SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2406.03793 · ICML · 2024

Low-Rank Similarity Mining for Multimodal Dataset Distillation

Yong-Lu Li, Cewu Lu, Yue Xu, Zhilin Lin, Yusong Qiu, ! 𝐼, " 𝐼 # 𝐼 $ 𝐼

arXiv · PDF · Open in the Atlas

Code that ran

12 of 16 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (1 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 1 violated it and 1 has no contract verdict; 9 failures of our instrument; 4 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/2406.03793.json · badge: /api/ran/2406.03793.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
silicx/LoRS_Distill canonical (classifier label)
classifier basis: census
11 ran · 3 no recorded run
silicx/lors_distill community
Papers with Code archive
1 ran · 1 no recorded run
FunctionStatusWhere it lives
LowRankSimilarityGenerator Ran
ran on a synthesized input · constructed an object; no result was computed
silicx/lors_distill/src/similarity_mining.py
code served (permissive licence) · get_code("68db47247c28a3d3")
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 an instance of its own class, LowRankSimilarityGenerator: the run constructed the object and did not call it, so no result was computed. 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)

ARGS = {
    'dim': 4,
    'rank': 2,
    'alpha': 0.1
}
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 LowRankSimilarityGenerator — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a LowRankSimilarityGenerator 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
VGG11 Ran
ran; our contract draft was wrong, not the code
silicx/LoRS_Distill/src/networks.py
pointer only (licence: BSD-3-Clause); the stored copy get_code serves, from BigML-CS-UCLA/MKDT, is pointer only (licence: NONE) · get_code("190dfb67aacf4767")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (BSD-3-Clause), but get_code serves the stored copy, first harvested from BigML-CS-UCLA/MKDT, 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.

VGG11BN Ran
ran; our contract draft was wrong, not the code
silicx/LoRS_Distill/src/networks.py
pointer only (licence: BSD-3-Clause); the stored copy get_code serves, from BigML-CS-UCLA/MKDT, is pointer only (licence: NONE) · get_code("6fa9b9b022f6dbfa")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (BSD-3-Clause), but get_code serves the stored copy, first harvested from BigML-CS-UCLA/MKDT, 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.

VGG13 Ran
ran; our contract draft was wrong, not the code
silicx/LoRS_Distill/src/networks.py
pointer only (licence: BSD-3-Clause); the stored copy get_code serves, from BigML-CS-UCLA/MKDT, is pointer only (licence: NONE) · get_code("3da7c8c828c4f0af")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (BSD-3-Clause), but get_code serves the stored copy, first harvested from BigML-CS-UCLA/MKDT, 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.

epoch Ran
ran, honoured the contract we drafted
silicx/LoRS_Distill/src/epoch.py
code served (permissive licence) · get_code("7d65df8b482e868e")
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 1, 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
import numpy as np
from math import ceil

torch.manual_seed(0)

class DummyNet(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 2)
    
    def forward(self, image, caption, epoch, similarity=None):
        batch_size = image.shape[0]
        logits = self.linear(image)
        loss = logits.sum()
        acc = torch.tensor(0.9)
        return loss, acc

class DummyArgs:
    def __init__(self):
        self.device = 'cpu'
        self.distill = False

# Create a simple dataloader
class SimpleDataLoader:
    def __init__(self, num_samples=8, batch_size=2):
        self.num_samples = num_samples
        self.batch_size = batch_size
    
    def __iter__(self):
        for i in range(0, self.num_samples, self.batch_size):
            batch_size = min(self.batch_size, self.num_samples - i)
            image = torch.randn(batch_size, 4)
            caption = torch.randn(batch_size, 3)
            index = torch.arange(i, i + batch_size)
            yield (image, caption, index)

# Create dataloader with similarity
class SimilarityDataloaderTest:
    def __init__(self, num_samples=8, batch_size=2):
        self.images_train = torch.randn(num_samples, 4)
        self.labels_train = torch.randint(0, 2, (num_samples,))
        self.similarity_train = torch.eye(num_samples)
        self.batch_size = batch_size
    
    def __iter__(self):
        size = self.images_train.shape[0]
        num_batch = int(ceil(size / self.batch_size))
        indices = np.arange(size)
        for _ in range(num_batch):
            np.random.shuffle(indices)
            ids = indices[:self.batch_size]
            yield (self.images_train[ids], self.labels_train[ids], self.similarity_train[ids[:, None], ids])

# Prepare arguments
net = DummyNet()
optimizer = torch.optim.SGD(net.parameters(), lr=0.01)
args = DummyArgs()
dataloader = SimpleDataLoader(num_samples=8, batch_size=2)
e = 0

