Deep learning–based knowledge tracing, 2015–2025: derived bibliometric and taxonomy tables (Scientometrics submission)

v4.0.0 updates the locked Scopus-core deposit (N=926; identity 1287 = 926 + 361 + 0) after a human-only recode of the 145 newly eligible records (batch A2-NEW-145-HUMAN-RECODE-v1). What changed vs v3.0.0 (2026-08-05):- Record-level TITLE-ABS-KW taxonomy in taxonomy_coding_record_level_n926.csv after human recode. Pooled UNCERTAIN = 335/926 (36.2%; NL2 268/781 = 34.3%; NEW 67/145 = 46.2%).- secondary_mechanisms coded on the 145 (NONE = 37/145, not 145/145).- Frequency tables regenerated on N=926 (v3 freq_primary_family.csv still used the NL2 n=781 subset).- Coding-pathway comparison (NL2 781 vs NEW 145 vs pooled 926).- Human-only IRR n=40 summaries (κ = 0.731 on primary_family); the n=18 AI-draft overlap is discarded.- Keyword modularity decomposition (B4 four-config factorial).- A3 residual-validation tables remain locked to the 2026-08-03 residual (N_U = 277; Horvitz–Thompson 15.1% / 30.4%) and are not recomputed on the current residual 335. Not included: licensed Scopus raw exports; copyrighted full-text PDFs; coordinator_only/ AI-draft labels; coder working sheets. Claim boundaries: TITLE-ABS-KW codes, not a full-text census of N=926; A3 is a stratified residual audit (n=194 AVAILABLE of 200), not prevalence for N=926; networks are exploratory and threshold-dependent. Companion manuscript: Scientometrics submission. Concept DOI (all versions): 10.5281/zenodo.21635880.