dataset February 18, 2026

Chinese-Fineweb-Edu-V2.2 Dataset: Quickstart Guide

opencsg/Fineweb-Edu-Chinese-V2.2 ↗

The Chinese-Fineweb-Edu-V2.2 dataset provides high‑quality Chinese educational text for large‑language‑model pre‑training and instruction fine‑tuning. It is organized into three tiers of pre‑training data (Score 4‑5, 3‑4, 2‑3) and two SFT formats (pure QA and full‑context), all readily loadable via Hugging Face’s `datasets` library. The collection is licensed under the OpenCSG Community License, permitting commercial use after approval.

Project Ideas

  1. Use the high‑score (4‑5) pre‑training data to improve the factual accuracy of Chinese LLMs in STEM domains.
  2. Fine‑tune a LLaMA‑3 model on the pure QA SFT set to create a specialized Chinese tutoring assistant.
  3. Leverage the full‑context SFT data for research on retrieval‑augmented generation and context‑aware reasoning.
  4. Combine Tier 2 and Tier 3 pre‑training data to enhance multilingual code‑switching capabilities in conversational agents.
  5. Develop curriculum‑aligned evaluation benchmarks using the dataset’s original source references for education‑focused LLM assessment.
← Back to all reports