		`@@ -411,7 +411,7 @@`
411	411	"caption": "Specialist screening of the 234 AI/ML landmark candidates: records by primary category and data-quality flag.",
412	412	"label": "tab:ai_landmark_flags",
413	413	"sources": [
414		− "./literature_synthesis/agent_E/contributions.csv"
	414	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/literature_synthesis/agent_E/contributions.csv"
415	415	],
416	416	"rows": 9
417	417	},
		`@@ -420,7 +420,7 @@`
420	420	"caption": "Contribution type of the 234 AI/ML landmark candidates by primary category (specialist judgement from abstract or metadata).",
421	421	"label": "tab:ai_contribution_types",
422	422	"sources": [
423		− "./literature_synthesis/agent_E/contributions.csv"
	423	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/literature_synthesis/agent_E/contributions.csv"
424	424	],
425	425	"rows": 9
426	426	},
		`@@ -429,7 +429,7 @@`
429	429	"caption": "The 234 AI/ML landmark candidates with the specialist's contribution type and data-quality flag (titles in literature\\_synthesis/agent\\_E/contributions.csv).",
430	430	"label": "tab:ai_landmarks",
431	431	"sources": [
432		− "./literature_synthesis/agent_E/contributions.csv"
	432	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/literature_synthesis/agent_E/contributions.csv"
433	433	],
434	434	"rows": 234
435	435	},
		`@@ -438,7 +438,7 @@`
438	438	"caption": "Eligible works in the five security-and-privacy categories by publication year: works carrying the category (all) and works whose primary category it is (prim.).",
439	439	"label": "tab:security_counts",
440	440	"sources": [
441		− "./results/graph/category_counts_yearly.csv"
	441	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/category_counts_yearly.csv"
442	442	],
443	443	"rows": 9
444	444	},
		`@@ -447,8 +447,8 @@`
447	447	"caption": "Growth and structural-change summary for the five security-and-privacy categories (multi-label counts).",
448	448	"label": "tab:security_summary",
449	449	"sources": [
450		− "./results/graph/category_counts_yearly.csv",
451		− "./results/econometrics/breaks.csv"
	450	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/category_counts_yearly.csv",
	451	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/econometrics/breaks.csv"
452	452	],
453	453	"rows": 5
454	454	},
		`@@ -457,8 +457,8 @@`
457	457	"caption": "Co-classification of security-and-privacy works with the LLM, AI-safety and AI-agents categories, by publication year: number of works carrying both categories and share of the security category's works.",
458	458	"label": "tab:security_llm_cooccurrence",
459	459	"sources": [
460		− "./results/graph/category_cooccurrence_yearly.csv",
461		− "./results/graph/category_counts_yearly.csv"
	460	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/category_cooccurrence_yearly.csv",
	461	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/category_counts_yearly.csv"
462	462	],
463	463	"rows": 9
464	464	},
		`@@ -467,7 +467,7 @@`
467	467	"caption": "Composition of the automatically selected landmark set of the security-and-privacy area, by category.",
468	468	"label": "tab:security_landmarks",
469	469	"sources": [
470		− "./literature_synthesis/landmarks/security_privacy_landmarks.csv"
	470	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/literature_synthesis/landmarks/security_privacy_landmarks.csv"
471	471	],
472	472	"rows": 5
473	473	},
		`@@ -476,8 +476,8 @@`
476	476	"caption": "The systems and infrastructure categories of Part VI: primary-category output, compound annual growth, and structural-change estimates of the monthly log-volume series.",
477	477	"label": "tab:systems_categories",
478	478	"sources": [
479		− "./results/bibliometrics/cagr_by_category.csv",
480		− "./results/econometrics/breaks.csv"
	479	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/cagr_by_category.csv",
	480	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/econometrics/breaks.csv"
481	481	],
482	482	"rows": 11
483	483	},
		`@@ -486,7 +486,7 @@`
486	486	"caption": "Eligible works in the six theory and software categories by publication year: works carrying the category (all) and works whose primary category it is (prim.).",
487	487	"label": "tab:theory_counts",
488	488	"sources": [
489		− "./results/graph/category_counts_yearly.csv"
	489	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/category_counts_yearly.csv"
490	490	],
491	491	"rows": 9
492	492	},
		`@@ -495,8 +495,8 @@`
495	495	"caption": "Growth and structural-change summary for the six theory and software categories (multi-label counts).",
496	496	"label": "tab:theory_summary",
497	497	"sources": [
498		− "./results/graph/category_counts_yearly.csv",
499		− "./results/econometrics/breaks.csv"
	498	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/category_counts_yearly.csv",
	499	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/econometrics/breaks.csv"
500	500	],
501	501	"rows": 6
502	502	},
		`@@ -505,8 +505,8 @@`
505	505	"caption": "Co-classification of theory and software works with the LLM category, by publication year: number of works carrying both categories and share of the category's multi-label count.",
506	506	"label": "tab:theory_llm_cooccurrence",
507	507	"sources": [
508		− "./results/graph/category_cooccurrence_yearly.csv",
509		− "./results/graph/category_counts_yearly.csv"
	508	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/category_cooccurrence_yearly.csv",
	509	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/category_counts_yearly.csv"
510	510	],
511	511	"rows": 9
512	512	},
		`@@ -515,9 +515,9 @@`
515	515	"caption": "Specialist audit of the automatically selected landmark candidates of the theory and software areas, by verdict and primary category.",
516	516	"label": "tab:theory_landmark_audit",
517	517	"sources": [
518		− "./literature_synthesis/agent_G/contributions.csv",
519		− "./literature_synthesis/landmarks/theory_landmarks.csv",
520		− "./literature_synthesis/landmarks/software_pl_landmarks.csv"
	518	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/literature_synthesis/agent_G/contributions.csv",
	519	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/literature_synthesis/landmarks/theory_landmarks.csv",
	520	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/literature_synthesis/landmarks/software_pl_landmarks.csv"
521	521	],
522	522	"rows": 9
523	523	},
		`@@ -526,7 +526,7 @@`
526	526	"caption": "Landmark candidates of the theory and software areas retained as in-scope research or survey contributions, with the contribution type judged from the abstract or metadata.",
527	527	"label": "tab:theory_landmarks_discussed",
528	528	"sources": [
529		− "./literature_synthesis/agent_G/contributions.csv"
	529	+ "/Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/literature_synthesis/agent_G/contributions.csv"
530	530	],
531	531	"rows": 37
532	532	}

