SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2206.00364 · NeurIPS · 2022

Elucidating the Design Space of Diffusion-Based Generative Models

Tero Karras, Timo Aila, Nvidia Miika, Aittala Nvidia, Nvidia Samuli, Laine Nvidia

arXiv · PDF · Open in the Atlas

Code that ran

16 of 21 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; 7 honoured the contract we drafted, 0 violated it and 5 have no contract verdict; 4 failures of our instrument; 5 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; 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/2206.00364.json · badge: /api/ran/2206.00364.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
copy not recorded community
not in the paper's declared repositories
5 ran · 3 no recorded run
friedmanroy/ctrl-shift community (classifier label)
classifier basis not recorded on the link
5 ran · 0 no recorded run
pals-ttic/sjc extension (classifier label)
classifier basis: classifier_2025
3 ran · 0 no recorded run
crowsonkb/k-diffusion extension (classifier label)
classifier basis: classifier_2025
2 ran · 1 no recorded run
plai-group/vcdm extension (classifier label)
classifier basis: classifier_2025
1 ran · 0 no recorded run
christopher-beckham/toy-edm reimplementation (classifier label)
classifier basis: classifier_2025
0 ran · 1 no recorded run
FunctionStatusWhere it lives
betas_for_alpha_bar Ran
ran, honoured the contract we drafted
this paper's copy was not recorded; identical code first harvested from NVlabs/DiffPure
pointer only; the stored copy get_code serves, from NVlabs/DiffPure, is pointer only (licence: NOASSERTION) · get_code("2ab2316ac6fdd869")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

calculate_fid_from_inception_stats Ran
ran, honoured the contract we drafted
this paper's copy was not recorded; identical code first harvested from sangyun884/fast-ode
pointer only; the stored copy get_code serves, from sangyun884/fast-ode, is pointer only (licence: NONE) · get_code("45e941764344e7f3")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

calculate_fid_from_inception_stats Ran
ran, honoured the contract we drafted
plai-group/vcdm/fid.py
pointer only (licence: NOASSERTION) · get_code("5a0fd9882013c6a5")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, plai-group/vcdm (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

downscale_pos Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
crowsonkb/k-diffusion/k_diffusion/models/image_transformer_v2.py
code served (permissive licence) · get_code("ce526ffedb8faa2e")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Sonnet 4.5, prompt python_fixture_v1) and returned Tensor, ndim 4, dtype family float, shape (2, 3, 5, 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 torch

torch.manual_seed(0)

# The function expects pos with shape (..., h*nh, w*nw, e)
# where nh=2 and nw=2 are the downscaling factors
# Let's create a simple case with:
# - batch dimension: 2
# - h=3, nh=2 -> height dimension is 6
# - w=5, nw=2 -> width dimension is 10
# - e=4 (embedding dimension)

pos = torch.randn(2, 6, 10, 4)

ARGS = {
    'pos': pos
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
Imports
einops, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 4, dtype family float, shape (2, 3, 5, 4) — shape (2, 3, 5, 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
edm_sample Ran
ran on a synthesized input
friedmanroy/ctrl-shift/sampling.py
code served (permissive licence) · get_code("34c52bd69ff1a403")
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. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 torch
import torch.nn as nn

torch.manual_seed(0)

class MockScoreNetwork(nn.Module):
    """Mock score network that satisfies the interface used in edm_sample"""
    def __init__(self):
        super().__init__()
        self.sigma_min = 0.002
        self.sigma_max = 80.0
        self.linear = nn.Linear(4, 4)
    
    def forward(self, x, t, class_labels=None):
        # x: [N, 4] or similar
        # t: scalar tensor
        # class_labels: optional tensor
        # Returns denoised output of same shape as x
        return self.linear(x)
    
    def round_sigma(self, sigma):
        # Takes a sigma value and returns it rounded
        return sigma

# Create the network
net = MockScoreNetwork()
net.eval()

# Create latents: batch size 2, 4 channels
latents = torch.randn(2, 4)

# Create class labels: batch size 2
class_labels = torch.tensor([0, 1])

ARGS = {
    'net': net,
    'latents': latents,
    'class_labels': class_labels,
    'num_steps': 4,
    'sigma_min': 0.002,
    'sigma_max': 80.0,
    'rho': 7.0,
    'verbose': False,
    'invert': False,
    'euler': False,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, tqdm — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is 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
edm_schedule Ran
ran on a synthesized input
friedmanroy/ctrl-shift/sampling.py
code served (permissive licence) · get_code("c0c71f52be7b3680")
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 1, dtype family float, shape (11,) as a later run recorded it. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 = {
    "num_steps": 10,
    "sigma_min": 0.002,
    "sigma_max": 80.0,
    "rho": 7.0,
    "device": "cpu"
}
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 Tensor, ndim 1, dtype family float, shape (11,) — shape (11,) 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
extend_coords Ran
ran on a synthesized input
friedmanroy/ctrl-shift/generate_dist_shift.py
code served (permissive licence) · get_code("206dcd48bf0f2883")
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 4, dtype family float, shape (2, 3, 4, 5) 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)

