# Passage Re-ranking with BERT

**Authors:** Rodrigo Nogueira, Kyunghyun Cho  
**arXiv:** [1901.04085](https://arxiv.org/abs/1901.04085)  
**Date:** 2019-01-13  
**Summary page:** https://www.summarizepaper.com/en/paper/1901.04085v1/

## Key points

- Authors Rodrigo Nogueira and Kyunghyun Cho discuss advancements in natural language processing achieved by neural models pretrained on language modeling tasks such as ELMo, OpenAI GPT, and BERT.
- The authors specifically focus on BERT and present a re-implementation of this model for query-based passage re-ranking.
- Their system demonstrated state-of-the-art performance on the TREC-CAR dataset and emerged as the top entry in the MS MARCO passage retrieval task leaderboard.
- Their approach outperformed previous benchmarks by 27% relative improvement in Mean Reciprocal Rank at 10 (MRR@10).
- The availability of their code for reproduction on GitHub enhances the reproducibility and accessibility of their work.
- This study highlights the effectiveness of leveraging pretrained neural models like BERT for enhancing information retrieval tasks, showcasing significant advancements in natural language processing.

## Summary

In their paper titled "Passage Re-ranking with BERT," authors Rodrigo Nogueira and Kyunghyun Cho discuss the advancements in natural language processing achieved by neural models pretrained on language modeling tasks such as ELMo, OpenAI GPT, and BERT. Specifically focusing on BERT, the authors present a re-implementation of this model for query-based passage re-ranking. Their system has demonstrated state-of-the-art performance on the TREC-CAR dataset and emerged as the top entry in the MS MARCO passage retrieval task leaderboard. Impressively, their approach outperformed previous benchmarks by 27% relative improvement in Mean Reciprocal Rank at 10 (MRR@10). The availability of their code for reproduction on GitHub further enhances the reproducibility and accessibility of their work. This study underscores the effectiveness of leveraging pretrained neural models like BERT for enhancing information retrieval tasks, showcasing significant advancements in the field of natural language processing.