When large language models are aligned via supervised fine-tuning, they may encounter new factual information that was not acquired through pre-training. It is often conjectured that this can teach the model the behavior of hallucinating factually incorrect responses, as the model is trained to generate facts that are not grounded in its pre-existing knowledge. In this work, we study the impact of such exposure to new knowledge on the capability of the fine-tuned model to utilize its pre-existing knowledge. To this end, we design a controlled setup, focused on closed-book QA, where we vary the proportion of the fine-tuning examples that introduce new knowledge. We demonstrate that large language models struggle to acquire new factual knowledge through fine-tuning, as fine-tuning examples that introduce new knowledge are learned significantly slower than those consistent with the model's knowledge. However, we also find that as the examples with new knowledge are eventually learned, they linearly increase the model's tendency to hallucinate. Taken together, our results highlight the risk in introducing new factual knowledge through fine-tuning, and support the view that large language models mostly acquire factual knowledge through pre-training, whereas fine-tuning teaches them to use it more efficiently.

翻译：当大型语言模型通过监督式微调进行对齐时，它们可能会遇到预训练阶段未获取的新事实信息。通常推测，这会教会模型产生事实错误回应的幻觉行为，因为模型被训练生成并非基于其已有知识的真实信息。本研究探讨了这种对新知识的暴露对微调模型运用其已有知识能力的影响。为此，我们设计了一个受控实验场景，聚焦闭卷问答任务，通过改变引入新知识的微调样本比例，发现大语言模型难以通过微调获取新的事实知识——引入新知识的微调样本学习速度显著低于与模型已有知识一致的样本。然而，我们也发现随着包含新知识的样本最终被学习，它们会线性增加模型的幻觉倾向。综合而言，我们的结果凸显了通过微调引入新事实知识的风险，并支持以下观点：大语言模型主要通过预训练获取事实知识，而微调则教导模型更高效地运用这些知识。