Mapping the Evaluation Frontier: An Empirical Survey of the Bias-Reliability Tradeoff Across Eleven Evaluator-Agent Conditions
2026-07-01 • Machine Learning
Machine LearningArtificial IntelligenceComputation and Language
AI summaryⓘ
The authors tested a theory that when evaluating large language models (LLMs), you can't have low evaluator agreement, diverse evaluation methods, and reliable measurements all at once with a small number of samples. They looked at 11 different evaluation setups and confirmed that if evaluators agree a lot, the results are stable but less diverse, and if they don't agree much, the results are noisy. They also found an unusual case with GPT-4o that showed perfect diversity and no agreement, likely due to a version change. Their work provides a dataset to help compare different evaluation methods.
Large Language Models (LLMs)Evaluator CouplingStrategy DiversityMeasurement ReliabilityBias-Reliability TradeoffCoefficient of VariationSample SizeBenchmark DatasetGPT-4oEvaluation Metrics
Authors
Zewen Liu
Abstract
The bias-reliability tradeoff conjectures that LLM evaluation systems are constrained in (gamma, H, CV) space, where evaluator coupling (gamma), strategy diversity (H), and small-sample measurement reliability (CV(N)) cannot be simultaneously optimized at fixed sample size N. Prior evidence rests on n=5 conditions with complete metrics from a single study. We expand the empirical base to 11 conditions, measuring gamma and H for all 11 (nine with valid weight vectors) and CV(N=5) for seven with sufficient seeds (N >= 5). Five conditions provide the complete (gamma, H, CV) triple. The data confirm the trade-off: conditions with low evaluator coupling (gamma < 0.2) exhibit high measurement noise (CV(N=5) > 1.0), while conditions with strong coupling (gamma > 0.9) achieve low noise (CV(N=5) < 0.16). The correlation r(H, gamma) = -0.989 (n=5, excluding GPT-4o conditions) confirms that evaluator coupling suppresses strategy diversity. Four GPT-4o conditions show gamma=0.000 and H=1.000 across all seeds -- a pattern we attribute to version drift in the June 2026 GPT-4o API. No condition occupies the region {gamma < 0.2, CV(N=5) < 0.3}. We release all per-condition metrics as a standardized benchmark dataset for evaluator comparison.