namanbnsl/rosettabench-150-stratified-compressed
收藏Hugging Face2026-04-24 更新2026-04-26 收录
下载链接：
https://hf-mirror.com/datasets/namanbnsl/rosettabench-150-stratified-compressed
下载链接
链接失效反馈官方服务：
资源简介：
RosettaBench 150分层压缩数据集是一个用于衡量学习而非记忆的无污染基准测试数据集。数据来源于LiveCodeBench的AtCoder平台问题，通过分层抽样选择了150个问题，覆盖了简单、中等和困难三个难度级别。每个问题至少有3个测试用例，平均每个问题有3.2个测试用例。数据集包含问题ID、标题、内容、难度、平台、所有测试用例和语言生成种子等列。
The RosettaBench 150 Stratified Compressed dataset is a contamination-free benchmark for measuring learning, not memorization. The data is sourced from AtCoder platform problems in LiveCodeBench, with 150 problems selected through stratified sampling across easy, medium, and hard difficulty tiers. Each problem has at least 3 test cases, with an average of 3.2 test cases per problem. The dataset includes columns for question ID, title, content, difficulty, platform, all test cases, and language generation seed.
提供机构：
namanbnsl
