大模型幻觉并非缺知识，而是答得太分散。
Hallucination as Commitment Failure: Larger LLMs Misfire Despite Knowing the Answer
- 用语义概念替代字面词，判断答案是否可用
- 47%幻觉发生时正确概念已有高概率，且随模型变大而上升
- 幻觉是概率分散所致，而非缺乏知识，适合研究模型决策机制者看
幻觉常被理解为知识缺失的结果：当正确答案不在生成时的分布中，模型就答错；反之则答对。我们引入一种语义层面的答案可用性概念，将表达相同答案概念的不同词元变体聚合，考察模型在做出回答前，正确概念是否已存在。在从0.8B到72B的Qwen与Llama系列模型（包含Instruct和Base版本）中，16%至47%的Instruct幻觉发生时，正确概念已有显著的概率质量，且该比例随模型规模单调上升。对比具有相同语义支持的正确生成，关键差异不在于正确概念是否存在，而在于其概率分布：正确生成将概率集中于单一表面形式，而幻觉则将其分散于多个替代形式。这一概率集中度的不对称性在多词元生成中持续存在，并可在生成前的隐藏状态中检测到。这些结果共同指向一个单一机制：指令微调随着模型规模增大，强化了回答承诺，使帮助性输出与自信幻觉成为同一内在倾向的两种表现。
原文摘要 · Abstract (English)
Hallucination is often viewed as a direct consequence of missing knowledge: a model answers incorrectly when the correct answer is absent from its generation-time distribution, and correctly when it is present. We test this assumption by introducing a semantic notion of answer availability that aggregates token-level variants expressing the same answer concept, and asks whether the correct concept is already available at the moment the model commits to an answer. Across Qwen and Llama models from 0.8B to 72B in both Instruct and Base variants, 16-47% of Instruct hallucinations occur with substantial probability mass already on the correct concept, and the rate rises monotonically with scale. Comparing such failures against correct generations with matched semantic support, the distinguishing factor is not whether the correct concept is represented, but how its probability is distributed: correct generations concentrate mass on a single surface form, hallucinations disperse it across alternatives. The same sharpening asymmetry extends across multi-token generation and is detectable in pre-generation hidden states. Together, these results identify a single mechanism: instruction tuning sharpens answer commitment with scale, making helpfulness and confident hallucination two consequences of the same underlying disposition.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。