High School Chemistry Question-Answer Pairs for Model Fine-Tuning

August 2025 dataset compiled by liushuaiqian for fine-tuning domain-specific large language models in high school chemistry. It contains structured Chinese question-answer pairs derived from mainstream textbooks, common exam equations, and popular study guides. The data is formatted as JSONL for direct use with machine learning libraries.

Use Cases

• Fine-tune a text generation model to produce correct answers from 'instruction' prompts containing chemistry problems.
• Train a question-answering model using the paired 'instruction' and 'output' fields as supervised examples.
• Evaluate model performance on high school chemistry concepts by using the 'output' field as a ground truth reference.
• Create a retrieval-augmented generation system for chemistry tutoring by indexing the 'instruction' and 'output' pairs.

Strengths

• Data sourced from three mainstream textbook editions and popular reference books, providing structured domain coverage.
• Format is JSONL, designed for direct compatibility with Hugging Face datasets library for model fine-tuning.

Limitations

• Dataset size, row count, and sample balance are unknown, making it difficult to assess scale for training.
• The author notes potential format errors in some JSONL rows, requiring manual script-based correction before use.

Provenance

Source

Mainstream Chinese high school textbooks (Su Jiao, Ren Jiao, Lu Jiao editions), common exam equations, and popular study guides.

Collection Method

Knowledge points from source materials were converted into a unified JSONL format.

Time Range

null

Freshness

Last updated on 2025-08 24.

Geography

China