arXiv AI

How to Run Statistics over LLM Judges and Trust the Results: Calibrated Inference for Small-Sample AI Evaluation with evalstats

arXiv Machine Learning
Sep 14

Can We Trust LLM Judges: A Study of Capability-Dependent Biases and Multi-Judge Ensemble for Bias Calibration

The paper investigates how large language models (LLMs) used as judges in absolute scoring tasks exhibit systematic biases that compromise reliability. It shows that a judge’s task accuracy strongly predicts both its judging accuracy and its directional bias, yet more capable examinee models consistently receive more lenient judgments. To mitigate these biases, the authors propose a calibrated weighted majority voting (WMV) ensemble that estimates judges’ error rates from inter-judge agreement patterns, achieving near-oracle performance without labeled data and improving both accuracy and fairness.

By Gemma Zhang, Prachi Badarayani, Asmi Kumar, Sadid Hasan, Sulaiman Vesal
arXiv Computation and Language
Sep 11

Rethinking Verbalized Confidence for LLM-as-a-Judge: A Compatibility Shift on Post-2025 Proprietary Models

The paper argues that verbalized confidence—once viewed as overconfident and coarse—has become the preferred soft‑scoring method for LLM‑as‑a‑Judge on top‑tier proprietary models released after 2025. Experiments on SummEval, AggreFact, and HelpSteer2 across up to 18 LLMs show that log‑probabilities are no longer the best signal, and that adding an overconfidence advisory and self‑debate further improves calibration and robustness. The authors note that these enhancements incur little accuracy loss on post‑2025 models but do affect pre‑2025 ones, highlighting a compatibility shift in how confidence should be measured.

By Yu-Chung Hsiao
arXiv Computation and Language
Oct 22

SANE: Automated Scorer-Specific Metamorphic Testing for LLM Judges

SANE is an agent‑based framework that automatically generates scorer‑specific perturbation recipes and expected score relations for large‑language‑model judges, using behavioral requirements encoded as reusable agent instructions. It validates each recipe against the scorer’s intent and the instantiated test case, producing a growing inventory of aligned recipes as more scorers are added. In experiments across five domains, SANE yields 3.7× more validated test families than the Judge Reliability Harness, detects deficiencies in all domains—including professionalism and toxicity—and achieves strong human alignment in judge ranking on faithfulness (Spearman ρ = 0.852).

By Ethan Ludwick, Yu-Chung Hsiao
arXiv Machine Learning
6d ago

Prediction-Powered LLM Ranking with Adaptively Collected Human Feedback

The paper introduces Prediction-Powered Ranking (PP‑Rank), a framework that blends large language model (LLM) judge predictions with limited human feedback to rank LLMs more precisely while maintaining valid inference under adaptive sampling. PP‑Rank uses data‑driven prediction weights to minimize asymptotic covariance, establishes asymptotic normality for general evaluation targets, and provides a rule for reducing exploration that balances cost and variance. Experiments on UltraFeedback‑Extended and synthetic data show that PP‑Rank improves score estimation, narrows confidence intervals, certifies rankings, and can cut evaluation costs without losing precision.

By Gabriel Sargent, Will Wei Sun, Zhengwu Zhang, Yufeng Liu
Hugging Face Trending Papers
Aug 18

Judge, Retrieve, or Abstain: Uncertainty-Guarded LLM Judging with Provable Risk Guarantees

The paper introduces a risk‑controlled framework for using large language models (LLMs) as judges in tasks without reference answers. By calibrating uncertainty thresholds on a held‑out set, the method ensures that the false discovery rate of accepted verdicts stays below a user‑specified level α with high probability, using finite‑sample Clopper–Pearson intervals. When the parametric judge lacks confidence, the instance is routed to a retrieval‑augmented mode with a second calibrated threshold, preserving the error guarantee while achieving higher coverage than single‑mode baselines.