Datasets:
GPT-5.6 Romance Multi-Turn
Private English romance roleplay training data generated by GPT-5.6 and independently validated and rechecked before approval.
Contents
conversations.jsonl: 4,000 approved assistant-prefix targets.manifest.json: source, expansion, review thresholds, row counts, and SHA-256 checksums.- 500 source conversations expanded into 8 targets each.
- 47 concise reasoning targets and 3,953 direct-prose targets.
Data contract
The examples are English, adult (21+), consensual, non-graphic romance roleplay. Characters use common English names, and the assistant is instructed to preserve the user's agency rather than write the user's actions or dialogue.
The approved artifact was checked for formatting, row counts, assistant-prefix expansion, grammar, literary quality, continuity, agency, and adult-content constraints. The private owned-transcript supplement is intentionally excluded from this repository.
This dataset is intended for the owner's private model-training workflow. Review the source and applicable licenses before sharing or publishing it.
Artifact SHA-256: 9c01cfee6f56332e50d420fce7955ea787b6973c26f2a709a40ad6f641e1e8f1
- Downloads last month
- 5