Latent Paraphrasing: Perturbation on Layers Improves Knowledge Injection in Language Models

Autor:	Kang, Minki, Hwang, Sung Ju, Lee, Gibbeum, Cho, Jaewoong
Rok vydání:	2024
Předmět:	Computer Science - Computation and Language Computer Science - Artificial Intelligence
Druh dokumentu:	Working Paper
Popis:	As Large Language Models (LLMs) are increasingly deployed in specialized domains with continuously evolving knowledge, the need for timely and precise knowledge injection has become essential. Fine-tuning with paraphrased data is a common approach to enhance knowledge injection, yet it faces two significant challenges: high computational costs due to repetitive external model usage and limited sample diversity. To this end, we introduce LaPael, a latent-level paraphrasing method that applies input-dependent noise to early LLM layers. This approach enables diverse and semantically consistent augmentations directly within the model. Furthermore, it eliminates the recurring costs of paraphrase generation for each knowledge update. Our extensive experiments on question-answering benchmarks demonstrate that LaPael improves knowledge injection over standard fine-tuning and existing noise-based approaches. Additionally, combining LaPael with data-level paraphrasing further enhances performance. Comment: NeurIPS 2024
Databáze:	arXiv
Externí odkaz:	http://arxiv.org/abs/2411.00686 Zobrazit plný text záznamu View this record from Arxiv