SpringDragon: Text Adventure Dataset for Language Model Training

SpringDragon is a text adventure dataset originally compiled by Henk717 and hosted by ToastyPigeon. The dataset includes a raw text file and a deduplicated JSON version optimized for training frameworks like axolotl. It was last updated on Hugging Face on January 14, 2025.

Use Cases

• Fine-tune language models for text adventure generation based on the interactive narrative format.
• Train models on conversational or branching dialogue structures suggested by the 'text adventure' description.
• Benchmark model performance on story continuation tasks using the provided completion set.
• Generate synthetic text adventures for game development based on the patterns in the original dataset.

Strengths

• Includes a deduplicated JSON version specifically formatted for use with the axolotl training framework.
• Provides both an original raw text file and a processed version, offering flexibility for different preprocessing needs.

Limitations

• Description metadata is limited; actual data quality, size, and content require manual inspection after download.
• Column-level documentation is absent; field semantics and structure must be inferred from the JSON file.
• Row count, file size, and license information are unknown, which may limit suitability assessment.

Provenance

Source

Henk717 (original compiler), ToastyPigeon (host).

Collection Method

Likely scraped or compiled from text adventure game sources.

Freshness

Last updated 2025-01-14 15:57:50; freshness should be verified.