WildStories and WildEdits Datasets for Iterative Prompting Analysis
September 15, 2026
Researchers introduce WildStories, a dataset of 275,635 story generation prompts, and WildEdits, containing 24,291 edit trees. These datasets model how users iteratively revise prompts to explore narrative branches, addressing the limitations of static, one-shot evaluation benchmarks.
HOW THIS AFFECTS YOU
●
researcherYou can use these datasets to evaluate how models handle iterative, multi-turn creative instructions.
●
designerThis provides insight into how users actually navigate creative workflows with LLMs.