Building domain-agnostic, agentic AI assessments for conceptual mastery through error detection in generated outputs using adversarial methods. Initial applications target recruiting and education. The framework also serves as an RL environment for LLMs, multimodal models, and computer-use agents....See more