Large multilingual language models can effectively represent orthographic and lexical variation across Occitan dialects without the need for extensive data normalization.