ARGS = {
    'shape': (2, 3, 4, 5),
    'value': 75,
    'target': None
}
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 4, dtype family float, shape (2, 3, 4, 5) — shape (2, 3, 4, 5) 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_denoised Ran
ran; our contract draft was wrong, not the code
this paper's copy was not recorded; identical code first harvested from onefly123/pas
pointer only · get_code("3eed7631aeebeabc")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

get_named_beta_schedule Ran
ran, honoured the contract we drafted
this paper's copy was not recorded; identical code first harvested from NVlabs/DiffPure
pointer only; the stored copy get_code serves, from NVlabs/DiffPure, is pointer only (licence: NOASSERTION) · get_code("a086d6286a40b889")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

karras_t_schedule Ran
ran, honoured the contract we drafted
pals-ttic/sjc/adapt.py
pointer only (licence: NOASSERTION) · get_code("8285678722ad9b94")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, pals-ttic/sjc (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

overlap_coords Ran
ran on a synthesized input
friedmanroy/ctrl-shift/generate_dist_shift.py
code served (permissive licence) · get_code("83cbd2453e3efe71")
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 4, dtype family float, shape (2, 4, 3, 5) 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)

ARGS = {
    "shape": (2, 4, 3, 5),
    "value": 75,
}
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 4, dtype family float, shape (2, 4, 3, 5) — shape (2, 4, 3, 5) 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_int_list Ran
ran, honoured the contract we drafted
this paper's copy was not recorded; identical code first harvested from newbeeer/stf
pointer only; the stored copy get_code serves, from newbeeer/stf, is pointer only (licence: NONE) · get_code("cacd4f6ec202d9b4")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

power_schedule Ran
ran, honoured the contract we drafted
pals-ttic/sjc/adapt.py
pointer only (licence: NOASSERTION) · get_code("c5ab2a66337ec866")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, pals-ttic/sjc (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

trunc_coords Ran
ran on a synthesized input
friedmanroy/ctrl-shift/generate_dist_shift.py
code served (permissive licence) · get_code("a049710a973819ad")
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 4, dtype family float, shape (2, 3, 4, 5) 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 = {
    "shape": (2, 3, 4, 5),
    "trunc": 50
}
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 Tensor, ndim 4, dtype family float, shape (2, 3, 4, 5) — shape (2, 3, 4, 5) 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
tsr_stats Ran
ran; our contract draft was wrong, not the code
pals-ttic/sjc/run_sjc.py
pointer only (licence: NOASSERTION) · get_code("f3cf3784fb4488db")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, pals-ttic/sjc (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

zero_init Ran
ran; our contract draft was wrong, not the code
crowsonkb/k-diffusion/k_diffusion/models/image_transformer_v2.py
code served (permissive licence) · get_code("d63461a574db572f")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Sonnet 4.5, prompt python_fixture_v1) and returned Linear. 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 torch
from torch import nn

torch.manual_seed(0)

# Create a simple linear layer with weight and bias
layer = nn.Linear(in_features=4, out_features=3)

ARGS = {
    'layer': layer
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Linear — a Linear output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Linear 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
ablation_sampler Not yet run
unverified (harvested, no recorded run)
this paper's copy was not recorded; identical code first harvested from newbeeer/stf
pointer only; the stored copy get_code serves, from newbeeer/stf, is pointer only (licence: NONE) · get_code("daa5186ef7b993ca")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

ablation_sampler Not yet run
unverified (harvested, no recorded run)
this paper's copy was not recorded; identical code first harvested from ghliu/mdm
pointer only · get_code("9fbf646e18eed1db")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

checkpoint Not yet run
unverified (harvested, no recorded run)
crowsonkb/k-diffusion/k_diffusion/models/image_transformer_v2.py
code served (permissive licence) · get_code("bed894f151989fee")
How it ran

no invocation recorded.

edm_generalised_stochastic_sampler Not yet run
unverified (harvested, no recorded run)
christopher-beckham/toy-edm/src/edm_generalised_stochastic_sampler.py
pointer only (licence: NONE) · get_code("16450717abebe540")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, christopher-beckham/toy-edm when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

edm_sampler Not yet run
unverified (harvested, no recorded run)
this paper's copy was not recorded; identical code first harvested from ghliu/mdm
pointer only · get_code("8b16b0674af254db")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

