In the era of pretrained and large language models, high quality datasets are still of high relevance. Indeed, there is a current trend to promote (new) datasets for pre-training models, e.g., by curating previous corpora or adding new resources (Li et al., 2024), or building new corpora to critically evaluate large language models (Röttger et al., 2024). In the case of TS as a downstream task of natural language generation, most often already pre-trained models are fine-tuned for automatic simplification. Therefore, smaller but parallel text pairs are required to fine-tune as well as to judge the quality of TS systems on unseen data.