SciCode Domain Code: 1.1 Billion Lines of Code from Scientific GitHub Repositories

Over 1.1 billion lines of code collected from GitHub repositories, focusing on interdisciplinary fields like biology, chemistry, and materials science. The dataset, created by liuhangbiao, is organized into 178 domain-specific topics across approximately 115 GB of data. It was last updated on 2026-04-05.

Use Cases

• Training domain-specific code generation models based on the large-scale collection of scientific code.
• Analyzing coding patterns and practices in fields like biology and chemistry based on the dataset's topical organization.
• Pre-training or fine-tuning models for code summarization or documentation based on the repository source code.
• Studying the intersection of software engineering and scientific research based on the interdisciplinary code samples.

Strengths

• Contains over 1.1 billion lines of code, providing substantial volume for model training.
• Covers 178 distinct domain topics, offering breadth across scientific fields.
• Sourced from GitHub repositories, providing real-world, practical code examples.

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.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

GitHub repositories

Collection Method

Collected and organized from public repositories.

Freshness

Last updated 2026-04-05 16:42:01; freshness should be verified.