ARGS = {
    'e': e,
    'dataloader': dataloader,
    'net': net,
    'optimizer': optimizer,
    'args': args
}
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 tuple, ndim 1, 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
formatting_result_all Ran
ran; our contract draft was wrong, not the code
silicx/LoRS_Distill/evaluate_only.py
code served (permissive licence) · get_code("7b3e6879e6bda6f8")
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 str. 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)

val_result = {
    'img_r1': 0.85,
    'img_r5': 0.92,
    'img_r10': 0.95,
    'txt_r1': 0.82,
    'txt_r5': 0.90,
    'txt_r10': 0.93,
    'r_mean': 0.8967
}

ARGS = {
    'val_result': val_result
}
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 str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
formatting_result_content Ran
ran; our contract draft was wrong, not the code
silicx/LoRS_Distill/evaluate_only.py
code served (permissive licence) · get_code("01b160948b4eabee")
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 str. 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)

val_result = {
    'img_r1': 85.50,
    'img_r5': 92.30,
    'img_r10': 95.75,
    'txt_r1': 78.20,
    'txt_r5': 88.10,
    'txt_r10': 91.45,
    'r_mean': 88.72
}

ARGS = {
    'val_result': val_result
}
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 str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
formatting_result_content_clean Ran
ran; our contract draft was wrong, not the code
silicx/LoRS_Distill/evaluate_only.py
code served (permissive licence) · get_code("e74d3f1a61087ec9")
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 str. 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)

ARGS = {
    'val_result': {
        'img_r1': 0.85,
        'img_r5': 0.92,
        'img_r10': 0.96,
        'txt_r1': 0.83,
        'txt_r5': 0.90,
        'txt_r10': 0.95,
        'r_mean': 0.90
    }
}
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 str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
itm_eval Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
silicx/LoRS_Distill/src/epoch.py
code served (permissive licence) · get_code("c7f85279d772ea19")
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 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 numpy as np
import torch

torch.manual_seed(0)

# scores_i2t: shape (num_images, num_texts) - similarity scores from images to texts
# Each row is scores for one image against all texts
scores_i2t = np.random.rand(2, 4).astype(np.float32)

# scores_t2i: shape (num_texts, num_images) - similarity scores from texts to images
# Each row is scores for one text against all images
scores_t2i = np.random.rand(4, 2).astype(np.float32)

# txt2img: dict mapping text index to image index
# For each text, which image is the correct match
txt2img = {
    0: 0,
    1: 1,
    2: 0,
    3: 1
}

# img2txt: dict mapping image index to list of text indices
# For each image, which texts are valid matches
img2txt = {
    0: [0, 2],
    1: [1, 3]
}

ARGS = {
    'scores_i2t': scores_i2t,
    'scores_t2i': scores_t2i,
    'txt2img': txt2img,
    'img2txt': img2txt
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, 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
make_timestamp Ran
ran; our contract draft was wrong, not the code
silicx/LoRS_Distill/distill_tesla_lors.py
code served (permissive licence) · get_code("1b3fdd7b1b5bf6c0")
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 str. 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 datetime

def make_timestamp(prefix: str='', suffix: str='') -> str:
    tmstamp = '{:%m%d_%H%M%S}'.format(datetime.datetime.now())
    return prefix + tmstamp + suffix

ARGS = {
    'prefix': 'test_',
    'suffix': '_end'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
nearest_neighbor Ran
ran, violated the contract we drafted
silicx/LoRS_Distill/src/vl_distill_utils.py
pointer only (licence: BSD-3-Clause); the stored copy get_code serves, from princetonvisualai/multimodal_dataset_distillation, is pointer only (licence: NONE) · get_code("c84ddf44fce519da")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (BSD-3-Clause), but get_code serves the stored copy, first harvested from princetonvisualai/multimodal_dataset_distillation, 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.

shuffle_files Ran
ran; our contract draft was wrong, not the code
silicx/LoRS_Distill/src/vl_distill_utils.py
code served (permissive licence) · get_code("bf0e428bf2a3f2f0")
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 2, 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 numpy as np

# Create sample file lists as numpy arrays
img_expert_files = np.array(['img_file_1.jpg', 'img_file_2.jpg', 'img_file_3.jpg', 'img_file_4.jpg'])
txt_expert_files = np.array(['txt_file_1.txt', 'txt_file_2.txt', 'txt_file_3.txt', 'txt_file_4.txt'])

