cs.AI, cs.CL, cs.SY, eess.SY

Hallucination Basins: A Dynamic Framework for Understanding and Controlling LLM Hallucinations

arXiv:2604.04743v1 Announce Type: new
Abstract: 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 ba…