---
dataset_info:
  features:
    - name: id
      dtype: int64
    - name: task_name
      dtype: string
    - name: variation
      dtype: int64
    - name: episode_num
      dtype: int64
    - name: instruction
      dtype: string
    - name: task_type
      dtype: string
  splits:
    - name: base
      num_examples: 48
    - name: common_sense
      num_examples: 48
    - name: complex
      num_examples: 48
    - name: spatial
      num_examples: 48
    - name: visual
      num_examples: 36
configs:
  - config_name: default
    data_files:
      - split: base
        path: data/base.jsonl
      - split: common_sense
        path: data/common_sense.jsonl
      - split: complex
        path: data/complex.jsonl
      - split: spatial
        path: data/spatial.jsonl
      - split: visual
        path: data/visual.jsonl
license: mit
task_categories:
  - robotics
tags:
  - embodied-ai
  - manipulation
  - robotic-manipulation
  - coppeliasim
  - benchmark
language:
  - en
size_categories:
  - n<1K
---

# EB-Manipulation Dataset

EB-Manipulation is a benchmark for evaluating LLM-controlled robotic manipulation in CoppeliaSim using a Franka Panda arm with a parallel gripper. It is part of the [EmbodiedBench](https://github.com/EmbodiedBench/EmbodiedBench) benchmark suite, designed for use with the [EASI](https://github.com/EvolvingLMMs-Lab/EASI) evaluation framework.

## Dataset Description

Agents must output sequences of 7D discrete gripper actions `[X, Y, Z, Roll, Pitch, Yaw, Gripper]` to complete manipulation tasks (picking, stacking, placing, wiping). The benchmark tests spatial reasoning, visual understanding, common sense, and complex instruction following.

### Subsets

| Subset | Description | Episodes |
|--------|-------------|----------|
| `base` | Standard manipulation tasks with direct instructions | 48 |
| `common_sense` | Tasks requiring commonsense reasoning about objects | 48 |
| `complex` | Complex multi-step manipulation instructions | 48 |
| `spatial` | Tasks with relative spatial references | 48 |
| `visual` | Tasks requiring visual property recognition | 36 |

### Task Types

| Task Type | Base Task | Description |
|-----------|-----------|-------------|
| `pick` | pick_cube | Pick up a target object and place it into a container |
| `stack` | stack_cubes | Stack cubes in a specified order |
| `place` | place_into_shape_sorter | Place objects into the correct shape sorter slots |
| `wipe` | wipe_table | Wipe a specified area on the table |

### Action Space

Each action is a 7D discrete array: `[X, Y, Z, Roll, Pitch, Yaw, Gripper_state]`

- **X, Y, Z**: 3D position in voxel grid (range [0, 100])
- **Roll, Pitch, Yaw**: Discrete Euler angles (range [0, 120], each unit = 3 degrees)
- **Gripper state**: 0 = close, 1 = open

## Dataset Structure

```
.
├── data/
│   ├── base.jsonl
│   ├── common_sense.jsonl
│   ├── complex.jsonl
│   ├── spatial.jsonl
│   └── visual.jsonl
├── simulator_data.zip          # Binary simulation files (auto-extracted by EASI)
│   ├── data/                   # Per-split episode data (.ttm, .pkl)
│   ├── vlm/                    # Task templates and object models
│   └── amsolver/robot_ttms/    # Robot model files
└── README.md
```

### Data Fields (JSONL)

Each row in the JSONL files contains:

- `id` (int): Unique identifier within the split
- `task_name` (string): Task variation name (e.g., `pick_cube_shape`)
- `variation` (int): Variation number within the task
- `episode_num` (int): Episode number within the variation
- `instruction` (string): Natural language task instruction
- `task_type` (string): Base task type (`pick`, `stack`, `place`, `wipe`)

### Simulator Data (simulator_data.zip)

Each episode's binary data is stored at:
`data/{split}/eval/{task_name}/variation{N}/episodes/episode{N}/`

- `task_base.ttm` — CoppeliaSim scene state
- `waypoint_sets.ttm` — Waypoint configuration
- `configs.pkl` — Episode metadata and success conditions

## Usage

### Loading with Datasets Library

```python
from datasets import load_dataset

# Load a specific split
dataset = load_dataset("oscarqjh/EB-Manipulation_easi", split="base")

# Access data
for example in dataset:
    print(example["instruction"])
    print(example["task_name"])
```

### Using with EASI

```bash
# Run evaluation on the base split
easi run ebmanipulation_base --agent react --backend openai --model gpt-4o

# List available manipulation splits
easi task list | grep ebmanipulation
```

## Requirements

- [CoppeliaSim](https://www.coppeliarobotics.com/) V4.1.0
- [PyRep](https://github.com/stepjam/PyRep) (CoppeliaSim Python binding)
- [AMSolver](https://github.com/EmbodiedBench/EmbodiedBench) (modified RLBench fork)

## Acknowledgements

This dataset is derived from the [EmbodiedBench](https://github.com/EmbodiedBench/EmbodiedBench) EB-Manipulation benchmark and uses [CoppeliaSim](https://www.coppeliarobotics.com/) as the simulation environment.