Overconfidence in LLM-as-a-Judge: Diagnosis and Confidence-Driven Solution
This addresses the need for trustworthy and adaptive evaluation pipelines in AI, though it is incremental as it builds on existing LLM-as-a-Judge methods.
The paper tackles the problem of overconfidence in LLMs used as automated judges, identifying that predicted confidence often exceeds actual correctness, and proposes an ensemble framework called LLM-as-a-Fuser that improves calibration and achieves superior reliability and accuracy compared to existing baselines.
Large Language Models (LLMs) are widely used as automated judges, where practical value depends on both accuracy and trustworthy, risk-aware judgments. Existing approaches predominantly focus on accuracy, overlooking the necessity of well-calibrated confidence, which is vital for adaptive and reliable evaluation pipelines. In this work, we advocate a shift from accuracy-centric evaluation to confidence-driven, risk-aware LLM-as-a-Judge systems, emphasizing the necessity of well-calibrated confidence for trustworthy and adaptive evaluation. We systematically identify the Overconfidence Phenomenon in current LLM-as-a-Judges, where predicted confidence significantly overstates actual correctness, undermining reliability in practical deployment. To quantify this phenomenon, we introduce TH-Score, a novel metric measuring confidence-accuracy alignment. Furthermore, we propose LLM-as-a-Fuser, an ensemble framework that transforms LLMs into reliable, risk-aware evaluators. Extensive experiments demonstrate that our approach substantially improves calibration and enables adaptive, confidence-driven evaluation pipelines, achieving superior reliability and accuracy compared to existing baselines.