metadata
license: gpl-3.0
task_categories:
- text-generation
language:
- en
tags:
- code
- python
pretty_name: NewMercury
size_categories:
- n<1K
NewMercury
This dataset is a part of an efficiency evaluation framework for LLM-generated code. It is based on Mercury, a competitive programming benchmark, with the following improvements:
- test generation facilities and extended test suites
- corrected test generation, input conversion and solution evaluation functions
- enhanced sandbox with additional library imports and adjusted time/recursion limits
See this repository on Github for more details.