SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2310.05264 · NeurIPS · 2023

The Emergence of Reproducibility and Generalizability in Diffusion Models

Peng Wang, Qing Qu, Liyue Shen, Huijie Zhang, Jinfan Zhou, Yifu Lu, Minzhe Guo

arXiv · PDF · Open in the Atlas

Code that ran

3 of 4 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; 0 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 3 failures of our instrument; 1 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/2310.05264.json · badge: /api/ran/2310.05264.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
huijieZH/Diffusion-Model-Generalizability canonical (classifier label)
classifier basis: classifier_2025
3 ran · 1 no recorded run
FunctionStatusWhere it lives
TransformSamples Ran
ran; our contract draft was wrong, not the code
huijieZH/Diffusion-Model-Generalizability/edm/sscd.py
code served (permissive licence) · get_code("640adf50ba465604")
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 torch
import torch.nn as nn

torch.manual_seed(0)

# Create simple sample data - list of small tensors
samples = [
    torch.randn(4),
    torch.randn(4),
]

# Create a simple transform function that takes a tensor and returns a tensor
class SimpleTransform(nn.Module):
    def forward(self, x):
        return x * 2.0

transform = SimpleTransform()

ARGS = {
    "samples": samples,
    "transform": transform,
}
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 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
get_exp_bias_batch Ran
ran; our contract draft was wrong, not the code
huijieZH/Diffusion-Model-Generalizability/edm/generate.py
code served (permissive licence) · get_code("553a32f93376dc01")
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 max. 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

torch.manual_seed(0)

# Analyze the function:
# - x: tensor, shape (bs_x, ...) - used with unsqueeze(1) and arithmetic
# - y_batch: tensor, shape (bs_y, ...) - used in arithmetic with x
# - s: scalar or broadcastable - multiplied with y_batch
# - std: scalar - used in division (std ** 2)

# The function computes:
# xb = x.unsqueeze(1)  -> shape (bs_x, 1, ...)
# (xb - s * y_batch) ** 2  -> broadcasted to (bs_x, bs_y, ...)
# .view(bs_x, bs_y, -1).sum(dim=-1)  -> shape (bs_x, bs_y)
# .max(dim=1)  -> returns (values, indices)

# Create small tensors with distinct dimensions
bs_x = 2
bs_y = 3
feature_dim = 4

x = torch.randn(bs_x, feature_dim)  # shape (2, 4)
y_batch = torch.randn(bs_y, feature_dim)  # shape (3, 4)
s = 0.5  # scalar multiplier
std = 1.0  # scalar for normalization

ARGS = {
    "x": x,
    "y_batch": y_batch,
    "s": s,
    "std": std
}
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 max — a max output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a max output the recorded ndim 2 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
normal_distribution_batch Ran
ran; our contract draft was wrong, not the code
huijieZH/Diffusion-Model-Generalizability/edm/generate.py
code served (permissive licence) · get_code("7effd95926009645")
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. 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

torch.manual_seed(0)

# Analyzing the function:
# - x: tensor, unsqueezed to (bs_x, 1), used in arithmetic with y_batch
# - y_batch: tensor with shape (bs_y, ...), used in arithmetic and has .shape[0]
# - s: scalar, used in multiplication
# - std: scalar, used in division
# - bias: tensor (default 0), has .unsqueeze(1) called on it

# The function performs operations like:
# - xb - s * y_batch (broadcasting)
# - bias.unsqueeze(1)
# So bias should be a tensor with at least 1 dimension

# Create small tensors with distinct batch sizes
bs_x = 2  # batch size for x
bs_y = 3  # batch size for y_batch
feature_dim = 4  # feature dimension

x = torch.randn(bs_x, feature_dim)  # shape: (2, 4)
y_batch = torch.randn(bs_y, feature_dim)  # shape: (3, 4)
s = 0.5  # scalar scaling factor
std = 1.0  # standard deviation
bias = torch.zeros(bs_x)  # shape: (2,) - matches bs_x for unsqueeze(1) to work

ARGS = {
    "x": x,
    "y_batch": y_batch,
    "s": s,
    "std": std,
    "bias": bias,
}
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 — 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_sampler Not yet run
unverified (harvested, no recorded run)
huijieZH/Diffusion-Model-Generalizability/edm/generate.py
code served (permissive licence) · get_code("00dcc1be26a2af0f")
How it ran

no invocation recorded.

Also present in other papers' code

Of 4 samples lifted from this paper's linked repositories, 4 carry a function identity and 0 do not. Of the 4 distinct functions with an identity, 4 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 4 later) and 0 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
TransformSamples 1 other paper
1 later
2409.02426 (2024-09)
Breaking the Curse of Dimensionality: Diffusion Models Efficiently Learn Low-Dimensional Distributions
edm_sampler 1 other paper
1 later
2409.02426 (2024-09)
Breaking the Curse of Dimensionality: Diffusion Models Efficiently Learn Low-Dimensional Distributions
get_exp_bias_batch 1 other paper
1 later
2409.02426 (2024-09)
Breaking the Curse of Dimensionality: Diffusion Models Efficiently Learn Low-Dimensional Distributions
normal_distribution_batch 1 other paper
1 later
2409.02426 (2024-09)
Breaking the Curse of Dimensionality: Diffusion Models Efficiently Learn Low-Dimensional Distributions

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

In this work, we investigate an intriguing and prevalent phenomenon of diffusion models which we term as "consistent model reproducibility": given the same starting noise input and a deterministic sampler, different diffusion models often yield remarkably similar outputs. We confirm this phenomenon through comprehensive experiments, implying that different diffusion models consistently reach the same data distribution and score function regardless of diffusion model frameworks, model architectures, or training procedures. More strikingly, our further investigation implies that diffusion models are learning distinct distributions affected by the training data size. This is supported by the fact that the model reproducibility manifests in two distinct training regimes: (i) "memorization regime," where the diffusion model overfits to the training data distribution, and (ii) "generalization regime," where the model learns the underlying data distribution. Our study also finds that this valuable property generalizes to many variants of diffusion models, including those for conditional generation, solving inverse problems, and model fine-tuning. Finally, our work raises numerous intriguing theoretical questions for future investigation and highlights practical implications regarding training efficiency, model privacy, and the controlled generation of diffusion models.

For agents

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

get_harvested_code_for_paper("2310.05264")
get_code_for_paper("2310.05264")
have("2310.05264")

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

curl https://syntology.ai/api/ran/2310.05264.json

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

[![Syntology run record](https://syntology.ai/api/ran/2310.05264.svg)](https://syntology.ai/paper/2310.05264)

Connect an agent — have() is free.