MENLO: From Preferences to Proficiency -- Evaluating and Modeling Native-like Quality Across 47 Languages
This addresses the problem of scalable multilingual evaluation for LLM developers, but it is incremental as it builds on existing preference alignment and reinforcement learning techniques.
The paper tackled the challenge of evaluating native-like quality in large language model responses across 47 languages by introducing MENLO, a framework that created a dataset of 6,423 human-annotated preference pairs and showed that fine-tuning with reinforcement learning improved LLM judges, though they still underperformed humans.
Ensuring native-like quality of large language model (LLM) responses across many languages is challenging. To address this, we introduce MENLO, a framework that operationalizes the evaluation of native-like response quality based on audience design-inspired mechanisms. Using MENLO, we create a dataset of 6,423 human-annotated prompt-response preference pairs covering four quality dimensions with high inter-annotator agreement in 47 language varieties. Our evaluation reveals that zero-shot LLM judges benefit significantly from pairwise evaluation and our structured annotation rubrics, yet they still underperform human annotators on our dataset. We demonstrate substantial improvements through fine-tuning with reinforcement learning, reward shaping, and multi-task learning approaches. Additionally, we show that RL-trained judges can serve as generative reward models to enhance LLMs' multilingual proficiency, though discrepancies with human judgment remain. Our findings suggest promising directions for scalable multilingual evaluation and preference alignment. We release our dataset and evaluation framework to support further research in multilingual LLM evaluation.