HomeSafeBench: A Benchmark for Embodied Vision-Language Models in Free-Exploration Home Safety Inspection
Paper • 2509.23690 • Published
Datasets:
meta dict | graph dict | dangers listlengths 1 5 |
|---|---|---|
{
"sample_id": "train_000001_env_3_bedroom_N1",
"env": 3,
"room": "bedroom",
"N": 1,
"character_init_position": [
10,
0,
-6
]
} | {
"nodes": [
{
"id": 11,
"category": "Rooms",
"class_name": "kitchen",
"prefab_name": "PRE_ROO_Kitchen_03",
"obj_transform": {
"position": [
0,
0,
0
],
"rotation": [
0,
0,
0,
1
],
... | [
{
"id": 0,
"danger_type": "falling",
"name": "bookshelf",
"objects": [
{
"id": 295,
"class_name": "keyboard"
}
]
}
] |
{"sample_id":"train_000002_env_3_bathroom_N1","env":3,"room":"bathroom","N":1,"character_init_positi(...TRUNCATED) | {"nodes":[{"id":11,"category":"Rooms","class_name":"kitchen","prefab_name":"PRE_ROO_Kitchen_03","obj(...TRUNCATED) | [
{
"id": 0,
"danger_type": "trip",
"name": "floor",
"objects": [
{
"id": 109,
"class_name": "microwave"
}
]
}
] |
{"sample_id":"train_000003_env_3_kitchen_N1","env":3,"room":"kitchen","N":1,"character_init_position(...TRUNCATED) | {"nodes":[{"id":11,"category":"Rooms","class_name":"kitchen","prefab_name":"PRE_ROO_Kitchen_03","obj(...TRUNCATED) | [{"id":0,"danger_type":"fire","name":"sofa","objects":[{"id":327,"class_name":"folder"},{"id":160,"c(...TRUNCATED) |
{"sample_id":"train_000004_env_0_bedroom_N1","env":0,"room":"bedroom","N":1,"character_init_position(...TRUNCATED) | {"nodes":[{"id":11,"category":"Rooms","class_name":"bathroom","prefab_name":"PRE_ROO_Bathroom_00","o(...TRUNCATED) | [
{
"id": 0,
"danger_type": "trip",
"name": "floor",
"objects": [
{
"id": 192,
"class_name": "box"
}
]
}
] |
{"sample_id":"train_000005_env_3_livingroom_N1","env":3,"room":"livingroom","N":1,"character_init_po(...TRUNCATED) | {"nodes":[{"id":11,"category":"Rooms","class_name":"kitchen","prefab_name":"PRE_ROO_Kitchen_03","obj(...TRUNCATED) | [{"id":0,"danger_type":"fire","name":"sofa","objects":[{"id":243,"class_name":"box"},{"id":160,"clas(...TRUNCATED) |
{"sample_id":"train_000006_env_0_livingroom_N1","env":0,"room":"livingroom","N":1,"character_init_po(...TRUNCATED) | {"nodes":[{"id":11,"category":"Rooms","class_name":"bathroom","prefab_name":"PRE_ROO_Bathroom_00","o(...TRUNCATED) | [
{
"id": 0,
"danger_type": "falling",
"name": "cabinet",
"objects": [
{
"id": 313,
"class_name": "microwave"
}
]
}
] |
{"sample_id":"train_000007_env_0_bathroom_N1","env":0,"room":"bathroom","N":1,"character_init_positi(...TRUNCATED) | {"nodes":[{"id":11,"category":"Rooms","class_name":"bathroom","prefab_name":"PRE_ROO_Bathroom_00","o(...TRUNCATED) | [
{
"id": 0,
"danger_type": "electric",
"name": "bathtub",
"objects": [
{
"id": 308,
"class_name": "toaster"
}
]
}
] |
{"sample_id":"train_000008_env_0_kitchen_N1","env":0,"room":"kitchen","N":1,"character_init_position(...TRUNCATED) | {"nodes":[{"id":11,"category":"Rooms","class_name":"bathroom","prefab_name":"PRE_ROO_Bathroom_00","o(...TRUNCATED) | [
{
"id": 0,
"danger_type": "children",
"name": "bench",
"objects": [
{
"id": 272,
"class_name": "cutleryfork"
}
]
}
] |
{"sample_id":"train_000009_env_0_kitchen_N1","env":0,"room":"kitchen","N":1,"character_init_position(...TRUNCATED) | {"nodes":[{"id":11,"category":"Rooms","class_name":"bathroom","prefab_name":"PRE_ROO_Bathroom_00","o(...TRUNCATED) | [
{
"id": 0,
"danger_type": "children",
"name": "bench",
"objects": [
{
"id": 56,
"class_name": "perfume"
}
]
}
] |
{"sample_id":"train_000010_env_3_kitchen_N1","env":3,"room":"kitchen","N":1,"character_init_position(...TRUNCATED) | {"nodes":[{"id":11,"category":"Rooms","class_name":"kitchen","prefab_name":"PRE_ROO_Kitchen_03","obj(...TRUNCATED) | [
{
"id": 0,
"danger_type": "fire",
"name": "stove",
"objects": [
{
"id": 327,
"class_name": "folder"
}
]
}
] |
End of preview. Expand in Data Studio
HomeSafeBench
This repository contains the official data release for the paper HomeSafeBench: Benchmarking Embodied Vision-Language Models in Free-Exploration Home Safety Inspection.
HomeSafeBench is a benchmark for free-exploration home safety inspection with embodied vision-language models, built on VirtualHome.
Dataset Contents
data/train/: 3,400 training tasks.data/test/: 1,000 human-validated evaluation tasks.
The CueBack data contains 3,158 aligned trajectories in three variants. They share the same first-person observations, tool calls, and tool feedback, but differ in the supervision text preceding each action:
cueback/sft_action.jsonl: executable actions without reasoning.cueback/sft_both.jsonl: the same actions with rationales generated directly from the trajectory prefix, current observation, and action.cueback/cueback.jsonl: the same actions with clue-based reasoning constructed by CueBack.cueback/images/: the referenced first-person observations.
The benchmark format, setup instructions, and evaluation code are documented in the HomeSafeBench code repository.
Citation
@misc{yao2026homesafebenchbenchmarkembodiedvisionlanguage,
title={HomeSafeBench: Benchmarking Embodied Vision-Language Models in Free-Exploration Home Safety Inspection},
author={Jiashu Yao and Haoyu Wen and Siyuan Gao and Yuhang Guo and Zeming Liu and Heyan Huang},
year={2026},
eprint={2509.23690},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2509.23690},
}
License
The HomeSafeBench dataset is released under the Creative Commons Attribution 4.0 International License.
- Downloads last month
- 2