Skip to content

Dataset · m-a-p

Feedback data for code generation

m-a-p/Code-Feedback

An English question-answering dataset for code generation systems with improvement.

What's inside

Code-Feedback is an English text dataset for question-answering tasks. In the card, it is linked to OpenCodeInterpreter, a family of open-source code generation systems that combine LLMs with execution and iterative refinement. The size is between 10k and 100k records.

What you can do with it

  • Training code generation models with execution
  • LoRA for iterative code correction
  • Evaluating code generation and refinement systems
Tasks
Question answering
Languages
English
Size
10K–100K rows
Downloads this month
2K
Likes
251

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("m-a-p/Code-Feedback")