The Compliance Paradox: Semantic-Instruction Decoupling in Automated Academic Code Evaluation

Devanshu Sahoo, Manish Prasad, Vasudev Majhi +5

The rapid integration of Large Language Models (LLMs) into educational assessment rests on the unverified assumption that instruction following capability translates directly to ob…