SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2302.00070 · 2023

Debiasing Vision-Language Models via Biased Prompts

arXiv · PDF · Open in the Atlas

Code that ran

10 of 17 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 7 have no contract verdict; 3 failures of our instrument; 7 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/2302.00070.json · badge: /api/ran/2302.00070.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
chingyaoc/debias_vl canonical (classifier label)
classifier basis not recorded on the link
10 ran · 7 no recorded run
FunctionStatusWhere it lives
basic_clean Ran
ran; our contract draft was wrong, not the code
chingyaoc/debias_vl/discriminative/clip/simple_tokenizer.py
pointer only (licence: MIT); the stored copy get_code serves, from ltpwy/MSCI, is pointer only (licence: NONE) · get_code("98f385d847636a3e")
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 ltpwy/MSCI, 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.

evaluate_waterbirds_predictions Ran
ran on a synthesized input
chingyaoc/debias_vl/discriminative/evaluate.py
code served (permissive licence) · get_code("35dc929d8eb8b23f")
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. 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

# Create a minimal dataset class that provides the required attributes
class MinimalDataset:
    def __init__(self):
        # Create targets and spurious attributes for a small batch
        # 4 samples total
        self.targets_all = {
            'target': np.array([0, 0, 1, 1]),
            'spurious': np.array([0, 1, 0, 1])
        }

# Create a minimal dataloader class that provides the required attributes
class MinimalDataLoader:
    def __init__(self):
        self.dataset = MinimalDataset()

# Create predictions as a numpy array matching the targets
# These should be shape (4,) with values in {0, 1}
predictions = np.array([0, 0, 1, 1])

# Create the dataloader
dataloader = MinimalDataLoader()

# Define arguments for the function
ARGS = {
    'predictions': predictions,
    'dataloader': dataloader
}
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
get_pairs Ran
ran; our contract draft was wrong, not the code
chingyaoc/debias_vl/discriminative/clip/simple_tokenizer.py
code served (permissive licence) · get_code("d919ae32e5e4e616")
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 set. 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)

# The function get_pairs takes a word parameter
# A word is represented as a tuple of symbols (variable-length strings)
# The function iterates through the word and creates pairs of consecutive symbols

word = ("hello", "world", "test")

ARGS = {
    "word": word
}
Setup block drafted by
a language model, Claude Haiku 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 set — a set output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a set 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
get_transform_celeba Ran
ran on a synthesized input
chingyaoc/debias_vl/discriminative/datasets/celebA.py
code served (permissive licence) · get_code("2ace1a1b277d6dff")
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 Compose. 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 torchvision.transforms as transforms

torch.manual_seed(0)

# For get_transform_celeba, we need:
# - model_type: str
# - train: bool

ARGS = {
    'model_type': 'resnet50',
    'train': True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy, pandas, torch, torchvision — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Compose — a Compose output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Compose 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
load_dataloaders Ran
ran on a synthesized input
chingyaoc/debias_vl/discriminative/datasets/waterbirds.py
code served (permissive licence) · get_code("075a8ffd49d83ae8")
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 os
import tempfile
import numpy as np
import pandas as pd
from PIL import Image
import torch
from torchvision import transforms

torch.manual_seed(0)

# Create a temporary directory structure with mock Waterbirds data
temp_dir = tempfile.mkdtemp()
data_subdir = os.path.join(temp_dir, 'y_place')
os.makedirs(data_subdir, exist_ok=True)

# Create minimal metadata.csv
metadata_data = {
    'img_filename': ['img_0.jpg', 'img_1.jpg', 'img_2.jpg', 'img_3.jpg'],
    'y': [0, 1, 0, 1],
    'place': [0, 1, 1, 0],
    'split': [0, 0, 1, 2]
}
metadata_df = pd.DataFrame(metadata_data)
metadata_df.to_csv(os.path.join(data_subdir, 'metadata.csv'), index=False)

# Create minimal image files
for i in range(4):
    img = Image.new('RGB', (4, 3), color=(73, 109, 137))
    img.save(os.path.join(data_subdir, f'img_{i}.jpg'))

# Create a mock args object
class Args:
    def __init__(self):
        self.root_dir = temp_dir
        self.target_name = 'y'
        self.confounder_names = ['place']
        self.arch = 'resnet50'
        self.num_workers = 0
        self.dataset = 'waterbirds'

args = Args()

# Create a simple transform
transform = transforms.Compose([
    transforms.Resize((4, 3)),
    transforms.ToTensor(),
])

ARGS = {
    'args': args,
    'train_shuffle': True,
    'val_correlation': None,
    'transform': transform
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy, pandas, torch, torchvision — 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
load_waterbirds Ran
ran on a synthesized input
chingyaoc/debias_vl/discriminative/datasets/waterbirds.py
code served (permissive licence) · get_code("6aa90b926f026199")
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 os
import tempfile
import numpy as np
import pandas as pd
from PIL import Image
import torch
from torchvision import transforms

torch.manual_seed(0)

# Create a temporary directory structure for the dataset
temp_dir = tempfile.mkdtemp()
root_dir = temp_dir
data_dir = os.path.join(root_dir, 'y_place')
os.makedirs(data_dir, exist_ok=True)

# Create mock metadata CSV
metadata = pd.DataFrame({
    'img_filename': ['img_0.jpg', 'img_1.jpg', 'img_2.jpg', 'img_3.jpg'],
    'y': [0, 1, 0, 1],
    'place': [0, 1, 1, 0],
    'split': [0, 0, 1, 2]
})
metadata.to_csv(os.path.join(data_dir, 'metadata.csv'), index=False)

