SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2202.04053 · 2022

DALL-Eval: Probing the Reasoning Skills and Social Biases of Text-to-Image Generation Models

arXiv · PDF · Open in the Atlas

Code that ran

10 of 10 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, 1 violated it and 1 has no contract verdict; 8 failures of our instrument; 0 no recorded run), measured 2026-11-21; 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/2202.04053.json · badge: /api/ran/2202.04053.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
j-min/dalleval canonical (classifier label)
classifier basis: plan_row
10 ran · 0 no recorded run
FunctionStatusWhere it lives
calculate_activation_statistics Ran
ran; our contract draft was wrong, not the code
j-min/dalleval/quality/FID/fid_score.py
code served (permissive licence) · get_code("177298ecedde13ed")
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. 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
import numpy as np

torch.manual_seed(0)

# Create a minimal model that mimics Inception behavior
class MinimalInceptionModel(nn.Module):
    def __init__(self, output_dims=2048):
        super(MinimalInceptionModel, self).__init__()
        self.conv = nn.Conv2d(3, output_dims, kernel_size=3, padding=1)
    
    def forward(self, x):
        # Returns a tuple where first element is the feature map
        out = self.conv(x)
        # Return as tuple to match model(batch)[0] indexing in get_activations
        return (out,)

# Create a minimal model instance
model = MinimalInceptionModel(output_dims=2048)

# Create dummy images as a list of batches
# Each batch should be a tensor of shape (batch_size, 3, height, width)
# Values must be between 0 and 1
batch1 = torch.rand(2, 3, 8, 8)  # 2 images, 3 channels, 8x8 spatial dims
batch2 = torch.rand(2, 3, 8, 8)
images = [batch1, batch2]

ARGS = {
    "images": images,
    "model": model,
    "batch_size": 2,
    "dims": 2048,
    "cuda": False,
    "verbose": True
}
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 — a tuple output has no array shape of its own; none is 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
calculate_frechet_distance Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
j-min/dalleval/quality/FID/fid_score.py
code served (permissive licence) · get_code("25ef415d07e43e6f")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Sonnet 4.5, prompt python_fixture_v1) and returned float64, 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 numpy as np

np.random.seed(0)

# Create two multivariate Gaussian distributions with different dimensions for clarity
# mu1, mu2 should be 1D arrays of the same length (feature dimension)
# sigma1, sigma2 should be 2D covariance matrices of shape (feature_dim, feature_dim)

feature_dim = 5

# Mean vectors for two distributions
mu1 = np.random.randn(feature_dim)
mu2 = np.random.randn(feature_dim)

# Covariance matrices - need to be positive semi-definite
# Create them by A @ A.T to ensure they are valid covariance matrices
A1 = np.random.randn(feature_dim, feature_dim)
sigma1 = A1 @ A1.T

A2 = np.random.randn(feature_dim, feature_dim)
sigma2 = A2 @ A2.T

eps = 1e-06

ARGS = {
    'mu1': mu1,
    'sigma1': sigma1,
    'mu2': mu2,
    'sigma2': sigma2,
    'eps': eps
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
Imports
numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float64, 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
cp_path_to_dir Ran
ran; our contract draft was wrong, not the code
j-min/dalleval/models/dalle_small/DALLE-pytorch/finetune_dalle.py
code served (permissive licence) · get_code("7ae941c6130274cb")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Sonnet 4.5, prompt python_fixture_v1) and returned PosixPath. 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.

from pathlib import Path
import tempfile
import os

# Create a temporary directory structure for testing
temp_dir = tempfile.mkdtemp()
temp_path = Path(temp_dir)

# Create a checkpoint file path (not an actual file, just a path)
cp_file_path = temp_path / "model_checkpoint.pth"

# Define the arguments
ARGS = {
    "cp_path": cp_file_path,
    "tag": "best"
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
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 PosixPath — a PosixPath output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a PosixPath 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
exists Ran
ran, violated the contract we drafted
j-min/dalleval/models/dalle_small/DALLE-pytorch/finetune_dalle.py
pointer only (licence: MIT); the stored copy get_code serves, from ThomasMrY/VCT, is pointer only (licence: NONE) · get_code("aa5486a3650902d8")
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 ThomasMrY/VCT, 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.

