Tag
The paper introduces a controlled benchmark of French surface variants (apostrophe omission, diacritic removal, keyboard typos) across seven tasks with 12,600 inputs, showing that accuracy drops and Brier score/NLL rise with perturbation intensity, while calibration combined with input restoration improves reliability.
This paper introduces FrenchNews-7, a benchmark for classifying French news editorial desks across multiple publishers, featuring a fine-tuned CamemBERT classifier and evaluations against LLM baselines.