[论文解读] Augmenting LLMs with Knowledge: A survey on hallucination prevention

Κωνσταντίνος Ανδριόπουλος, Johan Pouwelse|arXiv (Cornell University)|Sep 28, 2023

Topic Modeling被引用 9

一句话总结

简介知识增强的语言模型，获取并融合外部知识以减少幻觉，涵盖检索、融合与基于检索的增强方法。

ABSTRACT

Large pre-trained language models have demonstrated their proficiency in storing factual knowledge within their parameters and achieving remarkable results when fine-tuned for downstream natural language processing tasks. Nonetheless, their capacity to access and manipulate knowledge with precision remains constrained, resulting in performance disparities on knowledge-intensive tasks when compared to task-specific architectures. Additionally, the challenges of providing provenance for model decisions and maintaining up-to-date world knowledge persist as open research frontiers. To address these limitations, the integration of pre-trained models with differentiable access mechanisms to explicit non-parametric memory emerges as a promising solution. This survey delves into the realm of language models (LMs) augmented with the ability to tap into external knowledge sources, including external knowledge bases and search engines. While adhering to the standard objective of predicting missing tokens, these augmented LMs leverage diverse, possibly non-parametric external modules to augment their contextual processing capabilities, departing from the conventional language modeling paradigm. Through an exploration of current advancements in augmenting large language models with knowledge, this work concludes that this emerging research direction holds the potential to address prevalent issues in traditional LMs, such as hallucinations, un-grounded responses, and scalability challenges.

研究动机与目标

• 动机：推动将外部知识整合用于解决LLMs中的幻觉与知识局限。
• 调研并对LLMs的外部知识增强技术进行分类，包括非结构化与结构化来源。
• 解释实现检索、记忆增强及将检索内容与生成融合的核心架构。
• 强调知识增强LLMs在更新记忆、效率和安全性方面的挑战。

提出的方法

• 描述并比较检索增强生成（RAG）与解码器内融合（FiD）架构。
• 解释 REALM 风格的联合检索器-生成器预训练及其训练动力学。
• 详细说明 Atlas、RETRO、GRAFT-Net 与 PullNet 作为结构化知识整合和多跳问答的示例。
• 讨论搜索引擎增强生成与 SeeKeR 作为利用实时网页信息的方法。
• 总结基于图的和三元数据库方法（GCN/Relational GCN）用于结构化知识图谱。

Figure 1: Overview of knowledge augmentation of language models from the paper by Izacard et al. [ 7 ] . The input query (light yellow), along with a number of retrieved relevant documents (light blue), passes through the generative seq2seq model to produce an output response.

实验结果

研究问题

• RQ1 将LLMs与外部知识增强的主要架构范式有哪些？
• RQ2 检索与融合机制如何影响生成内容的真实性与基础性？
• RQ3 LLM中参数化知识存储与非参数化外部记忆之间的权衡是什么？
• RQ4 结构化知识图、三元组存储与文本语料如何促进多跳推理与问答？
• RQ5 对知识增强的LLMs在更新、可扩展性、安全性与来源可追溯性方面存在哪些开放挑战？

主要发现

• 指出从非结构化文本到结构化图的知识增强策略谱系。
• 显示检索增强模型可以减少对参数化记忆的依赖并改善基础性。
• 强调共同训练检索器与生成器以提升检索质量与最终任务性能的方法。
• 讨论将检索内容整合的不同融合策略（RAG 风格、FiD、FiD 类串联）的优点与局限。
• 指出在使用如搜索引擎的外部知识源时，扩展性、实时更新与安全性的重要性。

Figure 2: Overview of the Fusion-in-Decoder (FiD) [ 7 ] technique. The input question gets concatenated with each relevant passage and all concatenations get encoded in parallel. The embeddings that are produced are concatenated together (fusion) and are passed as input to the decoder.