File size: 11,048 Bytes
370b2cf
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
---
license: mit
pretty_name: CCORE Three.js 3D Scenes v2
language:
- en
task_categories:
- text-generation
size_categories:
- 1K<n<10K
tags:
- 3d
- threejs
- code
- reasoning
- conversational
- html
- javascript
- voxel
- multidomain
configs:
- config_name: default
  data_files:
  - split: train
    path:
    - data/train-00000.parquet
    - data/train-00001.parquet
    - data/train-00002.parquet
    - data/train-00003.parquet
    - data/train-00004.parquet
    - data/train-00005.parquet
    - data/train-00006.parquet
  - split: validation
    path:
    - data/validation-00000.parquet
  default: true
- config_name: previews
  data_files:
  - split: train
    path:
    - previews/train-00000.parquet
    - previews/train-00001.parquet
    - previews/train-00002.parquet
    - previews/train-00003.parquet
    - previews/train-00004.parquet
    - previews/train-00005.parquet
    - previews/train-00006.parquet
    - previews/train-00007.parquet
    - previews/train-00008.parquet
    - previews/train-00009.parquet
    - previews/train-00010.parquet
    - previews/train-00011.parquet
    - previews/train-00012.parquet
    - previews/train-00013.parquet
    - previews/train-00014.parquet
    - previews/train-00015.parquet
    - previews/train-00016.parquet
    - previews/train-00017.parquet
    - previews/train-00018.parquet
    - previews/train-00019.parquet
    - previews/train-00020.parquet
    - previews/train-00021.parquet
    - previews/train-00022.parquet
    - previews/train-00023.parquet
    - previews/train-00024.parquet
    - previews/train-00025.parquet
  - split: validation
    path:
    - previews/validation-00000.parquet
    - previews/validation-00001.parquet
dataset_info:
- config_name: default
  features:
  - name: id
    dtype: string
  - name: title
    dtype: string
  - name: level
    dtype: int32
  - name: domain
    dtype: string
  - name: voxel
    dtype: bool
  - name: messages
    list:
    - name: role
      dtype: string
    - name: content
      dtype: string
    - name: reasoning_content
      dtype: string
  splits:
  - name: train
    num_examples: 3263
    num_bytes: 296407746
  - name: validation
    num_examples: 182
    num_bytes: 16581829
  download_size: 107926293
  dataset_size: 312989575
- config_name: previews
  features:
  - name: id
    dtype: string
  - name: title
    dtype: string
  - name: level
    dtype: int32
  - name: domain
    dtype: string
  - name: voxel
    dtype: bool
  - name: image
    dtype: image
  splits:
  - name: train
    num_examples: 3263
    num_bytes: 883582393
  - name: validation
    num_examples: 182
    num_bytes: 49506275
  download_size: 932995397
  dataset_size: 933088668
---

<a href="#next-releases"><img src="assets/important-note.svg" alt="Important note — Next releases" width="380" /></a>

[![9 rendered scene examples](assets/scenes-grid-3x3.jpg)](assets/scenes-grid-3x3.jpg)

# CCORE Three.js 3D Scenes v2

**3,445 curated scene examples across 50 domains, including 1,431 voxel scenes, with scene prompts, curated reasoning traces, HTML/JavaScript responses and rendered references.**

ThreeJS-3D-Scenes-v2 extends [ThreeJS-3D-Scenes](https://huggingface.co/datasets/ccore/ThreeJS-3D-Scenes) with knowledge-oriented 3D scenes. The collection covers physical relationships, biological processes, ecological change, geography, history, technology, mathematics, everyday life and interactive systems.

Examples range from compact teaching objects to detailed environments, animated mechanisms, simulations and voxel worlds. Scene requests connect visible geometry or behavior to a concept, comparison or process.