ARGS = {
    'img_expert_files': img_expert_files,
    'txt_expert_files': txt_expert_files
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 2, 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
BaseSimilarityGenerator Not yet run
unverified (harvested, no recorded run)
silicx/lors_distill/src/similarity_mining.py
code served (permissive licence) · get_code("9e70711e612fe6a7")
How it ran

no invocation recorded.

build_model Not yet run
unverified (harvested, no recorded run)
silicx/LoRS_Distill/src/model.py
code served (permissive licence) · get_code("1808da638d8ebc0f")
How it ran

no invocation recorded.

epoch_test Not yet run
unverified (harvested, no recorded run)
silicx/LoRS_Distill/src/epoch.py
code served (permissive licence) · get_code("daec924451bcf2dc")
How it ran

no invocation recorded.

get_dataset Not yet run
unverified (harvested, no recorded run)
silicx/LoRS_Distill/src/utils.py
code served (permissive licence) · get_code("2a69179a8e3a05a4")
How it ran

no invocation recorded.

Also present in other papers' code

Of 16 samples lifted from this paper's linked repositories, 16 carry a function identity and 0 do not. Of the 16 distinct functions with an identity, 6 are also present in other papers whose code we hold (6 with at least one paper dated earlier, 3 the same month, 6 later) and 10 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
build_model 119 other papers
73 dated earlier, 3 the same month, 36 later, 5 same year, order not determinable, 2 undated
1809.11096 (2018-09)
Large Scale GAN Training for High Fidelity Natural Image Synthesis
1901.07441 (2019-01)
PadChest: A large chest x-ray image dataset with multi-label annotated reports
2104.14294 (2021-04)
Emerging Properties in Self-Supervised Vision Transformers
and 116 more not named here
VGG11BN 28 other papers
18 dated earlier, 1 the same month, 8 later, 1 same year, order not determinable
2203.01531 (2022-03)
CAFE: Learning to Condense Dataset by Aligning Features
2204.07513 (2022-04)
Synthesizing Informative Training Samples with GAN
2210.16774 (2022-10)
Dataset Distillation via Factorization
and 25 more not named here
VGG11 27 other papers
17 dated earlier, 1 the same month, 8 later, 1 same year, order not determinable
2203.01531 (2022-03)
CAFE: Learning to Condense Dataset by Aligning Features
2204.07513 (2022-04)
Synthesizing Informative Training Samples with GAN
2210.16774 (2022-10)
Dataset Distillation via Factorization
and 24 more not named here
VGG13 19 other papers
14 dated earlier, 4 later, 1 same year, order not determinable
2204.07513 (2022-04)
Synthesizing Informative Training Samples with GAN
2210.16774 (2022-10)
Dataset Distillation via Factorization
2211.04446 (2022-11)
Private Set Generation with Discriminative Information
and 16 more not named here
itm_eval 13 other papers
12 dated earlier, 1 later
2107.07651 (2021-07)
Align before Fuse: Vision and Language Representation Learning with Momentum Distillation
2202.10401 (2022-02)
Vision-Language Pre-Training with Triple Contrastive Learning
2206.00621 (2022-06)
Cross-View Language Modeling: Towards Unified Cross-Lingual Cross-Modal Pre-training
and 10 more not named here
nearest_neighbor 2 other papers
1 dated earlier, 1 later
2308.07545 (2023-08)
Vision-Language Dataset Distillation
2407.20299 (2024-07)
Dataset Distillation for Offline Reinforcement Learning

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

Though dataset distillation has witnessed rapid development in recent years, the distillation of multimodal data, e.g., image-text pairs, poses unique and under-explored challenges. Unlike unimodal data, image-text contrastive learning (ITC) data lack inherent categorization and should instead place greater emphasis on modality correspondence. In this work, we propose Low-Rank Similarity Mining (LoRS) for multimodal dataset distillation, that concurrently distills a ground truth similarity matrix with image-text pairs, and leverages low-rank factorization for efficiency and scalability. The proposed approach brings significant improvement to the existing algorithms, marking a significant contribution to the field of visual-language dataset distillation. We advocate adopting LoRS as a foundational synthetic data setup for image-text dataset distillation. Our code is available at https: //github.com/silicx/LoRS_Distill.

For agents

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

get_harvested_code_for_paper("2406.03793")
get_code_for_paper("2406.03793")
have("2406.03793")

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

curl https://syntology.ai/api/ran/2406.03793.json

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

[![Syntology run record](https://syntology.ai/api/ran/2406.03793.svg)](https://syntology.ai/paper/2406.03793)

Connect an agent — have() is free.