qqe
Evaluates whether citation growth outpaces publication expansion across leading AI/NLP conferences, measuring the elasticity of scholarly impact relative to scale-driven growth over a decade. Use when
SKILL.mdskill-factory/output/qqe/SKILL.md
GitHubqqe
Has ACL Lost Its Crown? A Decade-Long Quantitative Analysis of Scale and Impact Across Leading AI Conferences — Ma et al. (2025) (arXiv:2512.04448, 2025)
What this evaluates
Evaluates whether citation growth outpaces publication expansion across leading AI/NLP conferences, measuring the elasticity of scholarly impact relative to scale-driven growth over a decade.
Datasets
- Leading AI/NLP Conferences (ACL, EMNLP, NAACL, AAAI, IJCAI, ICLR, NeurIPS) — total ?; splits: 2014-2024 (-1)
Metrics
QQE(primary) — range: other (unitless ratio)- Quality-Quantity Elasticity; measures the elasticity of citation growth relative to publication/acceptance growth. Calculated as the ratio of annual citation growth rate to annual publication growth rate.
AC— range: other- Average citations per paper; total citations divided by total accepted papers.
MC— range: other- Median citations per paper; the middle value of the citation distribution.
HCR— range: percent- Highly-cited ratio; percentage of papers exceeding a citation threshold.
ZCR— range: percent- Zero-cited ratio; percentage of papers with zero citations.
Gini— range: [0, 1]- Gini coefficient; measures inequality in the distribution of citations across papers.
Input / output format
Input: Annual counts of accepted papers and their subsequent citation counts for each conference from 2014 to 2024.
Output: Computed bibliometric indices per conference, including QQE, average/median citations, highly-cited/zero-cited ratios, H-index variants, and citation network shares.
Scoring recipe
def compute_bibliometrics(papers):
pub_count = len(papers)
cite_counts = [p.citations for p in papers]
ac = sum(cite_counts) / pub_count
mc = sorted(cite_counts)[pub_count // 2]
hcr = sum(1 for c in cite_counts if c > threshold) / pub_count
zcr = sum(1 for c in cite_counts if c == 0) / pub_count
return {'AC': ac, 'MC': mc, 'HCR': hcr, 'ZCR': zcr}
def compute_qqe(pub_series, cite_series):
pub_growth = (pub_series[-1] - pub_series[0]) / pub_series[0]
cite_growth = (cite_series[-1] - cite_series[0]) / cite_series[0]
return cite_growth / pub_growth if pub_growth != 0 else float('inf')
Common pitfalls
- QQE values fluctuate significantly year-to-year (e.g., negative values in some years), so trend analysis requires smoothing or long-term windows.
- Normalized metrics (e.g., Norm-H) adjust for scale, making direct comparison with absolute metrics (e.g., H-index) misleading without context.
- Citation counts depend heavily on the observation window; newer conferences or recent years will appear artificially low.
Evidence (verbatim from paper)
The QQE index evaluates whether citation growth outpaces publication expansion. All venues exhibit QQE > 1.0, indicating that scale growth has not diluted impact density.
Citation
@misc{ma2025hasacllostitscrown,
title={Has ACL Lost Its Crown? A Decade-Long Quantitative Analysis of Scale and Impact Across Leading AI Conferences},
author={Ma et al. (2025)},
year={2025},
note={arXiv:2512.04448}
}
- arXiv: 2512.04448