Compliance without coherence: fluent failure and the ethics of alignment evaluation in multi-agent language models

作者：Pantaleon Fassbender · 发表于：AI and Ethics · 年份：2026 · DOI：10.1007/s43681-026-01252-7