English

zELO: ELO-inspired Training Method for Rerankers and Embedding Models

Artificial Intelligence 2025-09-17 v1

Abstract

We introduce a novel training methodology named zELO, which optimizes retrieval performance via the analysis that ranking tasks are statically equivalent to a Thurstone model. Based on the zELO method, we use unsupervised data in order train a suite of state-of-the-art open-weight reranker models: zerank-1 and zerank-1-small. These models achieve the highest retrieval scores in multiple domains, including finance, legal, code, and STEM, outperforming closed-source proprietary rerankers on both NDCG@10 and Recall. These models also demonstrate great versatility, maintaining their 0-shot performance on out-of-domain and private customer datasets. The training data included 112,000 queries and 100 documents per query, and was trained end-to-end from unannotated queries and documents in less than 10,000 H100-hours.

Keywords

Cite

@article{arxiv.2509.12541,
  title  = {zELO: ELO-inspired Training Method for Rerankers and Embedding Models},
  author = {Nicholas Pipitone and Ghita Houir Alami and Advaith Avadhanam and Anton Kaminskyi and Ashley Khoo},
  journal= {arXiv preprint arXiv:2509.12541},
  year   = {2025}
}

Comments

13 pages, 9 sections, 17 figures and tables

R2 v1 2026-07-01T05:38:09.352Z