LongPage

Summary

Current language models struggle with long-form creative writing because they lack hierarchical planning capabilities. LongPage provides **explicit reasoning traces** that show models how to think through character development, plot progression, and thematic coherence at scale—the **Chain of Thought for creative writing** that the field has been missing.

Keywords

hf-dataset language-modeling · -reasoning language-modeling text2text-generation machine-generated found monolingual original parquet optimized-parquet text datasets dask polars mlcroissant long-context cot reasoning creative-writing cold start reasoning data

Topics

Language Modeling, Reasoning

Research notes

• downloads=918; likes=149