向内看，为何大语言模型会产生幻觉：一种因果视角
Look Within, Why LLMs Hallucinate: A Causal Perspective
AI总结：
本文从因果视角干预大语言模型的自注意力层，发现禁用前端或尾部特定层可缓解幻觉问题，为理解与缓解幻觉开辟新途径。
摘要：
大语言模型（LLMs）的出现是生成式人工智能的一个里程碑，在文本理解和生成任务中取得了显著成功。尽管LLMs在许多下游任务中取得了巨大成功，但它们仍存在严重的幻觉问题，这对LLMs的实际应用构成了重大挑战。大多数关于LLMs幻觉的研究工作都集中在数据质量上。自注意力是基于Transformer的LLMs中的核心模块，然而其与LLMs幻觉的潜在关系却鲜有研究。为填补这一空白，我们从因果视角研究该问题。我们提出一种干预LLMs自注意力层的方法，并保持其结构和大小不变。具体而言，我们在多个流行的开源LLMs中禁用不同的自注意力层，然后将其幻觉程度与原始模型进行比较。我们在幻觉评估基准上评估了被干预的LLMs，并得出结论：禁用LLMs前端或尾部的某些特定自注意力层可以缓解幻觉问题。该研究为理解和缓解LLMs的幻觉开辟了新途径。
英文摘要：
The emergence of large language models (LLMs) is a milestone in generative artificial intelligence, achieving significant success in text comprehension and generation tasks. Despite the tremendous success of LLMs in many downstream tasks, they suffer from severe hallucination problems, posing significant challenges to the practical applications of LLMs. Most of the works about LLMs' hallucinations focus on data quality. Self-attention is a core module in transformer-based LLMs, while its potential relationship with LLMs' hallucination has been hardly investigated. To fill this gap, we study this problem from a causal perspective. We propose a method to intervene in LLMs' self-attention layers and maintain their structures and sizes intact. Specifically, we disable different self-attention layers in several popular open-source LLMs and then compare their degrees of hallucination with the original ones. We evaluate the intervened LLMs on hallucination assessment benchmarks and conclude that disabling some specific self-attention layers in the front or tail of the LLMs can alleviate hallucination issues. The study paves a new way for understanding and mitigating LLMs' hallucinations.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
