SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2605.07971 · 2026

DVD: Discrete Voxel Diffusion for 3D Generation and Editing

Yingzhen Li, Heliang Zheng, Jiaqi Wu, Fupeng Sun, Zhengrui Xiang

arXiv · PDF · Open in the Atlas

Code that ran

10 of 13 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 6 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 4 failures of our instrument; 3 no recorded run), measured 2026-11-22; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2605.07971.json · badge: /api/ran/2605.07971.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
TeCai/DVD canonical (classifier label)
classifier basis: census
10 ran · 3 no recorded run
FunctionStatusWhere it lives
asset_label Ran
ran; our contract draft was wrong, not the code
TeCai/DVD/app_dvd_image.py
code served (permissive licence) · get_code("303d1ea417a4133a")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

from pathlib import Path

ARGS = {
    'path': Path('test_asset_name.txt')
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_first_directory Ran
ran; our contract draft was wrong, not the code
TeCai/DVD/dataset_toolkits/build_metadata.py
code served (permissive licence) · get_code("29323106f1b1a408")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import os
import tempfile
import shutil

# Create a temporary directory with subdirectories for testing
temp_dir = tempfile.mkdtemp()

# Create a subdirectory so get_first_directory has something to find
subdir_1 = os.path.join(temp_dir, "first_directory")
os.makedirs(subdir_1, exist_ok=True)

# Create a file (which should be skipped since we only want directories)
file_path = os.path.join(temp_dir, "some_file.txt")
with open(file_path, 'w') as f:
    f.write("test")

ARGS = {
    "path": temp_dir
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
image_cfg_schedule Ran
ran, honoured the contract we drafted
TeCai/DVD/trellis/pipelines/dvd_voxel.py
code served (permissive licence) · get_code("45e03dd9d228fb83")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned float, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# image_cfg_schedule takes a single float parameter t
# and returns a float based on a conditional check
# No complex objects needed - just a simple float value

t_value = 0.3  # Example: t < 0.5, so should return 0.4

ARGS = {
    "t": t_value
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
list_asset_files Ran
ran, honoured the contract we drafted
TeCai/DVD/app_dvd_image.py
code served (permissive licence) · get_code("1ef2d7f4f60a1cdc")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import tempfile
from pathlib import Path

# Create a temporary directory structure for testing
temp_dir = tempfile.TemporaryDirectory()
temp_path = Path(temp_dir.name)

# Create test subdirectory with some files
test_subdir = temp_path / "assets"
test_subdir.mkdir(exist_ok=True)

# Create some test files with different suffixes
(test_subdir / "file1.txt").touch()
(test_subdir / "file2.txt").touch()
(test_subdir / "image.png").touch()
(test_subdir / "data.json").touch()
(test_subdir / "readme.md").touch()

# Temporarily override ROOT_DIR for testing
import sys
from types import ModuleType

# Create a mock module with ROOT_DIR set to our temp directory
mock_module = ModuleType('mock_list_asset_files')
mock_module.ROOT_DIR = temp_path
sys.modules['mock_list_asset_files'] = mock_module

ARGS = {
    "directory": "assets",
    "suffixes": {".txt", ".png", ".json"}
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
load_voxel_coords Ran
ran, honoured the contract we drafted
TeCai/DVD/examples_dvd_image_editing_api.py
code served (permissive licence) · get_code("076e2ec5ed8bb0de")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family float, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import os
import torch
import numpy as np
import tempfile

torch.manual_seed(0)

# Create temporary files with voxel coordinate data
temp_dir = tempfile.gettempdir()

# Create a .npy file with voxel coordinates
npy_coords = np.random.rand(2, 4).astype(np.float32)
npy_path = os.path.join(temp_dir, "test_voxels.npy")
np.save(npy_path, npy_coords)

# Create a .pt file with voxel coordinates in a dict
pt_coords = torch.randn(2, 4)
pt_path = os.path.join(temp_dir, "test_voxels.pt")
torch.save({"coords": pt_coords}, pt_path)

ARGS = {
    "path": npy_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
load_voxel_coords Ran
ran, honoured the contract we drafted
TeCai/DVD/examples_dvd_text_editing_api.py
code served (permissive licence) · get_code("5441444d27b951cc")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family float, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import os
import numpy as np
import torch
import tempfile

torch.manual_seed(0)

# Create temporary directory for test files
temp_dir = tempfile.mkdtemp()

# Test with .npy file
npy_path = os.path.join(temp_dir, 'voxels.npy')
npy_data = np.random.rand(2, 4).astype(np.float32)
np.save(npy_path, npy_data)

# Test with .pt file (dict with 'coords' key)
pt_path = os.path.join(temp_dir, 'voxels.pt')
pt_data = {'coords': torch.randn(2, 4)}
torch.save(pt_data, pt_path)

