GSMA/leaderboard
收藏资源简介:
该数据集名为Open Telco Leaderboard Scores,是一个电信领域基准测试得分数据集。它包含了84个模型在7个电信领域基准测试中的得分,数据来源于MWC(世界移动通信大会)排行榜。数据集仅发布得分信息,不包含能源指标。主要文件包括:leaderboard_scores.csv(扁平表格,用于数据集查看器)和leaderboard_scores.json(结构化JSON,包含每个模型的基准测试得分和标准误差)。数据模式中,核心列有:model(模型名称)、provider(模型提供商,如OpenAI、Google、Meta)、rank(按平均得分降序排列的排名)、average(可用基准测试得分的平均值)、benchmarks_completed(具有得分的基准测试数量)。此外,每个基准测试对应一列,每单元格以JSON元组形式包含[score, stderr](得分和标准误差),如果未评估则为空。基准测试包括:teleqna(电信问答,多项选择)、teletables(表格理解)、oranbench(O-RAN知识)、srsranbench(srsRAN知识)、telemath(电信数学问题)、telelogs(电信日志分析)、three_gpp(3GPP规范知识)。数据集总大小为19773字节,包含87个训练样本,下载大小为13690字节。使用示例展示了如何通过Hugging Face的datasets库加载数据集。
The dataset is named Open Telco Leaderboard Scores. It provides benchmark scores for 84 models across 7 telecom-domain benchmarks, sourced from the MWC leaderboard. This dataset publishes scores only, without energy metrics. The main files include: leaderboard_scores.csv (a flat table for the dataset viewer) and leaderboard_scores.json (structured JSON with per-model benchmark scores and standard errors). The schema features core columns: model (model name), provider (model provider, e.g., OpenAI, Google, Meta), rank (rank by average score in descending order), average (mean of available benchmark scores), benchmarks_completed (number of benchmarks with scores). Additionally, there is one column per benchmark, where each cell contains [score, stderr] as a JSON tuple, or is empty if not evaluated. Benchmarks include: teleqna (Telecom Q&A, multiple choice), teletables (Table understanding), oranbench (O-RAN knowledge), srsranbench (srsRAN knowledge), telemath (Telecom math problems), telelogs (Telecom log analysis), three_gpp (3GPP specification knowledge). The total dataset size is 19773 bytes, with 87 training examples, and a download size of 13690 bytes. Usage example demonstrates loading the dataset via Hugging Faces datasets library.




