JudgmentBench: Comparing Rubric and Preference Evaluation for Quality Assessment
For researchers evaluating LLM outputs in high-expertise domains without ground truth, this work provides empirical evidence that preference-based evaluation is more reliable and efficient than rubric-based scoring.
The paper introduces JudgmentBench, a benchmark of 30 legal tasks with expert annotations, and finds that comparative pairwise judgments recover quality ordering much better than rubric-based scoring (Spearman correlation 0.908 vs. 0.150) while taking less than half the annotation time.
Two methodologies dominate current practices of benchmarking: rubric-based scoring evaluates items against predefined criteria, whereas comparative judgment elicits pairwise preferences between outputs. Although both methodologies are widely used, the choice between them is rarely justified. We release JudgmentBench, a benchmark of 30 real-world legal tasks, paired with 1,539 rubric scores and 1,530 pairwise preference judgments collected from practicing attorneys--including at major U.S. law firms--with substantial experience. The annotations constitute the first publicly available dataset in a high-expertise domain in which both supervision signals are elicited from the same experts on the same items. Using LLM-generated outputs at three constructed quality levels, we provide an initial empirical comparison: comparative judgments recover the intended quality ordering substantially better than rubrics (mean Spearman's rank correlation of 0.908 vs. 0.150, estimated difference = 0.758 [0.494, 1.021]) while requiring less than half the annotation time. The patterns hold for human annotators and LLM autograders. Beyond this initial comparison, the paired structure of the dataset supports a broader research agenda on how expert judgment should be elicited, aggregated, and used as supervision in domains without verifiable ground truth.