LLMs as Repositories of Factual Knowledge: Limitations and Solutions
This addresses the problem of factual inaccuracies and inconsistencies in LLMs for users relying on them for up-to-date information, though it is incremental as it builds on existing fine-tuning approaches.
The study assessed the reliability of 24 state-of-the-art LLMs as repositories of factual knowledge, finding limitations in accuracy and consistency when handling time-sensitive questions, and proposed ENAF, a neurosymbolic fine-tuning method that improved performance.
LLMs' sources of knowledge are data snapshots containing factual information about entities collected at different timestamps and from different media types (e.g. wikis, social media, etc.). Such unstructured knowledge is subject to change due to updates through time from past to present. Equally important are the inconsistencies and inaccuracies occurring in different information sources. Consequently, the model's knowledge about an entity may be perturbed while training over the sequence of snapshots or at inference time, resulting in inconsistent and inaccurate model performance. In this work, we study the appropriateness of Large Language Models (LLMs) as repositories of factual knowledge. We consider twenty-four state-of-the-art LLMs that are either closed-, partially (weights), or fully (weight and training data) open-source. We evaluate their reliability in responding to time-sensitive factual questions in terms of accuracy and consistency when prompts are perturbed. We further evaluate the effectiveness of state-of-the-art methods to improve LLMs' accuracy and consistency. We then propose "ENtity-Aware Fine-tuning" (ENAF), a soft neurosymbolic approach aimed at providing a structured representation of entities during fine-tuning to improve the model's performance.