Datasets:

AIGeeksGroup
/

Code4D

text stringlengths 0 324
1. A breeze stirs through the autumn forest, gently swaying the entire tree as leaves dance in the wind.

2. A 10-second time-lapse of a summer forest day: warm pre-dawn air gives way to a bright sunrise, strong midday sun flickers through lush green canopy, late afternoon light turns golden, then a soft sunset fades into a humid, moonlit night with light mist.

3. The midday sun illuminated countless trees standing tall. Several leaves detached from their branches and fell straight down. Narrow, curved, slightly withered leaves with a yellow, matte appearance revealed brown spots fell and spun silently as they descended, finally coming to rest on the bright pile of fallen leaves.

4. A dense, lush green forest during a heavy rainstorm. Rain streaks are falling rapidly and diagonally across the entire scene. The branches and leaves of tall trees sway gently in the wind and rain.

5. A peaceful desert afternoon with a gentle breeze. Sand is flowing like liquid silk along the sharp ridgeline of a sand dune, slowly cascading down the leeward side.

6. A vibrant underwater scene. An ethereal, translucent jellyfish glowing with soft blue and purple bioluminescence drifts gracefully through the water. Its gelatinous, umbrella-shaped bell pulsates rhythmically, contracting and expanding slowly to propel itself forward.

7. A weathered chopped tree rooted in the dark forest ground. It is burning fiercely. Bright yellow flames rise from the center, while the edges of the stump are glowing with intense red embers and ash. The flickering light casts dynamic, shifting shadows on the gnarled roots spreading out into the dirt.

8. On the living room coffee table, a tall, straight-sided, thick-walled ceramic cup lay tipped on its side. The mug featured a wide, curved handle and a deep blue matte glaze. As it fell, a large volume of water poured from its wide rim, quickly flooding the tabletop, flowing toward the edge, and dripping down.

9. A brown glass bottle rolled slowly across the sunlit living room floor.

10. A cozy bedroom with warm lighting. A desk holds a classic ceramic coffee cup, its smooth surface reflecting the soft glow of the room. Steam rises gently from the cup in soft wisps, disappearing into the warm light.

Dataset Card for Code4D (Code2Worlds)

Dataset Summary

The Code4D benchmark is a dataset designed to evaluate the capability of Large Language Models (LLMs) in generating physically grounded 4D environments. It pairs natural language prompts with complex 3D scenes (provided here as .blend files) that exhibit temporal evolution, physical interactions, and atmospheric changes.

Unlike existing text-to-3D datasets that focus solely on static structures, Code4D challenges models on dynamic fidelity, including fluid dynamics, particle systems, rigid-body dynamics, and soft-body simulations.

This dataset supports the Code2Worlds framework, which formulates 4D generation as language-to-simulation code generation using a dual-stream architecture (Object Stream and Scene Stream).

Supported Tasks and Leaderboards

Text-to-4D Scene Generation: Generating dynamic 3D scenes from text descriptions.
Procedural Code Generation: Evaluating LLMs on generating Blender/Infinigen API calls.
Physics Simulation Benchmarking: Assessing the realism of generated physical interactions.

Languages

The prompts and documentation are in English.

Dataset Structure

Data Instances

Each instance in the dataset consists of a text prompt and its corresponding Blender project file (.blend).

Example:

Prompt: "A breeze stirs through the autumn forest, gently swaying the entire tree as leaves dance in the wind."
File: scene_1.blend

Data Fields

prompt (string): The natural language instruction describing the scene and desired dynamics.
blend_file (file): The Blender 3D project file containing the scene layout, assets, and simulation settings.

Dataset Creation

Curation Rationale

The dataset was constructed to address the "semantic-physical execution gap" in generative models. It specifically targets scenarios where monolithic generation fails, requiring precise control over both local object structures and global environmental layouts.

Considerations for Using the Data

Software Dependencies

To open and render the .blend files properly, you need:

Blender 4.3 or higher.
Infinigen libraries.

Computational Requirements

The benchmark scenes are designed for high-fidelity rendering.

Nature Scenes: Configured for 1920x1080 resolution, 240 frames, 128 samples.
Indoor Scenes: Configured for 1920x1080 resolution, 120 frames, 196 samples.

Citation

If you use this dataset in your research, please cite the following paper:

@article{zhang2026code2worlds,
  title={Code2Worlds: Empowering Coding LLMs for 4D World Generation},
  author={Zhang, Yi and Wang, Yunshuang and Zhang, Zeyu and Tang, Hao},
  journal={arXiv preprint arXiv:2602.11757},
  year={2026}
}

Downloads last month: 6

Paper:

Code2Worlds: Empowering Coding LLMs for 4D World Generation

Repository:

GitHub

Size of the auto-converted Parquet files:

2.72 kB

Number of rows:

Paper for AIGeeksGroup/Code4D

Code2Worlds: Empowering Coding LLMs for 4D World Generation

Paper • 2602.11757 • Published 4 days ago • 1