get_activations Ran
ran; our contract draft was wrong, not the code
j-min/dalleval/quality/FID/fid_score.py
code served (permissive licence) · get_code("eb21740d88a30ecb")
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 (4, 2048) as a later run recorded it. 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, 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 as nn
import numpy as np

torch.manual_seed(0)

# Create a minimal Inception-like model that returns a tensor with spatial dims
class MinimalInceptionModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.conv = nn.Conv2d(3, 2048, kernel_size=3, padding=1)
    
    def forward(self, x):
        # Returns tuple where first element is the feature map
        out = self.conv(x)
        return (out,)

# Create model instance
model = MinimalInceptionModel()

# Create dummy image data: (n_images, 3, height, width)
# Values between 0 and 1
images_list = [
    torch.rand(2, 3, 8, 8),  # First batch of 2 images
    torch.rand(2, 3, 8, 8),  # Second batch of 2 images
]

ARGS = {
    'images': images_list,
    'model': model,
    'batch_size': 2,
    'dims': 2048,
    'cuda': False,
    'verbose': True,
}
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 (4, 2048) — shape (4, 2048) 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
get_trainable_params Ran
ran; our contract draft was wrong, not the code
j-min/dalleval/models/dalle_small/DALLE-pytorch/finetune_dalle.py
code served (permissive licence) · get_code("863c2d358b6285ad")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Sonnet 4.5, prompt python_fixture_v1) and returned list. 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)

class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.layer1 = nn.Linear(4, 3)
        self.layer2 = nn.Linear(3, 2)
        self.frozen_layer = nn.Linear(2, 5)
        
        # Freeze one layer to test the requires_grad filtering
        for param in self.frozen_layer.parameters():
            param.requires_grad = False
    
    def forward(self, x):
        x = self.layer1(x)
        x = self.layer2(x)
        x = self.frozen_layer(x)
        return x

model = SimpleModel()

ARGS = {
    'model': model
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list — a list output has no array shape of its own; none is 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
np_mad Ran
ran on a synthesized input
j-min/dalleval/biases/compute_mad.py
code served (permissive licence) · get_code("35f46cc480ac8cea")
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 float64, ndim 0, 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 a simple numpy array as test data
data = np.array([[1.0, 2.0, 3.0, 4.0],
                 [5.0, 6.0, 7.0, 8.0],
                 [9.0, 10.0, 11.0, 12.0]])

ARGS = {
    "data": data,
    "axis": None
}
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 float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
run Ran
ran; our contract draft was wrong, not the code
j-min/dalleval/models/dalle_small/DALLE-pytorch/collect_env.py
code served (permissive licence) · get_code("0365a0718555a518")
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 subprocess
import sys
import locale

# The function 'run' takes a single parameter: 'command'
# It executes a shell command and returns a tuple of (return_code, stdout, stderr)
# The parameter should be a string representing a valid shell command

ARGS = {
    'command': 'echo "Hello, World!"'
}
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 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
run_and_parse_first_match Ran
ran; our contract draft was wrong, not the code
j-min/dalleval/models/dalle_small/DALLE-pytorch/collect_env.py
pointer only (licence: MIT); the stored copy get_code serves, from NolanKoblischke/SpectraFM_NeurIPS_FM4Science, is pointer only (licence: NONE) · get_code("ab7aeddfa9e9fb2e")
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 NolanKoblischke/SpectraFM_NeurIPS_FM4Science, 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.

run_and_read_all Ran
ran; our contract draft was wrong, not the code
j-min/dalleval/models/dalle_small/DALLE-pytorch/collect_env.py
pointer only (licence: MIT); the stored copy get_code serves, from NolanKoblischke/SpectraFM_NeurIPS_FM4Science, is pointer only (licence: NONE) · get_code("84905132fbc263e3")
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 NolanKoblischke/SpectraFM_NeurIPS_FM4Science, 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.

