CodeRM-UnitTest: Synthetic Python Unit Tests for Code Reward Modeling

A curated collection of high-quality synthetic Python unit tests derived from two code instruction tuning datasets: CodeFeedback-Filtered-Instruction and the training set of TACO. The dataset was created by author KAKA22 and last updated on 2025-01-20. It was used to train CodeRM-8B, a unit test generation model.

Use Cases

• Training code reward models based on synthetic unit test quality.
• Benchmarking automated unit test generation systems based on curated Python examples.
• Fine-tuning language models for code instruction tasks based on derived datasets.
• Studying the properties of synthetic code data for AI training.

Strengths

• Dataset is derived from two established sources: CodeFeedback-Filtered-Instruction and TACO.
• Last updated on 2025-01-20, indicating recent maintenance.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect source bias inherent to the original code instruction datasets.

Provenance

Source

Derived from CodeFeedback-Filtered-Instruction and the training set of TACO.

Collection Method

Curated collection of synthetic unit tests, as described in the paper 'Dynamic Scaling of Unit Tests for Code Reward Modeling'.

Freshness

Last updated 2025-01-20 11:57:06.