modified manuscript/tables/tab_ai_altdates.tex +2 −2

		`@@ -1,9 +1,9 @@`
1		−% generated by tables/make_tables.py from ./results/econometrics/breaks.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/econometrics/breaks.csv
2	2	\begin{table}[htbp]\centering\caption{Interrupted-time-series level changes of the nine AI categories at the alternative event dates (GPT-3, DALL-E 2, GPT-4, o1).}\label{tab:ai_altdates}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrrrr}
5	5	\toprule
6		−\textbf{Category} & \textbf{GPT-3 (2020-06)} & \textbf{p} & \textbf{DALL-E 2 (2022-05)} & \textbf{p } & \textbf{GPT-4 (2023-04)} & \textbf{p } & \textbf{o1 (2024-09)} & \textbf{p } \\
	6	+\textbf{Category} & \textbf{GPT-3 (2020-06)} & \textbf{p} & \textbf{DALL-E 2 (2022-05)} & \textbf{p } & \textbf{GPT-4 (2023-04)} & \textbf{p } & \textbf{o1 (2024-10)} & \textbf{p } \\
7	7	\midrule
8	8	LLM & -0.380 & 0.007 & -0.182 & 0.002 & 0.236 & 0.002 & 0.457 & 0.001 \\
9	9	Foundation & -0.164 & 0.049 & -0.221 & 0.002 & 0.023 & 0.756 & 0.084 & 0.173 \\

