BTR: Binary Token Representations for Efficient Retrieval Augmented Language Models

arXiv:2310.01329v2 Announce Type: replace
Abstract: Retrieval augmentation addresses many critical problems in large language models such as hallucination, staleness, and privacy leaks. However, running retrieval-augmented language models (LMs) is slow and difficult to scale due to processing large amounts of retrieved text. We introduce binary token representations (BTR), which use 1-bit vectors to precompute every token in passages, significantly reducing computation during inference. Despite the potential loss of accuracy, our new calibration techniques and training objectives restore performance. Combined with …

abstract arxiv augmentation binary cs.ai cs.cl hallucination however language language models large language large language models leaks lms privacy processing retrieval retrieval-augmented running scale text token type