Pageshift-Entertainment has announced a major update to its LongPage dataset, a valuable resource for those developing large language models (LLMs) with the goal of generating complex narrative content. ## Details of the LongPage Dataset The LongPage dataset stands out for the inclusion of "reasoning traces" associated with each novel. These traces offer a hierarchical breakdown of the plot, starting from a high-level idea and arriving at the detailed structure in chapters and scenes. This approach facilitates the training of LLMs capable of handling the complexity of writing an entire book. The new version of the dataset significantly expands its scope, growing from approximately 300 to over 6,000 novels. Pageshift-Entertainment is currently training a full-book writing model using LongPage and plans to release it as soon as the output quality reaches an acceptable level. ## Implications for Content Generation The availability of datasets like LongPage represents a significant step forward in the field of automatic content generation. The ability to train LLMs on a vast range of novels, together with reasoning traces, could lead to models capable of producing more coherent, complex, and engaging stories. This opens new perspectives for the entertainment industry, publishing, and content creation in general.