File size: 2,706 Bytes
1fea1aa
 
 
fc02817
1fea1aa
 
 
 
 
 
 
4017e5f
1fea1aa
 
 
 
 
 
 
 
 
 
4017e5f
1fea1aa
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
f4d9615
 
 
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
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
---
license: cc-by-4.0
task_categories:
  - text-to-video
  - robotics
tags:
  - world-model
  - physics-simulation
  - action-conditioned
  - video-prediction
  - robotics

pretty_name: ACWM-Phys
size_categories:
  - 10K<n<100K
---

# ACWM-Phys Dataset

**ACWM-Phys: Investigating Generalized Physical Interaction in Action-Conditioned Video World Models**

> Haotian Xue†, Yipu Chen\*, Liqian Ma\*, Zelin Zhao, Lama Moukheiber, Yongxin Chen  
> Georgia Institute of Technology

[[Project Page]](https://xavihart.github.io/ACWM-Phys) · [[Paper]](#) · [[Checkpoints]](https://huggingface.co/t1an/ACWM-Phys-checkpoints)

---

## Overview

ACWM-Phys is a benchmark dataset for evaluating action-conditioned video world models under diverse physical dynamics. It spans **8 environments** across 4 physics regimes, each with 1,000 training trajectories and controlled in-distribution (InD) / out-of-distribution (OoD) test splits.

| Category | Environments | OoD Axis |
|---|---|---|
| Rigid-Body | Push Cube, Stack Cube | Unseen workspace / object count |
| Deformable | Push Rope, Cloth Move | Unseen stiffness / cloth size |
| Particle | Push Sand, Pour Water | Doubled particle count / unseen water volume |
| Kinematics | Robot Arm, Reacher | Expanded workspace / corner-sector goals |

---

## Repository Structure

```
rigid_dynamics/
├── push_block/
│   ├── ind_train/   (1,000 episodes)
│   ├── ind_test/
│   └── ood_test/
└── stack_cube/      (same structure)
deformable/
├── push_rope/
└── clothmove/
particle/
├── push_sand/
└── pour_water/
kinematics/
├── robot_arm_64/
└── reacher/
```

Each split directory contains:
- **`episode_{i}.mp4`** — RGB video at 10 fps, 240×240 (240×400 for Push Sand)
- **`metadata.pt`** — `torch.load` → list of episode dicts

Each episode dict has:

| Field | Type | Description |
|---|---|---|
| `video_path` | `str` | Filename relative to split dir, e.g. `episode_0.mp4` |
| `actions` | `FloatTensor [T, action_dim]` | Per-step action sequence |
| `length` | `int` | Number of frames T |
| `seed` | `int` | Simulation random seed |
| `episode_idx` | `int` | Global episode index (some environments) |

---

## Download

```bash
huggingface-cli download t1an/ACWM-Phys --repo-type dataset --local-dir ./data
export ACWM_DATA_ROOT=./data
```

---

## Usage Example

```python
import torch

metadata = torch.load("data/rigid_dynamics/push_block/ind_train/metadata.pt", weights_only=False)
entry = metadata[0]
# entry["video_path"]  → "episode_0.mp4"
# entry["actions"]     → Tensor of shape [T, 2]
# entry["length"]      → 16
```

---

## Citation

## Citation

Coming soon.