The results suggest that development-set performance is insufficient evidence for deployment validity on a shifted distribution, and that point-estimate judges can obscure the heterogeneity in subjective human evaluation.
Abstract
LLM-as-a-Judge evaluators are increasingly used to score open-ended generation, yet a judge's correlation with human ratings on its development set may not guarantee valid measurement when outputs are closely matched and human preferences are subjective. We study this failure mode through psychological depth in short stories. Seven human readers and an LLM-judge ensemble selected on the original scalar Psychological Depth Scale dataset ($\rho = 0.646$) evaluated 60 blinded, prompt-matched story pairs from GPT-5 vs.\ GPT-4o and DeepSeek-R1 vs.\ DeepSeek-V3. Human preferences showed no universal reasoning advantage: GPT-5 was modestly preferred over GPT-4o (60.0--62.9\%), whereas DeepSeek-R1 trailed V3 (42.9\%), and inter-reader agreement was near chance (Krippendorff's $\alpha = 0.070$), with within-reader consistency and recurring weighting patterns suggesting structured heterogeneity rather than random responding. The judge, by contrast, favored reasoning outputs in 89.0\% of dimension-level comparisons and 59 of 60 pairs on aggregate PDS, uniformly across all five evaluator configurations, and its scores were associated with surface features such as sentence length and lexical diversity. These results suggest that development-set performance is insufficient evidence for deployment validity on a shifted distribution, and that point-estimate judges can obscure the heterogeneity in subjective human evaluation.
Researchers increasingly use Large Language Models as judges (LLM-as-a-judge) to evaluate model outputs. Yet there are no standards for how to design these judges. Typically, researchers choose the prompt, rating scale, and model intuitively. If these choices change the judge's verdicts, two studies can reach different...
The results show that high self-consistency does not necessarily indicate high agreement with human judgments when using local LLMs as automatic judges, and highlight the need to evaluate both consistency and human alignment when using local LLMs as automatic judges.
A systematic comparison using a naturally occurring online persuasion corpus in which original posters explicitly verify whether a reply changed their view is conducted, highlighting the risk of treating LLM judgments as faithful proxies for human belief updating and point to structural differences in how LLMs and huma...
The results argue that useful self-reports remain sensitive to elicitation, correlated errors, and benchmark noise, and that useful self-reports remain sensitive to elicitation, correlated errors, and benchmark noise.
L. Meyer, Sofia Rossi, Wei Chen et al.· arXiv.org· 0 citations
Language models are now routinely graded by other language models. If post-training makes a model's own writing more predictable, it may also teach the same model, acting as a judge, to reward predictable writing, so that progress on creativity would be invisible to automated evaluation. We follow two open model famili...
Evaluations that use a large language model (LLM) as a judge have begun to borrow reliability statistics from classical test theory and its extensions. We examine three such statistics that need one administration and no gold labels. None of them can isolate the judge, because one judge under one prompt supplies no var...
L. Zhu· 1 citation
Related blog posts
MIT News · Artificial Intelligence· news.mit.eduAug 27, 2026
A new machine-learning framework aims to improve the success rate of computational protein design while moving away from results that reproduce sequences found in nature.