Crimson Hexagonal Archive — Hugging Face Dataset Work Plan v3 is an implementation plan for turning the archive into a research corpus.
The proposed dataset retains one row per deposit, a chunked training view, thematic subsets, and views grouped by classifier-attributed heteronym. Each record preserves original creator metadata while adding a machine-generated attribution proposal and confidence signals.
The classifier is also intended to distinguish human-primary, human-directed AI-assisted, mixed, AI-directed, and provenance-anchored generated material. Multiple text renderings would test whether provenance affects downstream training behavior when it is visible to a model rather than stored only as metadata.
Version 3 is a plan, not the completed dataset. Automated reattribution is analytical metadata until a separate human review accepts any canonical correction. The original deposit remains the primary source.