SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2502.04320 · ICML · 2025

ConceptAttention: Diffusion Transformers Learn Highly Interpretable Features

Chau, Benjamin Hoover, Pinar Yanardag, Alec Helbling, Tuna Han, Salih Meral, Duen Horng

arXiv · PDF · Open in the Atlas

Code that ran

4 of 9 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (2 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 4 have no contract verdict; 0 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/2502.04320.json · badge: /api/ran/2502.04320.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
helblazer811/ConceptAttention canonical (classifier label)
classifier basis: classifier_2025
4 ran · 5 no recorded run
FunctionStatusWhere it lives
Modulation Ran
ran on a synthesized input
helblazer811/ConceptAttention/concept_attention/flux/dit_block.py
pointer only (licence: NONE) · get_code("68e13022e7ff1c3e")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, helblazer811/ConceptAttention 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.

ModulationOut Ran
ran on a synthesized input · constructed an object; no result was computed
helblazer811/ConceptAttention/concept_attention/flux/dit_block.py
pointer only (licence: NONE) · get_code("a9abef6f4da5afe3")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, helblazer811/ConceptAttention 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.

QKNorm Ran
ran on a synthesized input
helblazer811/ConceptAttention/concept_attention/flux/dit_block.py
pointer only (licence: NONE) · get_code("2f43209f61e8219e")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, helblazer811/ConceptAttention 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.

RMSNorm Ran
ran on a synthesized input · constructed an object; no result was computed
helblazer811/ConceptAttention/concept_attention/flux/dit_block.py
pointer only (licence: NONE) · get_code("5c52ab157dc1c15c")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, helblazer811/ConceptAttention 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.

ModifiedDoubleStreamBlock Not yet run
unverified (harvested, no recorded run)
helblazer811/ConceptAttention/concept_attention/flux/dit_block.py
pointer only (licence: NONE) · get_code("54f6ec9f0d9af990")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, helblazer811/ConceptAttention 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.

SelfAttention Not yet run
unverified (harvested, no recorded run)
helblazer811/ConceptAttention/concept_attention/flux/dit_block.py
pointer only (licence: NONE) · get_code("44b8130635f7bada")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, helblazer811/ConceptAttention 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.

apply_rope Not yet run
unverified (harvested, no recorded run)
helblazer811/ConceptAttention/concept_attention/flux/dit_block.py
pointer only (licence: NONE) · get_code("7f1b2506d32e33e3")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, helblazer811/ConceptAttention 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.

attention Not yet run
unverified (harvested, no recorded run)
helblazer811/ConceptAttention/concept_attention/flux/dit_block.py
pointer only (licence: NONE) · get_code("6d1fbe9be366c5ae")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, helblazer811/ConceptAttention 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.

scaled_dot_product_attention Not yet run
unverified (harvested, no recorded run)
helblazer811/ConceptAttention/concept_attention/flux/dit_block.py
pointer only (licence: NONE) · get_code("1af01c699046b8d4")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, helblazer811/ConceptAttention 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.

Also present in other papers' code

Of 9 samples lifted from this paper's linked repositories, 9 carry a function identity and 0 do not. Of the 9 distinct functions with an identity, 6 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 6 the same month, 6 later) and 3 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
apply_rope
brick of this paper's assembly: helper
9 other papers
3 dated earlier, 1 the same month, 5 later · 3 only as an assembly brick
2409.00587 (2024-09)
FLUX that Plays Music
2411.04746 (2024-11)
Taming Rectified Flow for Inversion and Editing
2412.07517 (2024-12)
FireFlow: Fast Inversion of Rectified Flow for Image Semantic Editing
and 6 more not named here
Modulation
brick of this paper's assembly: helper
5 other papers
2 the same month, 3 later · 3 only as an assembly brick
2502.01105 (2025-02)
LayerTracer: Cognitive-Aligned Layered SVG Synthesis via Diffusion Transformer
2502.17363 (2025-02)
KV-Edit: Training-Free Image Editing for Precise Background Preservation
2504.02160 (2025-04)
Less-to-More Generalization: Unlocking More Controllability by In-Context Generation
and 2 more not named here
ModulationOut
brick of this paper's assembly: helper
5 other papers
2 the same month, 3 later · 3 only as an assembly brick
2502.01105 (2025-02)
LayerTracer: Cognitive-Aligned Layered SVG Synthesis via Diffusion Transformer
2502.17363 (2025-02)
KV-Edit: Training-Free Image Editing for Precise Background Preservation
2504.02160 (2025-04)
Less-to-More Generalization: Unlocking More Controllability by In-Context Generation
and 2 more not named here
QKNorm
brick of this paper's assembly: helper
5 other papers
2 the same month, 3 later · 3 only as an assembly brick
2502.01105 (2025-02)
LayerTracer: Cognitive-Aligned Layered SVG Synthesis via Diffusion Transformer
2502.17363 (2025-02)
KV-Edit: Training-Free Image Editing for Precise Background Preservation
2504.02160 (2025-04)
Less-to-More Generalization: Unlocking More Controllability by In-Context Generation
and 2 more not named here
RMSNorm
brick of this paper's assembly: helper
3 other papers
1 the same month, 2 later · 2 only as an assembly brick
2502.17363 (2025-02)
KV-Edit: Training-Free Image Editing for Precise Background Preservation
2602.00883 (2026-02)
DIAMOND: Directed Inference for Artifact Mitigation in Flow Matching Models
2606.17584 (2026-06)
Root-Selecting Fixed-Point Inversion for Rectified Flows via Trajectory Straightness
SelfAttention
brick of this paper's assembly: helper
3 other papers
2 the same month, 1 later
2502.01105 (2025-02)
LayerTracer: Cognitive-Aligned Layered SVG Synthesis via Diffusion Transformer
2502.17363 (2025-02)
KV-Edit: Training-Free Image Editing for Precise Background Preservation
2606.17584 (2026-06)
Root-Selecting Fixed-Point Inversion for Rectified Flows via Trajectory Straightness

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

Do the rich representations of multi-modal diffusion transformers (DiTs) exhibit unique properties that enhance their interpretability? We introduce CONCEPTATTENTION, a novel method that leverages the expressive power of DiT attention layers to generate high-quality saliency maps that precisely locate textual concepts within images. Without requiring additional training, CONCEPTATTENTION repurposes the parameters of DiT attention layers to produce highly contextualized concept embeddings, contributing the major discovery that performing linear projections in the output space of DiT attention layers yields significantly sharper saliency maps compared to commonly used cross-attention maps. CONCEPTATTENTION even achieves state-of-theart performance on zero-shot image segmentation benchmarks, outperforming 15 other zeroshot interpretability methods on the ImageNet-Segmentation dataset. CONCEPTATTENTION works for popular image models and even seamlessly generalizes to video generation. Our work contributes the first evidence that the representations of multi-modal DiTs are highly transferable to vision tasks like segmentation.

For agents

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

get_harvested_code_for_paper("2502.04320")
get_code_for_paper("2502.04320")
have("2502.04320")

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

curl https://syntology.ai/api/ran/2502.04320.json

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

[![Syntology run record](https://syntology.ai/api/ran/2502.04320.svg)](https://syntology.ai/paper/2502.04320)

Connect an agent — have() is free.