Qwen3-4B Instruct 100K Backup: Instruction-Tuning Dataset

TY233 hosts a backup copy of the Qwen3-4B instruction dataset containing 99,987 examples. The dataset is provided in JSONL and Parquet formats to support collaboration and reproducibility. It was last updated on June 10, 2026.

Use Cases

• Fine-tune language models for instruction-following based on the structured conversation data.
• Benchmark model performance on instruction-response tasks using the provided examples.
• Reproduce training results for the Qwen3-4B model architecture.
• Analyze the structure and content of instruction-tuning datasets for research.

Strengths

• Contains 99,987 instruction-response examples.
• Provides data in both JSONL and Parquet formats for flexibility.
• Explicitly shared for reproducibility purposes.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

TY233

Collection Method

Backup mirror of dataset artifacts; original creation method is unspecified.

Freshness

Last updated 2026-06-10 22:34:47; freshness should be verified.