Synthetic Japanese Conversations for Instruction Tuning, 10K Records

Approximately 10,000 synthetic Japanese conversation records generated for instruction tuning of large language models. Author Aratako created the dataset by applying the Magpie method to the NVIDIA Nemotron-4-340B-Instruct model via DeepInfra and published the generation code on July 5, 2024.

Use Cases

• Fine-tuning instruction-following LLMs based on synthetic Japanese conversation data.
• Benchmarking model performance on Japanese conversational tasks.
• Augmenting training data for multilingual chat applications.
• Studying the characteristics of model-generated synthetic data.

Strengths

• Contains approximately 10,000 records, providing a substantial volume for training.
• The generation code is publicly available, enabling reproducibility and modification.
• Specifically targets the Japanese language, filling a niche for non-English instruction data.

Limitations

• The author notes no post-generation filtering was applied, so low-quality records may be present.
• Column-level documentation is absent; the structure and fields of the data must be inferred after download.
• The dataset's license is unknown, which may restrict commercial or redistribution use.

Provenance

Source

huggingface user Aratako

Collection Method

Synthetically generated by applying the Magpie method to the NVIDIA Nemotron-4-340B-Instruct model using DeepInfra.

Freshness

Last updated 2024-07-05 13:57:08.