# Test with .pth file (dict with 'voxels' key)
pth_path = os.path.join(temp_dir, 'voxels.pth')
pth_data = {'voxels': torch.randn(2, 4)}
torch.save(pth_data, pth_path)

# Use the .npy path as the primary test case
ARGS = {
    'path': npy_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
parse_camera_position Ran
ran, honoured the contract we drafted
TeCai/DVD/app_dvd_image.py
code served (permissive licence) · get_code("2b3eb28e2f587267")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

# The function parse_camera_position takes a single string parameter
# and parses it as three comma-separated floats representing camera position coordinates

ARGS = {
    'value': '-180,90,3'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
parse_range Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
TeCai/DVD/examples_dvd_text_editing_api.py
code served (permissive licence) · get_code("c67209bea664e4d4")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# parse_range expects:
# - values: list[int] - a list of two integers
# - resolution: int - an integer representing the resolution

ARGS = {
    "values": [10, 50],
    "resolution": 100
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family int — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
slugify Ran
ran; our contract draft was wrong, not the code
TeCai/DVD/examples_dvd_text_editing_api.py
code served (permissive licence) · get_code("3d1dc67364d11f73")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import re

def slugify(text: str, max_length: int=48) -> str:
    slug = re.sub('[^a-zA-Z0-9]+', '_', text.lower()).strip('_')
    return slug[:max_length].strip('_') or 'text_prompt'

ARGS = {
    'text': 'Hello World! This is a test prompt with special @#$ characters',
    'max_length': 48
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
text_cfg_schedule Ran
ran, honoured the contract we drafted
TeCai/DVD/trellis/pipelines/dvd_voxel.py
code served (permissive licence) · get_code("66c05bec38f964d9")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned float, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

ARGS = {
    "t": 0.5
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_aesthetic_model Not yet run
unverified (harvested, no recorded run)
TeCai/DVD/dataset_toolkits/calculate_aesthetic_scores.py
code served (permissive licence) · get_code("89a6971a43060990")
How it ran

no invocation recorded.

load_dvd_denoiser Not yet run
unverified (harvested, no recorded run)
TeCai/DVD/trellis/pipelines/dvd_voxel.py
code served (permissive licence) · get_code("82fa2aaf8ab4cc11")
How it ran

no invocation recorded.

need_process Not yet run
unverified (harvested, no recorded run)
TeCai/DVD/dataset_toolkits/build_metadata.py
code served (permissive licence) · get_code("ae3ac644f6a269e7")
How it ran

no invocation recorded.

Also present in other papers' code

Of 13 samples lifted from this paper's linked repositories, 13 carry a function identity and 0 do not. Of the 13 distinct functions with an identity, 3 are also present in other papers whose code we hold (3 with at least one paper dated earlier, 0 the same month, 0 later) and 10 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
get_first_directory 2 other papers
2 dated earlier
2506.01853 (2025-06)
ShapeLLM-Omni: A Native Multimodal LLM for 3D Generation and Understanding
2507.12465 (2025-07)
PhysX-3D: Physical-Grounded 3D Asset Generation
need_process 2 other papers
2 dated earlier
2506.01853 (2025-06)
ShapeLLM-Omni: A Native Multimodal LLM for 3D Generation and Understanding
2507.12465 (2025-07)
PhysX-3D: Physical-Grounded 3D Asset Generation
get_aesthetic_model 1 other paper
1 dated earlier
2411.15262 (2024-11)
MovieBench: A Hierarchical Movie Level Dataset for Long Video Generation

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

We introduce Discrete Voxel Diffusion (DVD), a discrete diffusion framework to generate, assess, and edit sparse voxels for SLAT (Structured LATent) based 3D generative pipelines. Although discrete diffusion has not generally displaced continuous diffusion in image-like generation, we show that it can be an effective first-stage prior for sparse voxel scaffolds. By treating voxel occupancy as a native discrete variable, DVD avoids continuous-to-discrete thresholding and provides a simple framework for voxel generation, uncertainty estimation, and editing. Beyond quality gains, DVD provides more interpretable generation dynamics through explicit categorical modeling. Furthermore, we leverage the predictive entropy as a robust uncertainty metric to identify ambiguous voxel regions and complicated samples, facilitating tasks such as data filtering and quality assessment. Finally, we propose a lightweight fine-tuning strategy using block-structured perturbation patterns. This approach empowers the model to inpaint and edit voxels within a single sampling round, requiring negligible auxiliary computation and no additional model evaluations. Code is available at https://github.com/TeCai/DVD.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2605.07971")
get_code_for_paper("2605.07971")
have("2605.07971")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2605.07971.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2605.07971.svg)](https://syntology.ai/paper/2605.07971)

Connect an agent — have() is free.