IS-Bench

What is IS-Bench?

IS-Bench is the first evaluation benchmark dedicated to assessing the safety of embodied agents during their interaction with home environments. It encompasses…

Released

2025-07-07

Modality

multimodal

Evaluates

多模态, Multimodal, 安全, Safety, 智能体, Agent, task planning, 安全对齐, Safety Alignment, 任务执行, Task Execution, 跨模态推理, Cross-modal Reasoning, 不支持, Unsupported

Openness

open

Importer

OpenCompass Hub

Reported scores

0

Source provenance

• Original evidence https://hub.opencompass.org.cn/dataset-detail/IS-Bench

IS-Bench paper, code and dataset

• Paper https://arxiv.org/abs/2506.16402
• Code https://github.com/AI45Lab/IS-Bench
• Dataset https://huggingface.co/datasets/Ursulalala/IS_Bench_dataset
• Dataset https://huggingface.co/datasets/Ursulalala/IS_Bench_scenes

Which sources cite IS-Bench?

• IS-Bench

Related 多模态, Multimodal, 安全, Safety, 智能体, Agent, task planning, 安全对齐, Safety Alignment, 任务执行, Task Execution, 跨模态推理, Cross-modal Reasoning, 不支持, Unsupported benchmarks

• 3DCodeBench
• A-Bench
• A-OKVQA
• AbsPyramid
• ActBench
• ActionAtlas
• ActiveFly-Bench
• Ada-LEval