README.md

---
pretty_name: MAP-Bench
language:
- en
license: other
license_name: hm3d-derived-benchmark-terms
license_link: LICENSE.md
task_categories:
- robotics
size_categories:
- 10K<n<100K
viewer: false
tags:
- robotics
- navigation
- benchmark
- map-bench
- hm3d
---

<p align="center">
  <img src="assets/map.png" width="260" alt="MAP: Manipulation Anchor Pose" />
</p>
<h1 align="center">MAP-Bench</h1>
<p align="center">
  <a href="https://arxiv.org/abs/2609.39388">Paper</a> ·
  <a href="https://fysics-ai.github.io/uniwam.github.io/">Project</a> ·
  <a href="https://github.com/Fysics-AI/UniWAM">GitHub</a> ·
  <a href="https://huggingface.co/Fysics-AI/UniWAM">Checkpoints</a>
</p>

**MAP-Bench** evaluates instruction-conditioned navigation to a **Manipulation Anchor Pose**: the base position and heading from which a robot can begin the intended manipulation. Built in HM3D, it uses one forward RGB view and three distance levels: Short, Medium, and Long.

- **35,454 training trajectories** with RGB observations, navigation actions, instructions, target boxes, and MAP annotations.
- **15,000 evaluation episodes** with scene references, start poses, target MAPs, and instructions.
- **75,000 evaluation rollouts** and their results: 15,000 each for GR00T N1.7, π₀.₅, FastWAM, UniWAM-Base, and UniWAM-Scaled.

## Download

Run from your workspace directory to download the complete benchmark package:

```bash
python -m pip install -U huggingface_hub
hf download Fysics-AI/MAP-Bench --repo-type dataset --local-dir ./MAP-Bench
```

Training videos and evaluation rollouts are stored in compressed archives. Use the [GitHub download helper](https://github.com/Fysics-AI/UniWAM/blob/main/scripts/download_assets.py) with `--training --traces` to verify and extract them. Obtain [HM3D scene assets](https://aihabitat.org/datasets/hm3d/) separately for simulator evaluation.

## Usage

See GitHub for [benchmark setup and evaluation](https://github.com/Fysics-AI/UniWAM/blob/main/docs/EVALUATION.md), [navigation training](https://github.com/Fysics-AI/UniWAM/blob/main/training/NAVIGATION.md), and [results](https://github.com/Fysics-AI/UniWAM#map-bench-results).

HM3D-derived data are subject to the source dataset terms; see [LICENSE.md](LICENSE.md).

## Citation

Please cite the UniWAM paper and HM3D when using MAP-Bench:

```bibtex
@article{xue2026uniwam,
  title   = {UniWAM Technical Report: Unified Mobile Manipulation via Mixed-Stream World-Action Modeling and Manipulation Anchor Pose Supervision},
  author  = {Wei Xue and Keliang Liu and Mingzhang Cui and Jinhua Xie and Jinjie Wei and Jianan Hou and Jingcheng Lu and Lintao Wang and Kaixiang Qiu and Yizhou Liu and Xinghai Ye and Jinghang Han and Mingcheng Li and Jie Gu and Shunli Wang and Lihua Zhang and Dingkang Yang},
  journal = {arXiv preprint arXiv:2609.39388},
  year    = {2026},
  url     = {https://arxiv.org/abs/2609.39388}
}
```