Headlines You Won't Forget: Can Pronoun Insertion Increase Memorability?

For unrestricted data access, please go to the newest version of this dataset. This is the data repository for the CMCL 2026 paper "Headlines You Won't Forget: Can Pronoun Insertion Increase Memorability?" The zip-file "LLM revisions and Annotations" contains three files: annotation_samples.json and annotation_samples_2.json contain all LLM-revisions.annotations_of_LLM_revisions.json contains the non-aggregated annotation data for all LLM-revisions. Annotations for accuracy_subclass are ordered as follows: [misrepresentation, omission, addition]. Annotations for style subclass are ordered as [grammar, awkward style, inconsistent]. In the zip-file "Memory Experiment Data", we provide non-aggregated results of the memory experiment (experiment_data.json), as well as the recalled items matched to their respective headline id's (headline_items_matched). headlines.json provides information about the revision type, and group assignment of headline items. We further provide aggregated results in csv-files. Files for each study are provided in separate sub-directories. If you use this dataset in your research please cite the paper.