CERN/colliderml-benchmark-results
收藏资源简介:
ColliderML基准测试结果数据集是一个用于存储ColliderML基准测试任务机器评分排行榜结果的集合。该数据集包含多个基准任务,如粒子追踪(tracking)、喷注分类(jets)、异常检测(anomaly)等,每个任务使用特定的数据集(如ttbar_pu200)并基于主要指标(如TrackML加权效率、AUC等)进行评分。数据以JSON文件形式组织,存储了提交ID、任务名称、提交者、分数、时间戳等信息,并自动由ColliderML后端推送。该数据集支持ColliderML排行榜空间的公开表格渲染,并允许模型通过配置文件链接回此数据集以在模型卡片中自动显示分数。
The ColliderML Benchmark Results dataset is a collection of machine-scored leaderboard results for ColliderML benchmark tasks. It includes multiple benchmark tasks such as tracking, jets, anomaly detection, etc., each using specific datasets (e.g., ttbar_pu200) and evaluated based on primary metrics (e.g., TrackML weighted efficiency, AUC). The data is organized in JSON files containing fields like submission ID, task name, submitter, scores, and timestamps, and is automatically pushed by the ColliderML backend. This dataset powers the public leaderboard table in the ColliderML leaderboard space and allows models to link back to it via configuration files for automatic score display on model cards.




