← back to paper
arxiv: 2605.15416 · 2 revisions
Margin-Adaptive Confidence Ranking for Reliable LLM Judgement