![[2602.18824] UniRank: A Multi-Agent Calibration Pipeline for Estimating University Rankings from Anonymized Bibliometric Signals](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.18824] UniRank: A Multi-Agent Calibration Pipeline for Estimating University Rankings from Anonymized Bibliometric Signals
Summary
UniRank introduces a multi-agent pipeline that estimates university rankings using anonymized bibliometric data, achieving notable accuracy without revealing institutional identities.
Why It Matters
This research addresses the challenge of estimating university rankings while maintaining data privacy. By utilizing anonymized bibliometric signals, it offers a novel approach that could enhance the credibility and fairness of ranking systems, which are often criticized for transparency issues. The findings could influence how universities assess their performance and how rankings are perceived globally.
Key Takeaways
- UniRank employs a three-stage architecture for ranking estimation.
- The system uses anonymized data to prevent memorization and bias.
- Performance varies significantly across university tiers, indicating analytical reasoning.
- Achieves a Memorization Index of zero, ensuring genuine predictions.
- The findings may reshape perceptions of university rankings and their methodologies.
Computer Science > Social and Information Networks arXiv:2602.18824 (cs) [Submitted on 21 Feb 2026] Title:UniRank: A Multi-Agent Calibration Pipeline for Estimating University Rankings from Anonymized Bibliometric Signals Authors:Pedram Riyazimehr, Seyyed Ehsan Mahmoudi View a PDF of the paper titled UniRank: A Multi-Agent Calibration Pipeline for Estimating University Rankings from Anonymized Bibliometric Signals, by Pedram Riyazimehr and 1 other authors View PDF HTML (experimental) Abstract:We present UniRank, a multi-agent LLM pipeline that estimates university positions across global ranking systems using only publicly available bibliometric data from OpenAlex and Semantic Scholar. The system employs a three-stage architecture: (a) zero-shot estimation from anonymized institutional metrics, (b) per-system tool-augmented calibration against real ranked universities, and (c) final synthesis. Critically, institutions are anonymized -- names, countries, DOIs, paper titles, and collaboration countries are all redacted -- and their actual ranks are hidden from the calibration tools during evaluation, preventing LLM memorization from confounding results. On the Times Higher Education (THE) World University Rankings ($n=352$), the system achieves MAE = 251.5 rank positions, Median AE = 131.5, PNMAE = 12.03%, Spearman $\rho = 0.769$, Kendall $\tau = 0.591$, hit rate @50 = 20.7%, hit rate @100 = 39.8%, and a Memorization Index of exactly zero (no exact-match zero-width predictio...
