EmbodiedMemoryBench / README.md
lzLiang's picture
Add files using upload-large-folder tool
887343d verified
|
Raw History Blame
1.72 kB
---
license: cc-by-nc-4.0
task_categories:
- robotics
- question-answering
- visual-question-answering
- text-generation
language:
- en
pretty_name: EmbodiedMemoryBench Full 2554
size_categories:
- 10K<n<100K
tags:
- embodied-ai
- embodied-memory
- robotics
- ai2-thor
- procthor
---
# EmbodiedMemoryBench
EmbodiedMemoryBench is a benchmark for evaluating memory in long-horizon embodied tasks. This private dataset package contains the complete **2,554-episode** benchmark used by the project.
## Contents
| Task family | Episodes |
|---|---:|
| Passive Observation | 1,036 |
| Dynamic Tracking | 1,052 |
| Interaction Experience | 263 |
| Experience Generalization | 203 |
| **Total** | **2,554** |
The `episodes/` directory contains the materialized episode files and associated context images. The `manifests/` directory contains the frozen membership manifests and audit report. `browse.tsv` provides a searchable episode index, while `metadata.json` and `verification_report.json` record the dataset counts and integrity checks.
## Integrity
The canonical membership file is `manifests/full2554.jsonl`. Episode counts and family totals are recorded in `metadata.json`; file and episode verification results are recorded in `verification_report.json`.
## License
This dataset is released for non-commercial research use under the Creative Commons Attribution-NonCommercial 4.0 International license (CC BY-NC 4.0). Third-party environments, assets, and dependencies remain subject to their respective upstream terms.
## Project
- Project page: https://zju-omniai.github.io/EmbodiedMemoryBench/
- Code and evaluation interfaces: https://github.com/ZJU-OmniAI/EmbodiedMemoryBench