Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Pipitone, Nicholas, Alami, Ghita Houir, Avadhanam, Advaith, Kaminskyi, Anton, Khoo, Ashley
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2509.12541
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866909790228185088
author Pipitone, Nicholas
Alami, Ghita Houir
Avadhanam, Advaith
Kaminskyi, Anton
Khoo, Ashley
author_facet Pipitone, Nicholas
Alami, Ghita Houir
Avadhanam, Advaith
Kaminskyi, Anton
Khoo, Ashley
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2509_12541
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle zELO: ELO-inspired Training Method for Rerankers and Embedding Models
Pipitone, Nicholas
Alami, Ghita Houir
Avadhanam, Advaith
Kaminskyi, Anton
Khoo, Ashley
Artificial Intelligence
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.
title zELO: ELO-inspired Training Method for Rerankers and Embedding Models
topic Artificial Intelligence
url https://arxiv.org/abs/2509.12541