Tag
This paper introduces a dimension-level evaluation method for measuring intent fidelity in large language models using structured prompt ablation.