BOSS (Benchmark for Observation Space Shift)
收藏资源简介:
BOSS数据集是一个针对观察空间偏移问题的综合评估基准,由美国北卡罗来纳大学教堂山分校的研究团队创建。该数据集在Libero仿真平台上构建,包含三个挑战:单次谓词偏移、累积谓词偏移和真实长距离任务。这些挑战旨在验证观察空间偏移问题,并评估其对长距离任务的影响。数据集通过规则自动修改生成器(RAMG)扩展了现有Libero数据集的视觉多样性,为研究社区提供了有价值的研究资源。
The BOSS Dataset is a comprehensive evaluation benchmark for the observation space shift problem, created by a research team from the University of North Carolina at Chapel Hill, United States. Constructed on the Libero simulation platform, this benchmark includes three challenges: single-step predicate shift, cumulative predicate shift, and realistic long-horizon tasks. These challenges are designed to validate the observation space shift problem and evaluate its impact on long-horizon tasks. The dataset expands the visual diversity of the existing Libero dataset via the Rule-based Automatic Modification Generator (RAMG), providing a valuable research resource for the global research community.

- 1BOSS: Benchmark for Observation Space Shift in Long-Horizon Task美国北卡罗来纳大学教堂山分校 · 2025年



