正在加载学习内容…
Autonomous AI Agents Systematically Diverge From Human Preferences in Proof Formalization · AIScope