Optimizing Retrieval-augmented Reader Models via Token Elimination
This addresses efficiency issues for users of retrieval-augmented language models in open-domain tasks like question answering, representing an incremental optimization.
The paper tackles the decoding time bottleneck in Fusion-in-Decoder models by analyzing the necessity of retrieved passages and proposing token elimination to reduce non-essential information. The method reduces run-time by up to 62.2% with only a 2% performance reduction, sometimes even improving results.
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.