modified manuscript/tables/tab_ai_breaks.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/econometrics/breaks.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/econometrics/breaks.csv
2	2	\begin{table}[htbp]\centering\caption{Structural change at 30 November 2022 in the nine AI categories: interrupted-time-series level and slope changes, placebo p-values, Bai--Perron and PELT break dates.}\label{tab:ai_breaks}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrrrr}

modified manuscript/tables/tab_ai_contribution_types.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./literature_synthesis/agent_E/contributions.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/literature_synthesis/agent_E/contributions.csv
2	2	\begin{table}[htbp]\centering\caption{Contribution type of the 234 AI/ML landmark candidates by primary category (specialist judgement from abstract or metadata).}\label{tab:ai_contribution_types}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrrr}

modified manuscript/tables/tab_ai_counts.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/graph/category_counts_yearly.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/category_counts_yearly.csv
2	2	\begin{table}[htbp]\centering\caption{Yearly multi-label counts of the nine AI categories (all period rows summed; 2026 incomplete).}\label{tab:ai_counts}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrrrrr}

modified manuscript/tables/tab_ai_flows.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/graph/category_flows_yearly.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/category_flows_yearly.csv
2	2	\begin{longtable}{lrrrrrrrrr}
3	3	\caption{Destination of the in-corpus references of the nine AI categories in 2019, 2022 and 2025 (per cent of in-corpus references of the citing category).}\label{tab:ai_flows}\\
4	4	\toprule

modified manuscript/tables/tab_ai_growth.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/bibliometrics/cagr_by_category.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/cagr_by_category.csv
2	2	\begin{table}[htbp]\centering\caption{Volume, compound annual growth and corpus share of the nine AI categories, baseline (2018--2022) and post (2023--2025) periods.}\label{tab:ai_growth}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrrrrr}

modified manuscript/tables/tab_ai_landmark_flags.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./literature_synthesis/agent_E/contributions.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/literature_synthesis/agent_E/contributions.csv
2	2	\begin{table}[htbp]\centering\caption{Specialist screening of the 234 AI/ML landmark candidates: records by primary category and data-quality flag.}\label{tab:ai_landmark_flags}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrrrr}

modified manuscript/tables/tab_ai_landmarks.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./literature_synthesis/agent_E/contributions.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/literature_synthesis/agent_E/contributions.csv
2	2	\begin{longtable}{lrrrrrrr}
3	3	\caption{The 234 AI/ML landmark candidates with the specialist's contribution type and data-quality flag (titles in literature\_synthesis/agent\_E/contributions.csv).}\label{tab:ai_landmarks}\\
4	4	\toprule

modified manuscript/tables/tab_breaks.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/econometrics/breaks.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/econometrics/breaks.csv
2	2	\begin{longtable}{lrrrrrrrr}
3	3	\caption{Structural-change estimates by series: ITS level and slope changes at 2022-12 (HAC p-values), placebo p-values, Bai–Perron (BIC) and PELT break dates.}\label{tab:breaks}\\
4	4	\toprule

modified manuscript/tables/tab_cited_age.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/graph/cited_age_yearly.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/cited_age_yearly.csv
2	2	\begin{table}[htbp]\centering\caption{Age of cited works by citing year (resolved references of eligible works; corpus version of the result file at generation time).}\label{tab:cited_age}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrrr}

modified manuscript/tables/tab_collaboration.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/bibliometrics/team_size_yearly.csv, ./results/bibliometrics/international_collab_yearly.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/team_size_yearly.csv, /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/international_collab_yearly.csv
2	2	\begin{table}[htbp]\centering\caption{Team size and collaboration indicators by publication year (all eligible works; periods of the source file summed).}\label{tab:collaboration}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrrrr}

