Empirical Calibration, 35-Year Longitudinal Secular Trends (1991–2025), Human-vs.-AI Open-Review Concordance, and Structural Error Estimation across Quantitative Biology & Biomedicine
Click any bar to filter the table below to that cohort. Strict Failure (<5.0) and Major Flaw (<6.5) rates shown on right axis.
Updates dynamically as you search or filter by Cohort, Subfield, Venue Tier, Decade, or Modality below.
Established selective/society journals held their standards (-0.01 pts/decade, p = 0.96), while Open-Access Megajournals expanded from 0% to 51.4% of volume (88% Oaxaca–Blinder compositional dilution).
Mean Composite Score by Study Modality, Technology Maturity, and Funding Tier (NIH R01/U01/P01 vs. Unfunded).
Selective Flagship/Society Journals (eLife, PLOS Biol/Genet/Pathog/CompBiol) vs. Open-Review Megajournals (PLOS ONE, PeerJ, BMJ Open, RSOS).
Every paper in the Open-Review Cohort includes both our blinded Gemini 4 Argon review and the verbatim published human referee reports. Click "Has Human Peer Review" below to inspect any pair side-by-side.
Multi-reviewer technical scrutiny (135 human reviewers, median 3,977 words; 72% Thorough Technical, 0% Superficial) yields high published quality (7.49/10, 8.0% <5.0). Blinded AI recalls 97.7% (42/43) of unresolved human critiques while catching 172 additional quantitative/design flaws (3.4/paper).
Despite open peer review (126 human reviewers, median 2,462 words; 48.0% strict failure <5.0), 38% of human review records are Superficial/Boilerplate and 58% (29/50) exhibit anomalies: generic/LLM-leaked boilerplate (19), coercive citation demands (6; e.g., OPEN_ANON_001, OPEN_ANON_011), or single-reviewer acceptances (4; e.g., OPEN_ANON_036), missing 213 critical flaws (4.3/paper).
Example: TIME_ANON_0224 (Pathogens 2020, 3.7/10, 156 cites) claimed SARS-CoV-2 is a MERS-CoV recombinant from a 7-taxon ClustalW alignment; TIME_ANON_0922 (2.2/10, 139 cites) docked Lopinavir+Ritonavir simultaneously as a single 1,300 Da molecule.
Example: TIME_ANON_08 (BMC Evol Biol 2005, 4.8/10, 158 cites) curated relaxin genes but plotted PAML's dN/dS = 99 division-by-zero code (`dS = 0`) as real ω = 99 positive selection; TIME_ANON_0605 (4.9/10, 73 cites) cataloged 209 millet MYB genes with a copy-pasted "MAPK" conclusion.
Example: SP_ANON_109 (Nature 2016, 4.1/10, 556 cites) inferred ongoing HIV replication under ART from the first 6 months of viral decay in 3 patients; TIME_ANON_0867 (Mol Cancer 2004, 4.5/10, 111 cites) claimed cisplatin induces de novo methylation despite 0/18 genes methylated in vitro.
Example: 2024–2025 papers combining scRNA-seq + Mendelian randomization + LASSO Cox + molecular docking (e.g., C21 at 2.8/10, 39 cites; C86 at 4.4/10, 36 cites) cited rapidly as introduction filler by other template papers.
First/Senior Author (n = 42: Mean 8.22 ± 1.11, Median 8.50, 0% <5.0) vs. Middle Co-Author (n = 122: Mean 7.29 ± 1.39, 6.6% <5.0; p = 3.1 × 10⁻⁵).
All 164 PMC papers were scrubbed of author names, titles, journals, and identifiers before blinded review.
Zero secular decline over 22 years (2000s Mean 8.14, 2010s Mean 8.30, 2020s Mean 8.16; r = +0.039, p = 0.81). 76.2% (32/42) score ≥8.0/10.
Flagged 8 middle co-author papers below 5.0 (including SP_ANON_103 at 2.2 and SP_ANON_109 at 4.1) and caught the BA.5 amplicon-dropout artifact in SP_ANON_012 (6.4) that was subsequently refuted in SP_ANON_007 (8.9).
| ID / Code | Manuscript Title & Attributes | Venue & Year | Cohort / Subfield | Cites | Rigor /3.5 |
Claims /2.5 |
Bio /2.0 |
Repro /2.0 |
Composite /10.0 |
Dossier |
|---|---|---|---|---|---|---|---|---|---|---|
| Loading catalog.json... | ||||||||||