Optimizing Retrieval-augmented Reader Models via Token Elimination

TL;DR

This paper introduces a token-level elimination method for retrieval-augmented reader models that significantly reduces decoding time with minimal performance loss, and sometimes even enhances results.

Contribution

It proposes a novel token elimination technique that optimizes retrieval-augmented models by removing non-essential information during decoding.

Findings

Run-time reduced by up to 62.2%.

Performance drops by only 2% with the method.

In some cases, performance improves.

Abstract

Fusion-in-Decoder (FiD) is an effective retrieval-augmented language model applied across a variety of open-domain tasks, such as question answering, fact checking, etc. In FiD, supporting passages are first retrieved and then processed using a generative model (Reader), which can cause a significant bottleneck in decoding time, particularly with long outputs. In this work, we analyze the contribution and necessity of all the retrieved passages to the performance of reader models, and propose eliminating some of the retrieved information, at the token level, that might not contribute essential information to the answer generation process. We demonstrate that our method can reduce run-time by up to 62.2%, with only a 2% reduction in performance, and in some cases, even improve the performance results.