modified manuscript/tables/tab_countries.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/bibliometrics/countries_yearly.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/countries_yearly.csv
2	2	\begin{table}[htbp]\centering\caption{Fractional output of the fifteen largest countries by year.}\label{tab:countries}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrrrrr}

modified manuscript/tables/tab_country_concentration.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/bibliometrics/countries_yearly.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/countries_yearly.csv
2	2	\begin{table}[htbp]\centering\caption{Concentration of output across countries by year, and the two largest producers under full counting.}\label{tab:country_concentration}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrrrrr}

modified manuscript/tables/tab_country_network.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/bibliometrics/country_pairs_by_period.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/country_pairs_by_period.csv
2	2	\begin{table}[htbp]\centering\caption{The country co-authorship network by period: edges with at least fifty joint works.}\label{tab:country_network}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrr}

modified manuscript/tables/tab_country_pairs.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/bibliometrics/country_pairs_by_period.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/country_pairs_by_period.csv
2	2	\begin{table}[htbp]\centering\caption{The twenty strongest country pairs by period (number of joint works).}\label{tab:country_pairs}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrr}

modified manuscript/tables/tab_country_shares.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/bibliometrics/countries_yearly.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/countries_yearly.csv
2	2	\begin{table}[htbp]\centering\caption{Share of fractional output of the fifteen largest countries by year (per cent).}\label{tab:country_shares}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrrrrr}

modified manuscript/tables/tab_dblp_only.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/bibliometrics/dblp_only_yearly.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/dblp_only_yearly.csv
2	2	\begin{table}[htbp]\centering\caption{DBLP records of 2018--2026 without an OpenAlex counterpart, by DBLP record type and year.}\label{tab:dblp_only}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrrrr}

modified manuscript/tables/tab_emerging_h12.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/emerging/metrics_h12.json
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/emerging/metrics_h12.json
2	2	\begin{table}[htbp]\centering\caption{Out-of-time performance of the emerging-topic detectors at horizon 12 months (test cutoffs 2023-01–2024-10).}\label{tab:emerging_h12}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrr}

modified manuscript/tables/tab_emerging_h24.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/emerging/metrics_h24.json
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/emerging/metrics_h24.json
2	2	\begin{table}[htbp]\centering\caption{Out-of-time performance of the emerging-topic detectors at horizon 24 months (test cutoffs 2023-01–2024-04).}\label{tab:emerging_h24}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrr}

modified manuscript/tables/tab_field_flows.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/graph/field_flows_yearly.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/field_flows_yearly.csv
2	2	\begin{table}[htbp]\centering\caption{Destination of the resolved references of eligible works by citing year: inside the corpus, computer science outside it, and the seven largest other OpenAlex fields (per cent of resolved references).}\label{tab:field_flows}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrrrrrr}

modified manuscript/tables/tab_indexing_lag.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/bibliometrics/indexing_lag.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/indexing_lag.csv
2	2	\begin{table}[htbp]\centering\caption{The OpenAlex \texttt{created\_date} relative to the publication date, by publication year.}\label{tab:indexing_lag}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrr}

modified manuscript/tables/tab_institution_concentration.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/bibliometrics/top_institutions_by_period.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/top_institutions_by_period.csv
2	2	\begin{table}[htbp]\centering\caption{Institutions with at least two hundred works per period: coverage and concentration (fractional counting).}\label{tab:institution_concentration}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrr}

modified manuscript/tables/tab_institution_types.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/bibliometrics/institution_types_yearly.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/institution_types_yearly.csv
2	2	\begin{table}[htbp]\centering\caption{Share of typed affiliations by OpenAlex institution type and year (per cent of affiliation-work pairs).}\label{tab:institution_types}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrrrrr}

modified manuscript/tables/tab_known_set.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/emerging/known_set_lead_times.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/emerging/known_set_lead_times.csv
2	2	\begin{table}[htbp]\centering\caption{Preregistered known-emergence set: detection month, peak-growth month and lead time.}\label{tab:known_set}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrr}

