Skip to content

Commit

Permalink
add pdf link back
Browse files Browse the repository at this point in the history
  • Loading branch information
EC2 Default User committed Dec 5, 2023
1 parent b815931 commit 62aafc6
Showing 1 changed file with 1 addition and 1 deletion.
2 changes: 1 addition & 1 deletion README.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
# Scalable and Effective Generative Information Retrieval
This repo provides the source code and checkpoints for our paper [Scalable and Effective Generative Information Retrieval]() (RIPOR). We propose RIPOR, a optimization framework for generative retrieval. RIPOR is designed based on two often-overlooked fundamental design considerations in generative retrieval. To addresse the issues, we propose a novel prefix-oriented ranking optimization algorithm and relevance-based DocID initialization, which illustrated in the following Figure. The main experiment is conducted on large-scale information retrieval benchmark MSMARCO-8.8M, and evaluated on three evaluation sets MSMARCO-Dev, TREC'19 and 20. RIPOR surpasses state-of-the-art generative retrieval models by a large margin (e.g., 30.5% MRR improvements on MS MARCO Dev Set), and perform better on par with popular dense retrieval models.
This repo provides the source code and checkpoints for our paper [Scalable and Effective Generative Information Retrieval](https://arxiv.org/pdf/2311.09134.pdf) (RIPOR). We propose RIPOR, a optimization framework for generative retrieval. RIPOR is designed based on two often-overlooked fundamental design considerations in generative retrieval. To addresse the issues, we propose a novel prefix-oriented ranking optimization algorithm and relevance-based DocID initialization, which illustrated in the following Figure. The main experiment is conducted on large-scale information retrieval benchmark MSMARCO-8.8M, and evaluated on three evaluation sets MSMARCO-Dev, TREC'19 and 20. RIPOR surpasses state-of-the-art generative retrieval models by a large margin (e.g., 30.5% MRR improvements on MS MARCO Dev Set), and perform better on par with popular dense retrieval models.

<p align="center">
<img align="center" src="./arch.png" width="850" />
Expand Down

0 comments on commit 62aafc6

Please sign in to comment.