Also present in other papers' code

Of 10 samples lifted from this paper's linked repositories, 10 carry a function identity and 0 do not. Of the 10 distinct functions with an identity, 9 are also present in other papers whose code we hold (9 with at least one paper dated earlier, 2 the same month, 9 later) and 1 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
exists 486 other papers
78 dated earlier, 4 the same month, 401 later, 1 same year, order not determinable, 2 undated · 78 only as an assembly brick
1603.08155 (2016-03)
Perceptual Losses for Real-Time Style Transfer and Super-Resolution
1611.07004 (2016-11)
Image-to-Image Translation with Conditional Adversarial Networks
1804.04637 (2018-04)
EMBER: An Open Dataset for Training Static PE Malware Machine Learning Models
and 483 more not named here
calculate_frechet_distance 74 other papers
25 dated earlier, 1 the same month, 48 later
1610.09585 (2016-10)
Conditional Image Synthesis With Auxiliary Classifier GANs
1706.08500 (2017-06)
GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium
1807.11458 (2018-07)
To learn image super-resolution, use a GAN to learn how to do image degradation first
and 71 more not named here
run_and_parse_first_match 14 other papers
2 dated earlier, 12 later
2109.04003 (2021-09)
Improving Deep Metric Learning by Divide and Conquer
2109.08267 (2021-09)
CompilerGym: Robust, Performant Compiler Optimization Environments for AI Research
2309.06180 (2023-09)
Efficient Memory Management for Large Language Model Serving with PagedAttention
and 11 more not named here
run_and_read_all 14 other papers
2 dated earlier, 12 later
2109.04003 (2021-09)
Improving Deep Metric Learning by Divide and Conquer
2109.08267 (2021-09)
CompilerGym: Robust, Performant Compiler Optimization Environments for AI Research
2309.06180 (2023-09)
Efficient Memory Management for Large Language Model Serving with PagedAttention
and 11 more not named here
calculate_activation_statistics 3 other papers
1 dated earlier, 2 later
1904.01310 (2019-04)
DM-GAN: Dynamic Memory Generative Adversarial Networks for Text-to-Image Synthesis
2311.06322 (2023-11)
Post-training Quantization for Text-to-Image Diffusion Models with Progressive Calibration and Activation Relaxing
2311.15941 (2023-11)
Tell2Design: A Dataset for Language-Guided Floor Plan Generation
run 3 other papers
1 dated earlier, 2 later
2109.08267 (2021-09)
CompilerGym: Robust, Performant Compiler Optimization Environments for AI Research
2312.08977 (2023-12)
Weighted Ensemble Models Are Strong Continual Learners
2408.08295 (2024-08)
SLCA++: Unleash the Power of Sequential Fine-tuning for Continual Learning with Pre-training
cp_path_to_dir 2 other papers
1 dated earlier, 1 later
2102.12092 (2021-02)
Zero-Shot Text-to-Image Generation
2206.00100 (2022-06)
VALHALLA: Visual Hallucination for Machine Translation
get_activations 2 other papers
1 dated earlier, 1 later
1904.01310 (2019-04)
DM-GAN: Dynamic Memory Generative Adversarial Networks for Text-to-Image Synthesis
2311.15941 (2023-11)
Tell2Design: A Dataset for Language-Guided Floor Plan Generation
get_trainable_params 2 other papers
1 dated earlier, 1 later
2102.12092 (2021-02)
Zero-Shot Text-to-Image Generation
2206.00100 (2022-06)
VALHALLA: Visual Hallucination for Machine Translation

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 2026-11-21.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Recently, DALL-E, a multimodal transformer language model, and its variants, including diffusion models, have shown high-quality text-to-image generation capabilities. However, despite the realistic image generation results, there has not been a detailed analysis of how to evaluate such models. In this work, we investigate the visual reasoning capabilities and social biases of different text-to-image models, covering both multimodal transformer language models and diffusion models. First, we measure three visual reasoning skills: object recognition, object counting, and spatial relation understanding. For this, we propose PaintSkills, a compositional diagnostic evaluation dataset that measures these skills. Despite the high-fidelity image generation capability, a large gap exists between the performance of recent models and the upper bound accuracy in object counting and spatial relation understanding skills. Second, we assess the gender and skin tone biases by measuring the gender/skin tone distribution of generated images across various professions and attributes. We demonstrate that recent text-to-image generation models learn specific biases about gender and skin tone from web image-text pairs. We hope our work will help guide future progress in improving text-to-image generation models on visual reasoning skills and learning socially unbiased representations. Code and data: https://github.com/j-min/DallEval

For agents

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

get_harvested_code_for_paper("2202.04053")
get_code_for_paper("2202.04053")
have("2202.04053")

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

curl https://syntology.ai/api/ran/2202.04053.json

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

[![Syntology run record](https://syntology.ai/api/ran/2202.04053.svg)](https://syntology.ai/paper/2202.04053)

Connect an agent — have() is free.