SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2312.04567 · 2023

Scaling Laws of Synthetic Images for Model Training ... for Now

arXiv · PDF · Open in the Atlas

Code that ran

4 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; 1 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 3 failures of our instrument; 0 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/2312.04567.json · badge: /api/ran/2312.04567.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
google-research/syn-rep-learn canonical (classifier label)
classifier basis: classifier_2025
2 ran · 0 no recorded run
copy not recorded community
not in the paper's declared repositories
2 ran · 0 no recorded run
FunctionStatusWhere it lives
accuracy Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
this paper's copy was not recorded; identical code first harvested from liyunqianggyn/equal-bits-bnn
pointer only; the stored copy get_code serves, from liyunqianggyn/equal-bits-bnn, is pointer only (licence: NONE) · get_code("a5e37b0cab1cbb28")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. 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.

teacher_temp_schedule Ran
ran, honoured the contract we drafted
this paper's copy was not recorded; identical code first harvested from google-research/syn-rep-learn
pointer only · get_code("3cc84e91b2c222fd")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. 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.

validate Ran
ran; our contract draft was wrong, not the code
google-research/syn-rep-learn/Scaling/supervised/eval_imagenet_supervised.py
code served (permissive licence) · get_code("17ed3c435b35d928")
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. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 a small model that takes images and outputs logits
class SimpleModel(nn.Module):
    def __init__(self):
        super(SimpleModel, self).__init__()
        self.fc = nn.Linear(4, 10)  # 10 classes for top-5 accuracy
    
    def forward(self, x):
        # Flatten input and pass through linear layer
        x = x.view(x.size(0), -1)
        return self.fc(x)

# Create a small validation dataset (list of (images, targets) tuples)
val_loader = [
    (torch.randn(2, 1, 2, 2), torch.randint(0, 10, (2,))),
    (torch.randn(2, 1, 2, 2), torch.randint(0, 10, (2,))),
    (torch.randn(2, 1, 2, 2), torch.randint(0, 10, (2,))),
]

# Create a small model instance
model = SimpleModel()

# Create args object with necessary attributes
class Args:
    def __init__(self):
        self.gpu = None  # CPU only
        self.print_freq = 2  # Print every 2 batches

args = Args()

ARGS = {
    'val_loader': val_loader,
    'model': model,
    'args': args
}
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
validate_zeroshot Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
google-research/syn-rep-learn/Scaling/clip/eval_imagenet_clip.py
code served (permissive licence) · get_code("1c99b4c3698e1fc2")
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. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 torch.utils.data

torch.manual_seed(0)

class MockTokenizer:
    def __call__(self, texts):
        # Return a tensor of shape (len(texts), 77) with dummy token ids
        return torch.randint(0, 100, (len(texts), 77))

class MockCLIPModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.text_proj = nn.Linear(512, 512)
        self.image_proj = nn.Linear(512, 512)
    
    def encode_text(self, tokens):
        # tokens shape: (batch, 77)
        # Return embeddings of shape (batch, 512)
        batch_size = tokens.shape[0]
        return self.text_proj(torch.randn(batch_size, 512))
    
    def encode_image(self, images):
        # images shape: (batch, 3, 224, 224) or similar
        # Return embeddings of shape (batch, 512)
        batch_size = images.shape[0]
        return self.image_proj(torch.randn(batch_size, 512))

# Create model and move to CPU (don't use cuda)
model = MockCLIPModel()
model.eval()

# Create tokenizer
tokenizer = MockTokenizer()

# Create templates for zero-shot prompts
templates = ["a photo of a {}"]

# Create labels (class names)
labels = ["dog", "cat"]

# Create a small validation loader
# Each batch contains (images, targets)
val_loader = [
    (torch.randn(2, 3, 224, 224), torch.tensor([0, 1])),
    (torch.randn(2, 3, 224, 224), torch.tensor([1, 0])),
]

# Logit scale parameter
logit_scale = torch.tensor(100.0)

ARGS = {
    "val_loader": val_loader,
    "templates": templates,
    "labels": labels,
    "model": model,
    "tokenizer": tokenizer,
    "logit_scale": logit_scale,
}
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 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

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, 2 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 1 the same month, 1 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
accuracy 116 other papers
90 dated earlier, 24 later, 2 same year, order not determinable · 8 only as an assembly brick
1502.03044 (2015-02)
Show, Attend and Tell: Neural Image Caption Generation with Visual Attention
1610.05820 (2016-10)
Membership Inference Attacks Against Machine Learning Models
1905.03696 (2019-05)
HAWQ: Hessian AWare Quantization of Neural Networks with Mixed-Precision
and 113 more not named here
teacher_temp_schedule 2 other papers
1 dated earlier, 1 the same month
2306.00984 (2023-06)
StableRep: Synthetic Images from Text-to-Image Models Make Strong Visual Representation Learners
2312.17742 (2023-12)
Learning Vision from Models Rivals Learning Vision from Data

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

Recent significant advances in text-to-image models unlock the possibility of training vision systems using synthetic images, potentially overcoming the difficulty of collecting curated data at scale. It is unclear, however, how these models behave at scale, as more synthetic data is added to the training set. In this paper we study the scaling laws of synthetic images generated by state of the art text-to-image models, for the training of supervised models: image classifiers with label supervision, and CLIP with language supervision. We identify several factors, including text prompts, classifier-free guidance scale, and types of text-to-image models, that significantly affect scaling behavior. After tuning these factors, we observe that synthetic images demonstrate a scaling trend similar to, but slightly less effective than, real images in CLIP training, while they significantly underperform in scaling when training supervised image classifiers. Our analysis indicates that the main reason for this underperformance is the inability of off-the-shelf text-to-image models to generate certain concepts, a limitation that significantly impairs the training of image classifiers. Our findings also suggest that scaling synthetic data can be particularly effective in scenarios such as: (1) when there is a limited supply of real images for a supervised problem (e.g., fewer than 0.5 million images in ImageNet), (2) when the evaluation dataset diverges significantly from the training data, indicating the out-of-distribution scenario, or (3) when synthetic data is used in conjunction with real images, as demonstrated in the training of CLIP models.

For agents

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

get_harvested_code_for_paper("2312.04567")
get_code_for_paper("2312.04567")
have("2312.04567")

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

curl https://syntology.ai/api/ran/2312.04567.json

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

[![Syntology run record](https://syntology.ai/api/ran/2312.04567.svg)](https://syntology.ai/paper/2312.04567)

Connect an agent — have() is free.