Spaces:
Running
Running
File size: 2,663 Bytes
54708e8 3bf2770 03b0173 3bf2770 03b0173 b4d728d 3bf2770 b4d728d 3bf2770 b4d728d 54708e8 b4d728d 54708e8 b4d728d 54708e8 b4d728d 54708e8 b4d728d 54708e8 b4d728d 54708e8 b4d728d 3bf2770 03b0173 b4d728d 3bf2770 b4d728d 3bf2770 b4d728d 03b0173 3bf2770 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 | """UI copy."""
TITLE = """<h1 align="center" id="space-title">🧪 PhysInOne Benchmark Leaderboard</h1>"""
INTRODUCTION_TEXT = """
**PhysInOne Benchmark** — Visual Physics Learning & Reasoning in One Suite.
Created by [vLAR Group](https://vlar-group.github.io/) / HK PolyU.
**How to Participate**
1. Create a dataset under **your own** Hugging Face account (public is recommended), and upload your scene-level ZIP files using the structure below.
2. Fill in the "Submit" tab with: display name / task / `user_dataset` (for example, `username/repo`).
3. The system will enqueue the job, and the backend Worker will evaluate it **scene by scene** asynchronously. You can track progress in the "Queue" tab.
"""
DATASET_LAYOUT_TEXT = """
### Dataset Layout
```
<user_dataset>/
└── predictions/
└── <task_name>/
├── scene_001.zip
├── scene_002.zip
└── ...
```
- Package each scene prediction as a ZIP file (recommended size: < 200 MB).
- You only need to prepare ZIP files for the tasks you want to submit; see each task's "Expected layout" section for details.
- Missing scenes will be recorded as `missing` and will affect the final score according to the task's aggregation policy.
"""
SUBMIT_INSTRUCTIONS = """
## 📦 Submission Guide
| Field | Description |
|---|---|
| Display name | Public team label (can be changed later); 2-40 characters, allowing letters, numbers, Chinese characters, `_`, `-`, `.`, and spaces |
| Task | Select from the dropdown |
| User dataset | Format: `username/repo`; **your identity = the dataset owner**, so it cannot be forged |
After submission, the job will appear in the "Queue" tab. Status progresses as `pending → running → done / failed`, and the UI will show how many scenes have been completed.
"""
ABOUT_TEXT = """
## About This Leaderboard
- Architecture: public Frontend Space + private Worker Space + private dataset.
- Evaluation uses **scene** as the smallest atomic unit and supports resumable execution.
- User identity is derived from the owner of `user_dataset`, so **OAuth is not required** and name collisions are avoided.
- Ground truth data is fully isolated and inaccessible to participants.
- See [DESIGN.md](https://huggingface.co/spaces/vLAR/PhysInOne-Leaderboard/blob/main/DESIGN.md) for full details.
"""
CITATION_BUTTON_LABEL = "If this work is helpful to you, please consider citing:"
CITATION_BUTTON_TEXT = r"""@article{physinone2026,
title = {PhysInOne: Visual Physics Learning and Reasoning in One Suite},
author = {vLAR Group, HK PolyU},
year = {2026},
eprint = {2604.09415},
archivePrefix = {arXiv}
}
"""
|