carbonz0/alpaca-chinese-dataset

Dataset comprises instruction-output pairs generated through machine translation and self-instruct methodology, formatted identically to the original Alpaca dataset's JSON structure for seamless compatibility with existing LLM fine-tuning pipelines. Data generation combines automated translation of English instructions with self-bootstrapping techniques, though cleaning methodologies and keyword filtering rules remain in development. Targets Chinese language model instruction tuning, enabling adaptation of instruction-following capabilities to Mandarin without requiring manual annotation at scale.

397 stars. No commits in the last 6 months.