modified manuscript/tables/tab_net_ai_vs_rest.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/bibliometrics/category_actors_yearly.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/category_actors_yearly.csv
2	2	\begin{table}[htbp]\centering\caption{Actor shares of works whose primary category is in the AI/ML area against all other areas, by period (per cent).}\label{tab:net_ai_vs_rest}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrrrr}

modified manuscript/tables/tab_net_ai_vs_rest_yearly.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/bibliometrics/category_actors_yearly.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/category_actors_yearly.csv
2	2	\begin{table}[htbp]\centering\caption{China-affiliated, company-affiliated and international shares of works by year, AI/ML area against all other areas (per cent).}\label{tab:net_ai_vs_rest_yearly}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrr}

modified manuscript/tables/tab_net_area_flows.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/graph/area_flows_yearly.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/area_flows_yearly.csv
2	2	\begin{table}[htbp]\centering\caption{Citation-flow matrices between the eight areas of the ontology by period: share of each citing area's internal references received by each cited area (per cent).}\label{tab:net_area_flows}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrrrrr}

modified manuscript/tables/tab_net_category_actors.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/bibliometrics/category_actors_yearly.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/category_actors_yearly.csv
2	2	\begin{longtable}{lrrrrrrrrrr}
3	3	\caption{Shares of works with a China-affiliated author, a United States-affiliated author, a company affiliation and authors in at least two countries, by primary category and period (per cent).}\label{tab:net_category_actors}\\
4	4	\toprule

modified manuscript/tables/tab_net_category_gini.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/bibliometrics/category_impact_yearly.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/category_impact_yearly.csv
2	2	\begin{longtable}{lrrrrrrrrr}
3	3	\caption{Gini coefficient of 36-month citations $c^{(36)}$ by primary category and publication year, for years with a complete window; baseline mean, change of the last complete year against it, and share uncited at 36 months in that year.}\label{tab:net_category_gini}\\
4	4	\toprule

modified manuscript/tables/tab_net_category_to_ai.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/graph/category_flows_yearly.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/category_flows_yearly.csv
2	2	\begin{longtable}{lrrrrrrrrrr}
3	3	\caption{Share of each category's internal references received by the nine categories of the AI/ML area, by citing year (per cent).}\label{tab:net_category_to_ai}\\
4	4	\toprule

modified manuscript/tables/tab_net_gini_ai_vs_rest.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/bibliometrics/category_impact_yearly.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/category_impact_yearly.csv
2	2	\begin{table}[htbp]\centering\caption{Citation inequality and levels at 36 months, AI/ML area against all other areas, by publication year with a complete window.}\label{tab:net_gini_ai_vs_rest}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrrrrrr}

modified manuscript/tables/tab_net_modularity.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/graph/modularity_yearly.csv, ./results/graph/cross_area_share_yearly.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/modularity_yearly.csv, /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/cross_area_share_yearly.csv
2	2	\begin{table}[htbp]\centering\caption{Modularity of the yearly category citation-flow network at the category and area partitions, the within-category self-share, and the aggregate cross-area share of internal references, raw and composition-adjusted.}\label{tab:net_modularity}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrr}

modified manuscript/tables/tab_reference_coverage.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/graph/reference_coverage_yearly.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/reference_coverage_yearly.csv
2	2	\begin{table}[htbp]\centering\caption{Reference coverage by citing year: declared, resolved, internal and external references of eligible works.}\label{tab:reference_coverage}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrrr}

modified manuscript/tables/tab_security_counts.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/graph/category_counts_yearly.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/category_counts_yearly.csv
2	2	\begin{longtable}{lrrrrrrrrrr}
3	3	\caption{Eligible works in the five security-and-privacy categories by publication year: works carrying the category (all) and works whose primary category it is (prim.).}\label{tab:security_counts}\\
4	4	\toprule

