Tag
This paper demonstrates that rubric text alone can predict LLM judge outputs, challenging the assumption of rubric-based evaluation and raising concerns about its reliability in automated text generation assessment.