Just Ask Jev: Reinforcement Learning for Calibrated Decisions as a Zero-Shot Detector of AI Alignment Failures ▲ 5 💬 2 2026-09-27 arXiv HF 原文 GitHub ★14 摘要 （无摘要）