## Dataset contents

| Split | Examples | Rendered references |
| --- | ---: | ---: |
| Train | 3,263 | 3,263 |
| Validation | 182 | 182 |
| Total | 3,445 | 3,445 |

The release contains 1,431 voxel scenes and 2,014 scenes in other visual styles. Complexity levels run from 1 to 8 and describe the annotated task difficulty. They are curriculum labels.

Related examples belong to concept families. The validation split holds out one family in each domain using a deterministic selection, with every retained example from that family assigned to the same split. Text and image configurations share identifiers, splits and row order.

## Domains and classes

The `domain` field identifies one of 50 subject classes. Concepts span four curriculum stages, from structure and local relationships to contextual scenes and integrated systems.

| Domain class | Examples | Voxel | Other styles |
| --- | ---: | ---: | ---: |
| Accessibility Public Services | 76 | 27 | 49 |
| Agriculture Food Systems | 73 | 24 | 49 |
| Ancient History Daily Life | 64 | 22 | 42 |
| Animals Behavior | 75 | 26 | 49 |
| Architecture Construction | 72 | 23 | 49 |
| Cells Microbiology | 71 | 22 | 49 |
| Characters Animation | 76 | 39 | 37 |
| Chemistry Materials | 70 | 24 | 46 |
| City Africa Middle East | 53 | 39 | 14 |
| City Americas | 59 | 46 | 13 |
| City Asia Oceania | 61 | 44 | 17 |
| City Brazil | 34 | 26 | 8 |
| City Europe | 60 | 47 | 13 |
| Civic Society Cooperation | 71 | 35 | 36 |
| Clothing Personal Objects | 75 | 26 | 49 |
| Computation Algorithms | 68 | 21 | 47 |
| Culture Music Art | 73 | 37 | 36 |
| Daily Events Narratives | 76 | 26 | 50 |
| Earth Geology | 75 | 25 | 50 |
| Economics Production Exchange | 69 | 35 | 34 |
| Electricity Magnetism | 77 | 25 | 52 |
| Everyday Household | 73 | 27 | 46 |
| Furniture Interiors | 76 | 26 | 50 |
| Game Platform Puzzles | 76 | 43 | 33 |
| Game Sports Play | 72 | 37 | 35 |
| Game Strategy Systems | 76 | 45 | 31 |
| Genetics Evolution | 76 | 27 | 49 |
| Geography Landscapes Regions | 56 | 16 | 40 |
| Geography Maps Navigation | 75 | 26 | 49 |
| History Technology Science | 75 | 26 | 49 |
| History Trade Migration | 48 | 15 | 33 |
| Human Anatomy Ergonomics | 70 | 23 | 47 |
| Kitchen Food | 69 | 24 | 45 |
| Language Communication Symbols | 69 | 31 | 38 |
| Machines Fabrication | 66 | 33 | 33 |
| Mathematics Geometry | 73 | 25 | 48 |
| Mathematics Probability Data | 73 | 24 | 49 |
| Medieval Early Modern History | 71 | 25 | 46 |
| Modern History Events | 38 | 14 | 24 |
| Physics Fluids Heat | 73 | 26 | 47 |
| Physics Mechanics | 71 | 24 | 47 |
| Physics Waves Optics Sound | 74 | 24 | 50 |
| Plants Ecosystems | 72 | 25 | 47 |
| Shops Street Life | 76 | 26 | 50 |
| Space Astronomy | 63 | 33 | 30 |
| Sustainability Infrastructure | 70 | 36 | 34 |
| Tools Repairs | 73 | 26 | 47 |
| Toys Play Learning | 70 | 23 | 47 |
| Transport Infrastructure | 71 | 40 | 31 |
| Water Weather Climate | 72 | 22 | 50 |
| Total | 3,445 | 1,431 | 2,014 |

## Complexity

| Level | Examples |
| --- | ---: |
| 1 | 247 |
| 2 | 562 |
| 3 | 633 |
| 4 | 330 |
| 5 | 619 |
| 6 | 322 |
| 7 | 525 |
| 8 | 207 |

Machine-readable counts are available in [dataset-stats.json](stats/dataset-stats.json) and [domain-counts.csv](stats/domain-counts.csv).

