Experiment data for the SwiftEval benchmark
收藏NIAID Data Ecosystem2026-05-02 收录
下载链接：
https://zenodo.org/record/14445600
下载链接
链接失效反馈官方服务：
资源简介：
This dataset includes 28 hand-crafted problems for the SwiftEval benchmark and the results of 44 experiments run with various LLMs.
It has the following structure:
problems.json - contains 28 problems
experiments - folder contains 44 experiments data
{experiment_name}
completions.jsonl - contains raw LLM text generation
evaluations.jsonl - contains extracted code and evaluation results
metadata.jsonl - contains LLM model checkpoint and generation parameters
创建时间：
2025-02-20
