English

MILE-RefHumEval: A Reference-Free, Multi-Independent LLM Framework for Human-Aligned Evaluation

Computation and Language 2026-02-11 v1

Abstract

We introduce MILE-RefHumEval, a reference-free framework for evaluating Large Language Models (LLMs) without ground-truth annotations or evaluator coordination. It leverages an ensemble of independently prompted evaluators guided by a human-aligned schema, supporting both discrete and continuous scoring judgement. With task-specific prompts from best candidate selection, summarization and image captioning to dialogue, MILE-RefHumEval provides flexible, interpretable, and scalable assessments. Experiments show it aligns closely with human judgments, outperforms prior methods, and reduces computational overhead, offering an efficient, robust, and human-aligned solution for real-world LLM evaluation.

Keywords

Cite

@article{arxiv.2602.09624,
  title  = {MILE-RefHumEval: A Reference-Free, Multi-Independent LLM Framework for Human-Aligned Evaluation},
  author = {Nalin Srun and Parisa Rastin and Guénaël Cabanes and Lydia Boudjeloud Assala},
  journal= {arXiv preprint arXiv:2602.09624},
  year   = {2026}
}
R2 v1 2026-07-01T10:29:28.955Z