SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2410.22366 · 2024

One-Step is Enough: Sparse Autoencoders for Text-to-Image Diffusion Models

arXiv · PDF · Open in the Atlas

Code that ran

2 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; 2 failures of our instrument; 2 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/2410.22366.json · badge: /api/ran/2410.22366.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
surkovv/sdxl-unbox canonical (classifier label)
classifier basis: census
2 ran · 2 no recorded run
FunctionStatusWhere it lives
add_feature_on_area_base Ran
ran; our contract draft was wrong, not the code
surkovv/sdxl-unbox/utils/hooks.py
code served (permissive licence) · get_code("2cd254cbecb4c027")
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 5, 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
import torch.nn as nn

torch.manual_seed(0)

# Create a minimal SAE (Sparse Autoencoder) module
class MinimalSAE(nn.Module):
    def __init__(self, input_dim=4, latent_dim=8):
        super().__init__()
        self.encoder = nn.Linear(input_dim, latent_dim)
        self.decoder = nn.Linear(latent_dim, input_dim)
        self.device = torch.device('cpu')
    
    def encode(self, x):
        # x shape: (batch, height, width, channels)
        batch, h, w, c = x.shape
        x_flat = x.reshape(batch * h * w, c)
        encoded = self.encoder(x_flat)
        return encoded.reshape(batch, h, w, -1)

sae = MinimalSAE(input_dim=4, latent_dim=8)

# Create feature index
feature_idx = 3

# Create activation_map (2D or 3D tensor)
# Based on code: can be 2D and gets unsqueezed to 3D, or already 3D
activation_map = torch.rand(3, 5)  # 2D map

# Create a minimal module (not actually used in the function body)
module = nn.Identity()

# Create input and output
# input and output are tuples where [0] is a tensor
# The function expects output[0] to be indexable with [0] and [1]
# and to have shape (batch, channels, height, width) after permutation back
input_tensor = torch.rand(2, 4, 3, 5)  # (batch=2, channels=4, height=3, width=5)
output_tensor = torch.rand(2, 4, 3, 5)  # same shape

input_tuple = (input_tensor,)
output_tuple = (output_tensor,)

ARGS = {
    'sae': sae,
    'feature_idx': feature_idx,
    'activation_map': activation_map,
    'module': module,
    'input': input_tuple,
    'output': output_tuple
}
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 5, 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
retrieve Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
surkovv/sdxl-unbox/SDLens/hooked_sd_pipeline.py
code served (permissive licence) · get_code("8b3022444bceb9b4")
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

torch.manual_seed(0)

# The function accepts 'io' parameter which can be:
# 1. A tuple with exactly 1 element
# 2. A torch.Tensor
# We'll test with a tensor case

ARGS = {
    'io': torch.randn(2, 4)
}
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
add_feature Not yet run
unverified (harvested, no recorded run)
surkovv/sdxl-unbox/utils/hooks.py
code served (permissive licence) · get_code("809505b9f5ea00b0")
How it ran

no invocation recorded.

add_feature_on_area_base_both Not yet run
unverified (harvested, no recorded run)
surkovv/sdxl-unbox/utils/hooks.py
code served (permissive licence) · get_code("74d970e267d552a1")
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, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 4 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.

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

For large language models (LLMs), sparse autoencoders (SAEs) have been shown to decompose intermediate representations that often are not interpretable directly into sparse sums of interpretable features, facilitating better control and subsequent analysis. However, similar analyses and approaches have been lacking for text-to-image models. We investigate the possibility of using SAEs to learn interpretable features for SDXL Turbo, a few-step text-to-image diffusion model. To this end, we train SAEs on the updates performed by transformer blocks within SDXL Turbo's denoising U-net in its 1-step setting. Interestingly, we find that they generalize to 4-step SDXL Turbo and even to the multi-step SDXL base model (i.e., a different model) without additional training. In addition, we show that their learned features are interpretable, causally influence the generation process, and reveal specialization among the blocks. We do so by creating RIEBench, a representation-based image editing benchmark, for editing images while they are generated by turning on and off individual SAE features. This allows us to track which transformer blocks' features are the most impactful depending on the edit category. Our work is the first investigation of SAEs for interpretability in text-to-image diffusion models and our results establish SAEs as a promising approach for understanding and manipulating the internal mechanisms of text-to-image models.

For agents

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

get_harvested_code_for_paper("2410.22366")
get_code_for_paper("2410.22366")
have("2410.22366")

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

curl https://syntology.ai/api/ran/2410.22366.json

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

[![Syntology run record](https://syntology.ai/api/ran/2410.22366.svg)](https://syntology.ai/paper/2410.22366)

Connect an agent — have() is free.