Уровень 0 · материалов: 2
В кластер входят документы о методах предотвращения использования ИИ для выполнения работы, но не входят документы об автоматизации создания заданий для упрощения проверки.
Общие признаки: противодействие ИИ-генерации, защита от академического мошенничества, верификация реального выполнения работы, методы проверки авторства
Группа выше: Академическое мошенничество
Смысл: The author describes a method for creating educational assignments in hardware engineering that are resistant to AI cheating. By requiring information that can only be obtained through active simulation, analysis of fragmented repositories, or unreleased literature, the author forces students to actually perform the engineering work rather than prompting an LLM.
An engineering professor shares techniques for creating 'AI-proof' SystemVerilog assignments by hiding critical data in dynamic processes and fragmented codebases.
Смысл: The text discusses a creative way to filter out low-effort AI-generated code contributions in open-source projects by turning an AI-instruction file (AGENTS.md) into a self-disclosure mechanism and requiring physical hardware proof of testing.
A maintainer uses the AGENTS.md standard to trick AI agents into self-identifying in Pull Requests, then requires manual hardware logs to verify human oversight.