modified manuscript/tables/tab_security_landmarks.tex +6 −6

		`@@ -1,15 +1,15 @@`
1		−% generated by tables/make_tables.py from ./literature_synthesis/landmarks/security_privacy_landmarks.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/literature_synthesis/landmarks/security_privacy_landmarks.csv
2	2	\begin{table}[htbp]\centering\caption{Composition of the automatically selected landmark set of the security-and-privacy area, by category.}\label{tab:security_landmarks}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrrrrr}
5	5	\toprule
6	6	\textbf{Category} & \textbf{Works} & \textbf{Post-period} & \textbf{Dated 2026} & \textbf{Articles} & \textbf{Conf.} & \textbf{Preprints} & \textbf{Grey} & \textbf{Abstract} & \textbf{No DOI} \\
7	7	\midrule
8		−Cybersecurity & 25 & 16 & 9 & 13 & 1 & 1 & 10 & 20 & 5 \\
9		−Applied cryptography & 25 & 15 & 8 & 11 & 2 & 0 & 12 & 18 & 1 \\
10		−Blockchain systems & 25 & 6 & 2 & 14 & 1 & 5 & 5 & 19 & 0 \\
11		−Privacy-preserving computation & 25 & 14 & 8 & 16 & 3 & 1 & 5 & 21 & 1 \\
12		−Federated learning & 25 & 15 & 6 & 18 & 4 & 0 & 3 & 14 & 2 \\
	8	+Cybersecurity & 25 & 12 & 3 & 10 & 2 & 1 & 12 & 21 & 7 \\
	9	+Applied cryptography & 25 & 11 & 5 & 13 & 1 & 0 & 11 & 19 & 1 \\
	10	+Blockchain systems & 25 & 6 & 1 & 15 & 0 & 3 & 7 & 20 & 0 \\
	11	+Privacy-preserving computation & 25 & 12 & 6 & 16 & 3 & 1 & 5 & 22 & 2 \\
	12	+Federated learning & 25 & 15 & 6 & 18 & 5 & 0 & 2 & 15 & 2 \\
13	13	\bottomrule
14	14	\end{tabular}
15	15	\begin{tablenotes}\footnotesize\item Landmarks are the works with the highest field-normalised 36-month citation score (NCS36; 2025--2026 works ranked by 12-month citations) within each primary category, with a floor of fifteen per category, plus the 200 most cited works of the corpus overall (\texttt{literature\_synthesis/select\_landmarks.py}). Grey = dissertations, reports, book chapters and books as typed by OpenAlex. Abstract = works whose abstract is available in the metadata; the synthesis of Part VII reads these abstracts only, never full texts. The set is flagged for principal-investigator review; Part VII documents the records that are duplicates, misdated or misclassified.\end{tablenotes}

modified manuscript/tables/tab_security_llm_cooccurrence.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/graph/category_cooccurrence_yearly.csv, ./results/graph/category_counts_yearly.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/category_cooccurrence_yearly.csv, /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/category_counts_yearly.csv
2	2	\begin{longtable}{lrrrrrrrrrrrrrr}
3	3	\caption{Co-classification of security-and-privacy works with the LLM, AI-safety and AI-agents categories, by publication year: number of works carrying both categories and share of the security category's works.}\label{tab:security_llm_cooccurrence}\\
4	4	\toprule

modified manuscript/tables/tab_security_summary.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/graph/category_counts_yearly.csv, ./results/econometrics/breaks.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/category_counts_yearly.csv, /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/econometrics/breaks.csv
2	2	\begin{table}[htbp]\centering\caption{Growth and structural-change summary for the five security-and-privacy categories (multi-label counts).}\label{tab:security_summary}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrrrrrrr}

modified manuscript/tables/tab_source_types.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/bibliometrics/top_sources_by_period.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/top_sources_by_period.csv
2	2	\begin{table}[htbp]\centering\caption{Distribution of works over source types among the listed sources, by period (per cent).}\label{tab:source_types}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrrr}

