SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · aaai_28565 · AAAI · 2024

Compositional Inversion for Stable Diffusion Models

Tianyi Zhang, Qing Li, Zhaoxiang Zhang, Jinlin Wu, Zhen Lei, Xulu Zhang, Xiao-Yong Wei

Open in the Atlas

Code that ran

5 of 8 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 3 have no contract verdict; 2 failures of our instrument; 3 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/aaai_28565.json · badge: /api/ran/aaai_28565.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
zhangxulu1996/Compositional-Inversion canonical (classifier label)
classifier basis: census
5 ran · 3 no recorded run
FunctionStatusWhere it lives
center_crop Ran
ran on a synthesized input
zhangxulu1996/Compositional-Inversion/train_compositional_textual_inversion.py
code served (permissive licence) · get_code("81aae1af1d3e0807")
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 ndarray, ndim 2, dtype family float, shape (3, 3) 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)

# The function expects an image-like object with .shape attribute
# that supports integer indexing with slices
# Create a small image array (height=4, width=3)
img = np.random.rand(4, 3)

ARGS = {
    "img": img
}
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 ndarray, ndim 2, dtype family float, shape (3, 3) — shape (3, 3) 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
collate_fn Ran
ran; our contract draft was wrong, not the code
zhangxulu1996/Compositional-Inversion/train_compositional_custom_diffusion.py
code served (permissive licence) · get_code("b906198eb13e16e4")
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 dict. 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)

# Create example dictionaries that match the structure expected by collate_fn
examples = [
    {
        'placeholder_index': torch.tensor([[0]]),
        'instance_prompt_ids': torch.randint(0, 1000, (1, 4)),
        'reference_prompt_ids': torch.randint(0, 1000, (1, 4)),
        'instance_images': torch.randn(3, 8, 8),
        'mask': torch.randint(0, 2, (1, 8, 8)).float(),
        'class_prompt_ids': torch.randint(0, 1000, (1, 4)),
        'class_images': torch.randn(3, 8, 8),
        'class_mask': torch.randint(0, 2, (1, 8, 8)).float(),
    },
    {
        'placeholder_index': torch.tensor([[1]]),
        'instance_prompt_ids': torch.randint(0, 1000, (1, 4)),
        'reference_prompt_ids': torch.randint(0, 1000, (1, 4)),
        'instance_images': torch.randn(3, 8, 8),
        'mask': torch.randint(0, 2, (1, 8, 8)).float(),
        'class_prompt_ids': torch.randint(0, 1000, (1, 4)),
        'class_images': torch.randn(3, 8, 8),
        'class_mask': torch.randint(0, 2, (1, 8, 8)).float(),
    },
]

ARGS = {
    'examples': examples,
    'with_prior_preservation': True,
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
collate_fn Ran
ran on a synthesized input
zhangxulu1996/Compositional-Inversion/train_compositional_dreambooth.py
code served (permissive licence) · get_code("0ee3c73b333759c9")
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 dict. 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
import torch.nn as nn

torch.manual_seed(0)

# Create example dictionaries that match the structure expected by collate_fn
examples = [
    {
        'placeholder_index': torch.tensor([[0]]),
        'instance_prompt_ids': torch.randint(0, 100, (1, 4)),
        'reference_prompt_ids': torch.randint(0, 100, (1, 4)),
        'instance_images': torch.randn(3, 8, 8),
        'class_prompt_ids': torch.randint(0, 100, (1, 4)),
        'class_images': torch.randn(3, 8, 8),
    },
    {
        'placeholder_index': torch.tensor([[1]]),
        'instance_prompt_ids': torch.randint(0, 100, (1, 4)),
        'reference_prompt_ids': torch.randint(0, 100, (1, 4)),
        'instance_images': torch.randn(3, 8, 8),
        'class_prompt_ids': torch.randint(0, 100, (1, 4)),
        'class_images': torch.randn(3, 8, 8),
    },
]

ARGS = {
    'examples': examples,
    'with_prior_preservation': True,
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
contrastive_loss Ran
ran; our contract draft was wrong, not the code
zhangxulu1996/Compositional-Inversion/model/modeling_clip.py
pointer only (licence: MIT); the stored copy get_code serves, from microsoft/xpretrain, is pointer only (licence: NOASSERTION) · get_code("4675779aec92d6c0")
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 microsoft/xpretrain, and GitHub could not classify the licence of that repository (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.

margin_loss Ran
ran on a synthesized input
zhangxulu1996/Compositional-Inversion/train_compositional_textual_inversion.py
code served (permissive licence) · get_code("c71ca95da295a5eb")
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 0, dtype family float, shape () 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.functional as F

torch.manual_seed(0)

