LLM Judges Show Weak Link to General Ability in Writing Evaluation
A study accepted at the NeurIPS 2026 Can We Trust AI Evaluation? Workshop compares LLM judges with human preferences on three long-form writing tasks. Across four text generators, judge quality varied by task and criteria, correlated weakly with MMLU performance at Pearson’s ρ ≤ 0.2, and improved with paired comparisons; the strongest LLMs did not consistently reproduce human judgments.
The results caution teams against choosing LLM judges solely by general model strength, while supporting paired comparison as the more consistent evaluation format in the studied setup.
content type paperpublished November 2026
Evaluating the Evaluators: Investigating LLM Judges for Personalized Writing Style Assessment
AuthorsNatalie Mackraz‡, Andrew Silva, Aini Putkonen, Brihi Joshi†**, Joelle Alcaidinho, Arnav Arora§**, Barry-John Theobald, Katherine Metcalf‡
This paper was accepted at the Can We Trust AI Evaluation? Workshop at NeurIPS 2026.
We evaluate how good large language model (LLM) judges are for personalized writing tasks. This is important when a user’s preferred writing style is known (e.g., “upbeat”) and an LLM judge is used to evaluate whether generated text adheres to this preference. We examine how the writing task, the combination of generator and judge LLM, the evaluation objective, and general commonsense and reasoning ability impact LLM judge performance. To this end, we generate text with four LLMs for three long-form writing tasks, collect human labels evaluating which of two responses is better, collect judge-LLM, pairwise-ranking annotations, and then compare the human and judge LLM annotations. We find that judge quality correlates weakly with general LLM ability measured using MMLU (Pearson’s ρ ≤ 0.2) and varies by both writing task and evaluation criteria (i.e., general helpfulness versus style personalization). We also find that LLM evaluators are more consistent and performant using paired comparisons rather than scoring responses independently. Finally, we find that the “strongest” LLMs do not most reliably reproduce the human evaluations, which we hypothesize relates to exaggerated sensitivity to subtle details that human annotators and weaker LLMs ignore.