Models 模型
Select a model to view its evaluation videos on the 50 tasks below. Ground Truth is provided for reference only. 选择模型后,下方将显示该模型在 50 个任务上的评测视频。Ground Truth 仅供对照。
WorldArena 2.0 · Track 1 WorldArena 2.0 · Track 1
This announcement publishes the Track 1 top-10 evaluation videos on the 50 official tasks, with one original video per task. Twenty-five tasks are in-distribution scenes on a white table (ID). Twenty-five are out-of-distribution scenes on a colored or textured table (OOD). Switching models keeps the task set unchanged and replaces only the corresponding video. Ground Truth is the official reference and is not a participating submission. 本公告发布 Track 1 前十名模型在 50 个官方任务上的评测视频,每个任务对应一条原始评测视频。25 个任务为白桌场景(ID),25 个任务为彩桌或纹理桌场景(OOD)。切换模型时任务保持不变,仅更换对应视频。Ground Truth 为官方参考,不是参赛提交。
Models 模型
Select a model to view its evaluation videos on the 50 tasks below. Ground Truth is provided for reference only. 选择模型后,下方将显示该模型在 50 个任务上的评测视频。Ground Truth 仅供对照。