grow-ai-like-a-child/three-mountain-scaling

---
license: cc-by-4.0
task_categories:
- question-answering
- visual-question-answering
language:
- en
size_categories:
- 10M<n<100M
pretty_name: ThreeMountain_Scaling
dataset_info:
- config_name: lab
features:
- name: UID
dtype: string
- name: id
dtype: string
- name: question
dtype: string
- name: answer
dtype: string
- name: media
struct:
- name: prompt_img
list: string
- name: options
list: string
- name: choices
struct:
- name: A
dtype: string
- name: B
dtype: string
- name: C
dtype: string
- name: D
dtype: string
- name: scene
dtype: string
- name: setting
dtype: string
- name: category
dtype: string
- name: hint
dtype: string
- name: source
dtype: string
- name: prompt
dtype: string
- name: image0
dtype: image
- name: image1
dtype: image
- name: image2
dtype: image
- name: image3
dtype: image
- name: image4
dtype: image
splits:
- name: perspective_l2
- name: perspective_l1
- name: spatiality
- config_name: real
features:
- name: UID
dtype: string
- name: id
dtype: string
- name: question
dtype: string
- name: answer
dtype: string
- name: media
struct:
- name: prompt_img
list: string
- name: options
list: string
- name: choices
struct:
- name: A
dtype: string
- name: B
dtype: string
- name: C
dtype: string
- name: D
dtype: string
- name: scene
dtype: string
- name: setting
dtype: string
- name: category
dtype: string
- name: hint
dtype: string
- name: source
dtype: string
- name: prompt
dtype: string
- name: image0
dtype: image
- name: image1
dtype: image
- name: image2
dtype: image
- name: image3
dtype: image
- name: image4
dtype: image
splits:
- name: ego_exo
configs:
- config_name: lab
data_files:
- split: spatiality
path:
- spatiality.parquet
- spatiality_2obj.parquet
- split: perspective_l2
path: perspective_l2.parquet
- split: perspective_l1
path: perspective_mc.parquet
- config_name: real
data_files:
- split: ego_exo
path: ego-exo.parquet
---

# ThreeMountain_Scaling
| Segment | Meaning |
|----------------|-------------------------------------------------------------------------|
| `GO` | **Geometric Object** — indicates the object type used (e.g., `GO` for geometric, `RO` for real objects). |
| `L` / `Arc` | **Object Arrangement** — defines how objects are arranged spatially. `L` means L-shape arrangement; `Arc` means objects are placed in an arc. |
| `RC` | **Random Character Position** — `RC = True`: character position is randomized. |
| `FC` | **Fixed Character Position** — `FC = True`: character stays fixed. |
| `RS` | **Random Scale** — `RS = True`: objects are randomly scaled to vary in size. |

## Spatiality
| Field | Type | Description |
|--------------------|----------|------------------------------------------------------------------------------------------------------|
| `UID` | string | A unique identifier for the entry, sequentially assigned. |
| `id` | string | A human-readable identifier describing the view direction, e.g., `"1_left"`. |
| `question` | string | The natural language question posed to the user. |
| `media` | list | List of image paths. The first image is the reference, the rest are the candidate choices. |
| `answer` | string | The correct answer option among `'A'`, `'B'`, `'C'`, or `'D'`. |
| `choices` | dict | Maps choice labels (`A`-`D`) to image paths (relative to dataset root). |
| `scene` | string | The scene context for the question (e.g., `"Lab"`). |
| `setting` | string | The specific folder/configuration used to generate or organize the data (e.g., `"GO_L_RC_FC_RS"`). |
| `category` | string | High-level task category. Currently all entries are `"spatiality"`. |
| `encoding_schema` | string | Format used to encode the input, e.g., `"multi-image"`. |
| `hint` | string | Instruction reminding the user how to answer (e.g., "Answer with letter A, B, C, D"). |
| `creator` | string | Name or alias of the dataset creator. |
| `version` | string | Version of the dataset. |
| `source` | string | Where the data originated (e.g., `"self-generated"`). |
| `prompt` | string | Template text used in interactive settings; `<image>` placeholders are to be replaced with visuals. |

Example:
``` json

"UID": "00001",
"id": "1_left",
"question": "Look at the reference image...",
"media": [...],
"answer": "C",
"choices": ...,
"scene": "Lab",
"setting": "GO_L_RC_FC_RS",
"category": "spatiality",
"encoding_schema": "multi-image",
"hint": "Please answer with the option's letter...",
"creator": "kaia",
"version": "1.0",
"source": "self-generated",
"prompt": "<image>...<image>"

```