幻觉盆地：一种动态框架用于理解和控制LLM幻觉
Hallucination Basins: A Dynamic Framework for Understanding and Controlling LLM Hallucinations
AI总结：
本文提出几何动态系统框架，通过分析潜在空间中的盆地结构来理解LLM幻觉现象，发现任务依赖性显著影响分离性，并通过任务复杂度和多盆地定理展示几何感知引导可降低幻觉概率。
摘要：
大型语言模型（LLMs）会产生流畅但事实错误的输出，即幻觉。本文提出一个几何动态系统框架，其中幻觉源于潜在空间中任务依赖的盆地结构。通过使用自回归隐藏状态轨迹跨多个开源模型和基准，我们发现分离性强烈依赖于任务而非普遍存在：事实性设置可以显示更清晰的盆地分离，而总结和误解密集的设置通常不稳且常重叠。我们通过任务复杂度和多盆地定理正式化这种行为，表征L层变压器中的盆地出现，并展示几何感知引导可减少幻觉概率而无需重新训练。
英文摘要：
Large language models (LLMs) hallucinate: they produce fluent outputs that are factually incorrect. We present a geometric dynamical systems framework in which hallucinations arise from task-dependent basin structure in latent space. Using autoregressive hidden-state trajectories across multiple open-source models and benchmarks, we find that separability is strongly task-dependent rather than universal: factoid settings can show clearer basin separation, whereas summarization and misconception-heavy settings are typically less stable and often overlap. We formalize this behavior with task-complexity and multi-basin theorems, characterize basin emergence in L-layer transformers, and show that geometry-aware steering can reduce hallucination probability without retraining.
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。
