id int64 0 47 | task_name stringclasses 4 values | variation int64 0 4 | episode_num int64 0 5 | instruction stringlengths 23 77 | task_type stringclasses 4 values |
|---|---|---|---|---|---|
0 | pick_cube_shape | 0 | 0 | Pick up the star and place it into the silver container. | pick |
1 | pick_cube_shape | 0 | 1 | Pick up the star and place it into the blue container. | pick |
2 | pick_cube_shape | 0 | 2 | Pick up the star and place it into the red container. | pick |
3 | pick_cube_shape | 1 | 0 | Pick up the triangular prism and place it into the silver container. | pick |
4 | pick_cube_shape | 1 | 1 | Pick up the triangular prism and place it into the black container. | pick |
5 | pick_cube_shape | 1 | 2 | Pick up the triangular prism and place it into the navy container. | pick |
6 | pick_cube_shape | 2 | 0 | Pick up the cylinder and place it into the lime container. | pick |
7 | pick_cube_shape | 2 | 1 | Pick up the cylinder and place it into the teal container. | pick |
8 | pick_cube_shape | 2 | 2 | Pick up the cylinder and place it into the purple container. | pick |
9 | pick_cube_shape | 3 | 0 | Pick up the cube and place it into the lime container. | pick |
10 | pick_cube_shape | 3 | 1 | Pick up the cube and place it into the orange container. | pick |
11 | pick_cube_shape | 3 | 2 | Pick up the cube and place it into the white container. | pick |
12 | place_into_shape_sorter_color | 0 | 1 | Put the red star into the shape sorter. | place |
13 | place_into_shape_sorter_color | 0 | 3 | Put the red star into the shape sorter. | place |
14 | place_into_shape_sorter_color | 1 | 1 | Put the maroon star into the shape sorter. | place |
15 | place_into_shape_sorter_color | 1 | 2 | Put the maroon star into the shape sorter. | place |
16 | place_into_shape_sorter_color | 1 | 3 | Put the maroon star into the shape sorter. | place |
17 | place_into_shape_sorter_color | 2 | 0 | Put the lime star into the shape sorter. | place |
18 | place_into_shape_sorter_color | 2 | 1 | Put the lime star into the shape sorter. | place |
19 | place_into_shape_sorter_color | 2 | 3 | Put the lime star into the shape sorter. | place |
20 | place_into_shape_sorter_color | 3 | 0 | Put the green star into the shape sorter. | place |
21 | place_into_shape_sorter_color | 3 | 2 | Put the green star into the shape sorter. | place |
22 | place_into_shape_sorter_color | 4 | 0 | Put the blue star into the shape sorter. | place |
23 | place_into_shape_sorter_color | 4 | 3 | Put the blue star into the shape sorter. | place |
24 | stack_cubes_color | 0 | 0 | Stack the red cube and the black cube in sequence. | stack |
25 | stack_cubes_color | 0 | 1 | Stack the red cube and the navy cube in sequence. | stack |
26 | stack_cubes_color | 0 | 2 | Stack the red cylinder and the magenta cylinder in sequence. | stack |
27 | stack_cubes_color | 1 | 0 | Stack the maroon cylinder and the navy cylinder in sequence. | stack |
28 | stack_cubes_color | 1 | 1 | Stack the maroon triangular prism and the olive triangular prism in sequence. | stack |
29 | stack_cubes_color | 1 | 2 | Stack the maroon moon and the rose moon in sequence. | stack |
30 | stack_cubes_color | 2 | 0 | Stack the lime cylinder and the maroon cylinder in sequence. | stack |
31 | stack_cubes_color | 2 | 1 | Stack the lime moon and the red moon in sequence. | stack |
32 | stack_cubes_color | 2 | 2 | Stack the lime moon and the gray moon in sequence. | stack |
33 | stack_cubes_color | 3 | 0 | Stack the green triangular prism and the teal triangular prism in sequence. | stack |
34 | stack_cubes_color | 3 | 1 | Stack the green moon and the purple moon in sequence. | stack |
35 | stack_cubes_color | 3 | 2 | Stack the green cube and the blue cube in sequence. | stack |
36 | wipe_table_direction | 0 | 0 | Wipe the horizontal area. | wipe |
37 | wipe_table_direction | 0 | 1 | Wipe the horizontal area. | wipe |
38 | wipe_table_direction | 0 | 2 | Wipe the horizontal area. | wipe |
39 | wipe_table_direction | 0 | 3 | Wipe the horizontal area. | wipe |
40 | wipe_table_direction | 0 | 4 | Wipe the horizontal area. | wipe |
41 | wipe_table_direction | 0 | 5 | Wipe the horizontal area. | wipe |
42 | wipe_table_direction | 1 | 0 | Wipe the vertical area. | wipe |
43 | wipe_table_direction | 1 | 1 | Wipe the vertical area. | wipe |
44 | wipe_table_direction | 1 | 2 | Wipe the vertical area. | wipe |
45 | wipe_table_direction | 1 | 3 | Wipe the vertical area. | wipe |
46 | wipe_table_direction | 1 | 4 | Wipe the vertical area. | wipe |
47 | wipe_table_direction | 1 | 5 | Wipe the vertical area. | wipe |
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 benchmark suite, designed for use with the 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
.
βββ data/
β βββ base.jsonl
β βββ common_sense.jsonl
β βββ complex.jsonl
β βββ spatial.jsonl
β βββ visual.jsonl
βββ simulator_data.zip # Binary simulation files (auto-extracted by EASI)
β βββ data/ # Per-split episode data (.ttm, .pkl)
β βββ vlm/ # Task templates and object models
β βββ amsolver/robot_ttms/ # Robot model files
βββ README.md
Data Fields (JSONL)
Each row in the JSONL files contains:
id(int): Unique identifier within the splittask_name(string): Task variation name (e.g.,pick_cube_shape)variation(int): Variation number within the taskepisode_num(int): Episode number within the variationinstruction(string): Natural language task instructiontask_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.ttmβ CoppeliaSim scene statewaypoint_sets.ttmβ Waypoint configurationconfigs.pklβ Episode metadata and success conditions
Usage
Loading with Datasets Library
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
# 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 V4.1.0
- PyRep (CoppeliaSim Python binding)
- AMSolver (modified RLBench fork)
Acknowledgements
This dataset is derived from the EmbodiedBench EB-Manipulation benchmark and uses CoppeliaSim as the simulation environment.
- Downloads last month
- 12