SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2410.08709 · ICML · 2025

Distillation of Discrete Diffusion through Dimensional Correlations

Masaaki Imaizumi, Satoshi Hayakawa, Yuhta Takida, Yuki Mitsufuji, Hiromi Wakaki

arXiv · PDF · Open in the Atlas

Code that ran

6 of 7 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; 1 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 5 failures of our instrument; 1 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/2410.08709.json · badge: /api/ran/2410.08709.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
sony/di4c canonical (classifier label)
classifier basis: census
6 ran · 1 no recorded run
FunctionStatusWhere it lives
custom_lamda Ran
ran, honoured the contract we drafted
sony/di4c/maskgit-pytorch/Trainer/trainer.py
code served (permissive licence) · get_code("9ab699e19259b9d3")
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 (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
import numpy as np

torch.manual_seed(0)

# custom_lamda takes parameter x and returns x[:5]
# x needs to support slicing with [:5], so it should be a sequence/tensor/array
# Creating a simple tensor with more than 5 elements to make slicing meaningful

x = torch.randn(10, 4)  # Shape: (10, 4) - will slice to (5, 4)

ARGS = {
    'x': x
}
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 (5, 4) — shape (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
eval_loop Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
sony/di4c/sdtt/src/sdtt/harness_gpt2.py
pointer only (licence: MIT); the stored copy get_code serves, from jdeschena/sdtt, is pointer only (licence: NONE) · get_code("b0a49b09dfb49452")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from jdeschena/sdtt, and GitHub detected no licence in that repository 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. A run is recorded.

get_obj_from_str Ran
ran; our contract draft was wrong, not the code
sony/di4c/maskgit-pytorch/Network/Taming/util.py
code served (permissive licence) · get_code("221b2d116fdf1032")
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) and returned type. 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 importlib
import torch
import torch.nn as nn

torch.manual_seed(0)

# The function get_obj_from_str takes a string like "module.path.ClassName"
# and returns the class/object by:
# 1. Splitting on the last dot to get module path and class name
# 2. Optionally reloading the module if reload=True
# 3. Importing the module and getting the attribute (class/object)

# We need to provide:
# - string: a valid module.class path that exists
# - reload: boolean flag

# Using a built-in class that's guaranteed to exist
string_value = "torch.nn.Linear"
reload_value = False

ARGS = {
    "string": string_value,
    "reload": reload_value,
}
Setup block drafted by
a language model, Claude Haiku 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 type — a type output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a type 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
md5_hash Ran
ran; our contract draft was wrong, not the code
sony/di4c/maskgit-pytorch/Network/Taming/util.py
pointer only (licence: MIT); the stored copy get_code serves, from zh460045050/V2L-Tokenizer, is pointer only (licence: NONE) · get_code("07d90bd6ae1be7db")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from zh460045050/V2L-Tokenizer, and GitHub detected no licence in that repository 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. A run is recorded.

remap_image_torch Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
sony/di4c/maskgit-pytorch/Metrics/sample_and_eval.py
code served (permissive licence) · get_code("78d02565c457a57b")
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 int, 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)

# Create a small test image tensor
# Shape: (batch_size=2, channels=3, height=4, width=5)
# The function calls min/max operations on dimension -1 three times,
# then reshapes to (batch, 1, 1, 1), so we need a 4D tensor
image = torch.randn(2, 3, 4, 5, dtype=torch.float32)

ARGS = {
    "image": image
}
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 int, 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
transformer_timestep_embedding Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
sony/di4c/maskgit-pytorch/Network/transformer.py
code served (permissive licence) · get_code("d0434d43396aa8fb")
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 math

torch.manual_seed(0)

# Construct arguments for transformer_timestep_embedding
timesteps = torch.tensor([0, 5], dtype=torch.long)  # batch size 2
embedding_dim = 4
max_positions = 10000

ARGS = {
    "timesteps": timesteps,
    "embedding_dim": embedding_dim,
    "max_positions": max_positions,
}
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
instantiate_from_config Not yet run
unverified (harvested, no recorded run)
sony/di4c/maskgit-pytorch/Network/Taming/util.py
code served (permissive licence) · get_code("ef52d142b2c30ac2")
How it ran

no invocation recorded.

