Recent advancements in Large Language Models (LLMs) underscore the necessity of Retrieval Augmented Generation (RAG) to leverage external information. However, LLMs are sensitive to the position of relevant information within contexts and tend to generate incorrect responses when such information is placed in the middle, known as `Lost in the Middle' phenomenon. In this paper, we introduce a framework that generates consistent outputs for decoder-only models, irrespective of the input context order. Experimental results for three open domain question answering tasks demonstrate position invariance, where the model is not sensitive to input context order, and superior robustness to irrelevent passages compared to prevailing approaches for RAG pipelines.
@article{arxiv.2501.07523,
title = {Parallel Key-Value Cache Fusion for Position Invariant RAG},
author = {Philhoon Oh and Jinwoo Shin and James Thorne},
journal= {arXiv preprint arXiv:2501.07523},
year = {2025}
}