PyDebug-Gold-150k: Expert Python Debugging Pairs for LLM Training

150,000 expert-level Python debugging pairs intended for fine-tuning large language models and Direct Preference Optimization. The dataset is hosted on Kaggle and is categorized for computer science and natural language processing applications. The original author, organization, and specific creation date are not provided.

Use Cases

• Fine-tune code generation models based on expert debugging examples.
• Train models for automated code correction based on the provided debugging pairs.
• Optimize model alignment for programming tasks using Direct Preference Optimization (DPO).
• Benchmark LLM performance on Python debugging and code repair tasks.

Strengths

• Contains 150,000 data points, providing a substantial volume for training.
• Data is described as 'expert-level', suggesting high-quality annotations.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Last update date is unknown; freshness unverified.

Provenance

Source

kaggle