## Format

The default configuration contains training conversations in Parquet format.

| Field | Description |
| --- | --- |
| `id` | Stable identifier shared with the image configuration. |
| `title` | Scene title. |
| `level` | Annotated complexity from 1 to 8. |
| `domain` | Subject class. |
| `voxel` | Boolean annotation for voxel style. |
| `messages` | System instruction, user prompt and assistant target. |

Each message contains `role`, `content` and `reasoning_content`. The assistant message carries the complete response in `content` and the curated reasoning trace in `reasoning_content`. The system and user messages have a null `reasoning_content`.

The optional `previews` configuration contains `id`, `title`, `level`, `domain`, `voxel` and an `image` feature. Each image is a rendered reference for the corresponding scene.

## Loading and training

```python
from datasets import load_dataset

dataset = load_dataset("ccore/ThreeJS-3D-Scenes-v2", token=True)
sample = dataset["train"][0]

prompt_messages = sample["messages"][:-1]
assistant_target = sample["messages"][-1]
thinking = assistant_target["reasoning_content"]
response = assistant_target["content"]

# Optional visual references, aligned by split, row order and ID.
previews = load_dataset("ccore/ThreeJS-3D-Scenes-v2", "previews", token=True)
assert previews["train"][0]["id"] == sample["id"]
image = previews["train"][0]["image"]
```

For supervised fine-tuning, use the system and user messages as context and the assistant message as the target. Format the thinking trace and response with the target model's chat template. Renderers and serving code should extract the HTML document from the assistant response.

Keep the provided validation split separate during training. The reference images can support visual inspection and image-conditioned experiments. Animation and interaction can require multiple states to assess behavior.

## Review and scope

Included examples have rendered evidence and visual review. Scenes may use low-poly, voxel, schematic or deliberately simplified representations. Toy models and illustrative diagrams should be interpreted within the assumptions stated by each example.

This dataset is intended for research and training in 3D scene generation and domain-oriented creative coding.

## Related datasets

- [CCORE Three.js 3D Scenes](https://huggingface.co/datasets/ccore/ThreeJS-3D-Scenes)
- [CCORE Three.js 3D GLB Assets](https://huggingface.co/datasets/ccore/ThreeJS-3D-GLB-Assets)

## License

Released under the [MIT License](LICENSE), to the extent of the rights held by the dataset publisher.

## Next releases

Beyond this release, I want to build **larger, carefully reviewed datasets** and **more capable small models**, including specialized models in other areas.

I've developed and funded this work myself, and I'm releasing the **models, weights, and datasets free of charge to the community**.

Next, I'm planning **100,000 high-quality examples for 3D and general-purpose coding harnesses** to **train new small models** that can use tools, test their outputs, and refine their work through iteration.

I expect **creating these datasets and training the new models to require around 20 times the effort of the current project**. **I need the community's support to make this next stage possible.**

## Support the project

I cover the cost of data preparation, validation, experiments, and GPU training myself. If this dataset or the CCORE models are useful to you and you want to help fund larger datasets and future open releases, you can support the work directly:

### Ko-fi

[**Support CCORE on Ko-fi**](https://ko-fi.com/ccorecaio)

### Crypto donations

**Bitcoin (BTC) — Bitcoin Network (SegWit)**

`bc1qr2gnr9g0uzk8rmsq0k6xpjgg4l7r0rfyrq69e6`

**Ethereum (ETH) — Ethereum Mainnet (ERC20)**

`0xa17608FCdeAd9e717Cd42e04127175527b4E75CB`

**Solana (SOL / USDC / USDT) — Solana Network**

`GhNmYmezKi1uWGKZeVxgddjiRW2rYNHZrf84ZaMcnha`

Your support helps pay for compute, better datasets, and future public releases. Thank you for supporting independent open-source work.