# Create small tensors for testing
batch_size = 2
num_classes = 4

# scores: shape (batch_size, num_classes)
scores = torch.randn(batch_size, num_classes)

# targets: shape (batch_size,) - this should be class indices but the function
# treats it as if it's comparable to scores via F.mse_loss/F.l1_loss
# Looking at the function body, targets is used with F.mse_loss and F.l1_loss
# which expect same-shape tensors, so targets should also be (batch_size, num_classes)
targets = torch.randn(batch_size, num_classes)

# margin: float value
margin = 1.0

# embed_loss: string specifying loss type
embed_loss = 'l2'

ARGS = {
    'scores': scores,
    'targets': targets,
    'margin': margin,
    'embed_loss': embed_loss
}
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 0, dtype family float, shape () — shape () 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
clip_loss Not yet run
unverified (harvested, no recorded run)
zhangxulu1996/Compositional-Inversion/model/modeling_clip.py
pointer only (licence: MIT); the stored copy get_code serves, from PeterGriffinJin/InstructG2I, is pointer only (licence: NONE) · get_code("f2ca34eca05d40cd")
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 PeterGriffinJin/InstructG2I, 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. No invocation recorded.

parse_args Not yet run
unverified (harvested, no recorded run)
zhangxulu1996/Compositional-Inversion/train_compositional_custom_diffusion.py
code served (permissive licence) · get_code("82483e1d49e231a3")
How it ran

no invocation recorded.

parse_args Not yet run
unverified (harvested, no recorded run)
zhangxulu1996/Compositional-Inversion/train_compositional_dreambooth.py
code served (permissive licence) · get_code("681aeccd5b2a8880")
How it ran

no invocation recorded.

Also present in other papers' code

Of 8 samples lifted from this paper's linked repositories, 8 carry a function identity and 0 do not. Of the 8 distinct functions with an identity, 5 are also present in other papers whose code we hold (5 with at least one paper dated earlier, 0 the same month, 2 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
contrastive_loss 27 other papers
10 dated earlier, 7 later, 10 same year, order not determinable · 3 only as an assembly brick
2206.00621 (2022-06)
Cross-View Language Modeling: Towards Unified Cross-Lingual Cross-Modal Pre-training
2209.06430 (2022-09)
CLIP-ViP: Adapting Pre-trained Image-Text Model to Video-Language Alignment
2210.00312 (2022-10)
Multimodal Analogical Reasoning over Knowledge Graphs
and 24 more not named here
clip_loss 16 other papers
3 dated earlier, 6 later, 7 same year, order not determinable · 1 only as an assembly brick
2305.13050 (2023-05)
AudioToken: Adaptation of Text-Conditioned Diffusion Models for Audio-to-Image Generation
2306.02583 (2023-06)
Stable Diffusion is Unstable
2309.16429 (2023-09)
Diverse and Aligned Audio-to-Video Generation via Text-to-Video Model Adaptation
and 13 more not named here
center_crop 1 other paper
1 dated earlier
2211.09800 (2022-11)
InstructPix2Pix: Learning to Follow Image Editing Instructions
collate_fn 1 other paper
1 dated earlier
2312.08048 (2023-12)
Compositional Inversion for Stable Diffusion Models
parse_args 1 other paper
1 dated earlier
2312.08048 (2023-12)
Compositional Inversion for Stable Diffusion Models

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

Inversion methods, such as Textual Inversion, generate personalized images by incorporating concepts of interest provided by user images. However, existing methods often suffer from overfitting issues, where the dominant presence of inverted concepts leads to the absence of other desired concepts. It stems from the fact that during inversion, the irrelevant semantics in the user images are also encoded, forcing the inverted concepts to occupy locations far from the core distribution in the embedding space. To address this issue, we propose a method that guides the inversion process towards the core distribution for compositional embeddings. Additionally, we introduce a spatial regularization approach to balance the attention on the concepts being composed. Our method is designed as a post-training approach and can be seamlessly integrated with other inversion methods. Experimental results demonstrate the effectiveness of our proposed approach in mitigating the overfitting problem and generating more diverse and balanced compositions of concepts in the synthesized images. The source code is available at https://github.com/zhangxulu1996/Compositional-Inversion.

For agents

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

get_harvested_code_for_paper("aaai_28565")
get_code_for_paper("aaai_28565")
have("aaai_28565")

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

curl https://syntology.ai/api/ran/aaai_28565.json

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

[![Syntology run record](https://syntology.ai/api/ran/aaai_28565.svg)](https://syntology.ai/paper/aaai_28565)

Connect an agent — have() is free.