modified manuscript/tables/tab_systems_categories.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/bibliometrics/cagr_by_category.csv, ./results/econometrics/breaks.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/bibliometrics/cagr_by_category.csv, /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/econometrics/breaks.csv
2	2	\begin{table}[htbp]\centering\caption{The systems and infrastructure categories of Part VI: primary-category output, compound annual growth, and structural-change estimates of the monthly log-volume series.}\label{tab:systems_categories}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrrrrr}

modified manuscript/tables/tab_theory_counts.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/graph/category_counts_yearly.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/category_counts_yearly.csv
2	2	\begin{longtable}{lrrrrrrrrrrrr}
3	3	\caption{Eligible works in the six theory and software categories by publication year: works carrying the category (all) and works whose primary category it is (prim.).}\label{tab:theory_counts}\\
4	4	\toprule

modified manuscript/tables/tab_theory_landmark_audit.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./literature_synthesis/agent_G/contributions.csv, ./literature_synthesis/landmarks/theory_landmarks.csv, ./literature_synthesis/landmarks/software_pl_landmarks.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/literature_synthesis/agent_G/contributions.csv, /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/literature_synthesis/landmarks/theory_landmarks.csv, /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/literature_synthesis/landmarks/software_pl_landmarks.csv
2	2	\begin{table}[htbp]\centering\caption{Specialist audit of the automatically selected landmark candidates of the theory and software areas, by verdict and primary category.}\label{tab:theory_landmark_audit}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrrrrr}

modified manuscript/tables/tab_theory_landmarks_discussed.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./literature_synthesis/agent_G/contributions.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/literature_synthesis/agent_G/contributions.csv
2	2	\begin{longtable}{lrrrrr}
3	3	\caption{Landmark candidates of the theory and software areas retained as in-scope research or survey contributions, with the contribution type judged from the abstract or metadata.}\label{tab:theory_landmarks_discussed}\\
4	4	\toprule

modified manuscript/tables/tab_theory_llm_cooccurrence.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/graph/category_cooccurrence_yearly.csv, ./results/graph/category_counts_yearly.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/category_cooccurrence_yearly.csv, /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/category_counts_yearly.csv
2	2	\begin{longtable}{lrrrrrrrrrrrr}
3	3	\caption{Co-classification of theory and software works with the LLM category, by publication year: number of works carrying both categories and share of the category's multi-label count.}\label{tab:theory_llm_cooccurrence}\\
4	4	\toprule

modified manuscript/tables/tab_theory_summary.tex +1 −1

		`@@ -1,4 +1,4 @@`
1		−% generated by tables/make_tables.py from ./results/graph/category_counts_yearly.csv, ./results/econometrics/breaks.csv
	1	+% generated by tables/make_tables.py from /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/graph/category_counts_yearly.csv, /Users/simon-pierreboucher/Desktop/UQO/Recherche/Recherche-UQO/informatique/wp44/results/econometrics/breaks.csv
2	2	\begin{table}[htbp]\centering\caption{Growth and structural-change summary for the six theory and software categories (multi-label counts).}\label{tab:theory_summary}
3	3	\begin{threeparttable}\small
4	4	\begin{tabular}{lrrrrrrrrrrr}

modified manuscript/tables/tab_top_cited.tex +1 −1

Large diff (5 lines) is not loaded by default to keep the page fast.Load diff

modified manuscript/tables/tab_top_institutions.tex +1 −1

Large diff (5 lines) is not loaded by default to keep the page fast.Load diff

modified manuscript/tables/tab_top_sources.tex +1 −1

Large diff (5 lines) is not loaded by default to keep the page fast.Load diff

modified tables/make_tables.py +1 −1

Large diff (9 lines) is not loaded by default to keep the page fast.Load diff

spb/wp44_uqo

Theme