Also present in other papers' code

Of 21 samples lifted from this paper's linked repositories, 21 carry a function identity and 0 do not. Of the 21 distinct functions with an identity, 12 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 2 the same month, 12 later) and 9 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
betas_for_alpha_bar 92 other papers
11 dated earlier, 2 the same month, 79 later · 3 only as an assembly brick
2006.11239 (2020-06)
Denoising Diffusion Probabilistic Models
2010.02502 (2020-10)
Denoising Diffusion Implicit Models
2102.09672 (2021-02)
Improved Denoising Diffusion Probabilistic Models
and 89 more not named here
get_named_beta_schedule 63 other papers
10 dated earlier, 2 the same month, 51 later
2010.02502 (2020-10)
Denoising Diffusion Implicit Models
2108.02938 (2021-08)
ILVR: Conditioning Method for Denoising Diffusion Probabilistic Models
2111.11672 (2021-11)
Few-shot Image Generation with Mixup-based Distance Learning
and 60 more not named here
parse_int_list 37 other papers
37 later
2302.00670 (2023-02)
Stable Target Field for Reduced Variance Score Estimation in Diffusion Models
2302.04265 (2023-02)
PFGM++: Unlocking the Potential of Physics-Inspired Generative Models
2303.10137 (2023-03)
A Recipe for Watermarking Diffusion Models
and 34 more not named here
calculate_fid_from_inception_stats 23 other papers
23 later
2211.17091 (2022-11)
Refining Generative Process with Discriminator Guidance in Score-based Diffusion Models
2301.12003 (2023-01)
Minimizing Trajectory Curvature of ODE-based Generative Models
2302.09057 (2023-02)
Consistent Diffusion Models: Mitigating Sampling Drift by Learning to be Consistent
and 20 more not named here
ablation_sampler 12 other papers
12 later
2207.12598 (2022-07)
Classifier-Free Diffusion Guidance
2303.10137 (2023-03)
A Recipe for Watermarking Diffusion Models
2310.01236 (2023-10)
Mirror Diffusion Models for Constrained and Watermarked Generation
and 9 more not named here
edm_sampler 11 other papers
11 later
2301.11445 (2023-01)
3DShape2VecSet: A 3D Shape Representation for Neural Fields and Generative Diffusion Models
2303.10137 (2023-03)
A Recipe for Watermarking Diffusion Models
2306.06991 (2023-06)
Fast Diffusion Model
and 8 more not named here
get_denoised 5 other papers
5 later · 2 only as an assembly brick
2312.00094 (2023-12)
Fast ODE-based Sampling for Diffusion Models in Around 5 Steps
2409.19681 (2024-09)
Simple and Fast Distillation of Diffusion Models
2411.06503 (2024-11)
Diffusion Sampling Correction via Approximately 10 Parameters
and 2 more not named here
karras_t_schedule 3 other papers
3 later
2212.00774 (2022-12)
Score Jacobian Chaining: Lifting Pretrained 2D Diffusion Models for 3D Generation
2304.10261 (2023-04)
Anything-3D: Towards Single-view Anything Reconstruction in the Wild
2402.18842 (2024-02)
ViewFusion: Towards Multi-View Consistency via Interpolated Denoising
power_schedule 3 other papers
3 later
2212.00774 (2022-12)
Score Jacobian Chaining: Lifting Pretrained 2D Diffusion Models for 3D Generation
2304.10261 (2023-04)
Anything-3D: Towards Single-view Anything Reconstruction in the Wild
2402.18842 (2024-02)
ViewFusion: Towards Multi-View Consistency via Interpolated Denoising
tsr_stats 3 other papers
3 later
2212.00774 (2022-12)
Score Jacobian Chaining: Lifting Pretrained 2D Diffusion Models for 3D Generation
2303.11328 (2023-03)
Zero-1-to-3: Zero-shot One Image to 3D Object
2304.10261 (2023-04)
Anything-3D: Towards Single-view Anything Reconstruction in the Wild
ablation_sampler 2 other papers
2 later
2302.00670 (2023-02)
Stable Target Field for Reduced Variance Score Estimation in Diffusion Models
2302.04265 (2023-02)
PFGM++: Unlocking the Potential of Physics-Inspired Generative Models
zero_init 1 other paper
1 later
2412.03439 (2024-12)
CleanDIFT: Diffusion Features without Noise

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 argue that the theory and practice of diffusion-based generative models are currently unnecessarily convoluted and seek to remedy the situation by presenting a design space that clearly separates the concrete design choices. This lets us identify several changes to both the sampling and training processes, as well as preconditioning of the score networks. Together, our improvements yield new state-of-the-art FID of 1.79 for CIFAR-10 in a class-conditional setting and 1.97 in an unconditional setting, with much faster sampling (35 network evaluations per image) than prior designs. To further demonstrate their modular nature, we show that our design changes dramatically improve both the efficiency and quality obtainable with pre-trained score networks from previous work, including improving the FID of a previously trained ImageNet-64 model from 2.07 to near-SOTA 1.55, and after re-training with our proposed improvements to a new SOTA of 1.36.

For agents

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

get_harvested_code_for_paper("2206.00364")
get_code_for_paper("2206.00364")
have("2206.00364")

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

curl https://syntology.ai/api/ran/2206.00364.json

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

[![Syntology run record](https://syntology.ai/api/ran/2206.00364.svg)](https://syntology.ai/paper/2206.00364)

Connect an agent — have() is free.