将生成模型幻觉定义为估计与合理原因脱节，揭示最优估计仍会幻觉。
Are Hallucinations Bad Estimations?
- 将幻觉视为估计与潜在原因不匹配的结构性偏差
- 证明任意数据分布下幻觉率存在通用高概率下界
- 适用于对生成结果可信度敏感的场景，如问答与图文生成
我们把生成模型中的幻觉形式化为估计无法关联到任何合理原因的情况。在此解释下，我们证明即使损失最小化的最优估计器仍然会产生幻觉。通过针对一般数据分布的泛化高概率下界，验证了该结论。这将幻觉重新定义为损失最小化与人类可接受输出之间的结构错配，即由校准不当引发的估计误差。在硬币汇总、开放性问答和文本到图像任务上的实验支持了这一理论。
原文摘要 · Abstract (English)
We formalize hallucinations in generative models as failures to link an estimate to any plausible cause. Under this interpretation, we show that even loss-minimizing optimal estimators still hallucinate. We confirm this with a general high probability lower bound on hallucinate rate for generic data distributions. This reframes hallucination as structural misalignment between loss minimization and human-acceptable outputs, and hence estimation errors induced by miscalibration. Experiments on coin aggregation, open-ended QA, and text-to-image support our theory.
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。