DrBenchmark: A Large Language Understanding Evaluation Benchmark for French Biomedical Domain
Yanis Labrak, Adrien Bazoge, Oumaima El Khettari, Mickael Rouvier, Pacome Constant dit Beaufils, Natalia Grabar, Beatrice Daille, Solen Quiniou, Emmanuel Morin, Pierre-Antoine Gourraud, Richard Dufour
TL;DR
DrBenchmark tackles the lack of a standardized French biomedical evaluation by aggregating 20 diverse tasks into a single benchmark. The authors evaluate eight MLMs spanning French generalists, cross-linguals, and French biomedical models to assess cross-language and domain transfer, finding that no model dominates across all tasks, though domain-specific French biomedical models often perform best. The work provides a reproducible, MIT-licensed framework and data release on HuggingFace, enabling fair comparisons and future extensions. This benchmark enables rigorous, language- and domain-specific evaluation of PLMs in biomedicine and points toward leveraging generative/instruction-tuned models for tasks like QA and multi-label classification.
Abstract
The biomedical domain has sparked a significant interest in the field of Natural Language Processing (NLP), which has seen substantial advancements with pre-trained language models (PLMs). However, comparing these models has proven challenging due to variations in evaluation protocols across different models. A fair solution is to aggregate diverse downstream tasks into a benchmark, allowing for the assessment of intrinsic PLMs qualities from various perspectives. Although still limited to few languages, this initiative has been undertaken in the biomedical field, notably English and Chinese. This limitation hampers the evaluation of the latest French biomedical models, as they are either assessed on a minimal number of tasks with non-standardized protocols or evaluated using general downstream tasks. To bridge this research gap and account for the unique sensitivities of French, we present the first-ever publicly available French biomedical language understanding benchmark called DrBenchmark. It encompasses 20 diversified tasks, including named-entity recognition, part-of-speech tagging, question-answering, semantic textual similarity, and classification. We evaluate 8 state-of-the-art pre-trained masked language models (MLMs) on general and biomedical-specific data, as well as English specific MLMs to assess their cross-lingual capabilities. Our experiments reveal that no single model excels across all tasks, while generalist models are sometimes still competitive.