Also present in other papers' code

Of 7 samples lifted from this paper's linked repositories, 7 carry a function identity and 0 do not. Of the 7 distinct functions with an identity, 5 are also present in other papers whose code we hold (4 with at least one paper dated earlier, 4 the same month, 4 later) and 2 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_obj_from_str 54 other papers
39 dated earlier, 2 the same month, 12 later, 1 same year, order not determinable · 12 only as an assembly brick
2006.11239 (2020-06)
Denoising Diffusion Probabilistic Models
2102.05379 (2021-02)
Argmax Flows and Multinomial Diffusion: Learning Categorical Distributions
2112.00384 (2021-12)
Exploration into Translation-Equivariant Image Quantization
and 51 more not named here
md5_hash 30 other papers
23 dated earlier, 1 the same month, 5 later, 1 same year, order not determinable
2008.01777 (2020-08)
Making Sense of CNNs: Interpreting Deep Representations & Their Invariances with INNs
2104.07652 (2021-04)
Geometry-Free View Synthesis: Transformers and no 3D Priors
2112.00384 (2021-12)
Exploration into Translation-Equivariant Image Quantization
and 27 more not named here
instantiate_from_config 24 other papers
20 dated earlier, 4 later · 7 only as an assembly brick
2006.11239 (2020-06)
Denoising Diffusion Probabilistic Models
2102.05379 (2021-02)
Argmax Flows and Multinomial Diffusion: Learning Categorical Distributions
2206.02262 (2022-06)
Diffusion-GAN: Training GANs with Diffusion
and 21 more not named here
transformer_timestep_embedding 4 other papers
1 dated earlier, 1 the same month, 2 later · 1 only as an assembly brick
2402.04997 (2024-02)
Generative Flows on Discrete State-Spaces: Enabling Multimodal Flows with Applications to Protein Co-Design
2410.06264 (2024-10)
Think While You Generate: Discrete Diffusion with Planned Denoising
2412.10193 (2024-12)
Simple Guidance Mechanisms for Discrete Diffusion Models
and 1 more not named here
eval_loop 1 other paper
1 the same month
2410.21035 (2024-10)
Beyond Autoregression: Fast LLMs via Self-Distillation Through Time

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

Diffusion models have demonstrated exceptional performances in various fields of generative modeling, but suffer from slow sampling speed due to their iterative nature. While this issue is being addressed in continuous domains, discrete diffusion models face unique challenges, particularly in capturing dependencies between elements (e.g., pixel relationships in image, sequential dependencies in language) mainly due to the computational cost of processing high-dimensional joint distributions. In this paper, (i) we propose "mixture" models for discrete diffusion that are capable of treating dimensional correlations while remaining scalable, and (ii) we provide a set of loss functions for distilling the iterations of existing models. Two primary theoretical insights underpin our approach: First, conventional models with element-wise independence can well approximate the data distribution, but essentially require many sampling steps. Second, our loss functions enable the mixture models to distill such manystep conventional models into just a few steps by learning the dimensional correlations. Our experimental results show the effectiveness of the proposed method in distilling pretrained discrete diffusion models across image and language domains. The code used in the paper is available at https://github.com/sony/di4c.

For agents

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

get_harvested_code_for_paper("2410.08709")
get_code_for_paper("2410.08709")
have("2410.08709")

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

curl https://syntology.ai/api/ran/2410.08709.json

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

[![Syntology run record](https://syntology.ai/api/ran/2410.08709.svg)](https://syntology.ai/paper/2410.08709)

Connect an agent — have() is free.