04:00
2026-08-05
arxiv.org
large-language-models
Knowing the Form, Not the Function: Automatically Auditing Answer--Authority Decoupling in Legal Benchmarks
A study from arXiv (2608.02621v1) found that four large language models (LLMs) spontaneously cited legal authority in 238 Taiwan bar-examination items even when not prompted, but answer correctness an…