# Create mock image files
for i in range(4):
    img = Image.new('RGB', (4, 3), color=(i*60, i*60, i*60))
    img.save(os.path.join(data_dir, f'img_{i}.jpg'))

# Create mock transform
def mock_transform(img):
    return transforms.ToTensor()(img)

# Create mock args object
class Args:
    def __init__(self):
        self.root_dir = root_dir
        self.target_name = 'y'
        self.confounder_names = ['place']
        self.arch = 'resnet50'
        self.num_workers = 0
        self.dataset = 'waterbirds'

args = Args()

ARGS = {
    'args': args,
    'train_shuffle': True,
    'transform': mock_transform
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy, pandas, torch, torchvision — 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
log_metrics Ran
ran on a synthesized input
chingyaoc/debias_vl/discriminative/utils/logging.py
code served (permissive licence) · get_code("fe037f6fee7c4ffb")
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
from collections import defaultdict

class Args:
    def __init__(self):
        self.wilds_dataset = False
        self.verbose = True
        self.results_dict = defaultdict(list)

# Create sample metrics with group-wise correct/total counts
# Structure: (loss, correct, total, correct_by_groups, total_by_groups)
# correct_by_groups and total_by_groups are 2D arrays: [num_groups][num_subgroups]

correct_by_groups = np.array([
    [45, 38, 52],
    [28, 35, 41],
    [33, 29, 47]
], dtype=np.float32)

total_by_groups = np.array([
    [50, 40, 60],
    [50, 40, 60],
    [50, 40, 60]
], dtype=np.float32)

train_metrics = (0.25, 100, 500, correct_by_groups.copy(), total_by_groups.copy())
val_metrics = (0.35, 85, 400, correct_by_groups.copy() - 5, total_by_groups.copy())
test_metrics = (0.40, 80, 450, correct_by_groups.copy() - 8, total_by_groups.copy())

args = Args()

ARGS = {
    'train_metrics': train_metrics,
    'val_metrics': val_metrics,
    'test_metrics': test_metrics,
    'epoch': 1,
    'dataset_ix': 0,
    'args': args
}
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
summarize_acc Ran
ran on a synthesized input
chingyaoc/debias_vl/discriminative/utils/logging.py
code served (permissive licence) · get_code("5b644caaec292937")
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 numpy as np

# Create sample data for correct_by_groups and total_by_groups
# These are nested lists where each element represents a group
# Structure: correct_by_groups[yix][aix] = number of correct predictions
#            total_by_groups[yix][aix] = total predictions for that group

correct_by_groups = [
    [85, 92, 78],  # y_group 0
    [88, 95, 81],  # y_group 1
    [79, 86, 74]   # y_group 2
]

total_by_groups = [
    [100, 110, 95],   # y_group 0
    [102, 115, 98],   # y_group 1
    [105, 108, 100]   # y_group 2
]

ARGS = {
    'correct_by_groups': correct_by_groups,
    'total_by_groups': total_by_groups,
    'stdout': True,
    'return_groups': False
}
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 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
summarize_acc_from_predictions Ran
ran on a synthesized input
chingyaoc/debias_vl/discriminative/utils/logging.py
code served (permissive licence) · get_code("5d1b64d4af5d5aa1")
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 numpy as np

class SimpleDataset:
    def __init__(self, num_samples=10, num_classes=3):
        self.targets_all = {
            'target': np.array([0, 1, 2, 0, 1, 2, 0, 1, 2, 0], dtype=np.int64)[:num_samples],
            'spurious': np.array([0, 1, 0, 1, 0, 1, 0, 1, 0, 1], dtype=np.int64)[:num_samples]
        }

class SimpleDataLoader:
    def __init__(self, dataset):
        self.dataset = dataset

class SimpleArgs:
    def __init__(self):
        self.num_classes = 3

# Create dataset with 10 samples
dataset = SimpleDataset(num_samples=10, num_classes=3)
dataloader = SimpleDataLoader(dataset)

# Create predictions matching the number of samples
predictions = np.array([0, 1, 2, 0, 1, 2, 0, 1, 2, 0], dtype=np.int64)

args = SimpleArgs()

ARGS = {
    'predictions': predictions,
    'dataloader': dataloader,
    'args': args,
    'stdout': True,
    'return_groups': False
}
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 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
whitespace_clean Ran
ran; our contract draft was wrong, not the code
chingyaoc/debias_vl/discriminative/clip/simple_tokenizer.py
pointer only (licence: MIT); the stored copy get_code serves, from ltpwy/MSCI, is pointer only (licence: NONE) · get_code("9542161e9640b858")
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 ltpwy/MSCI, 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.

build_model Not yet run
unverified (harvested, no recorded run)
chingyaoc/debias_vl/discriminative/clip/model.py
code served (permissive licence) · get_code("d461b1d7c46b95ae")
How it ran

no invocation recorded.

get_backbone Not yet run
unverified (harvested, no recorded run)
chingyaoc/debias_vl/discriminative/clip/model.py
code served (permissive licence) · get_code("1beef3eb483e1be4")
How it ran

no invocation recorded.

get_clip_embeddings Not yet run
unverified (harvested, no recorded run)
chingyaoc/debias_vl/discriminative/network/clip.py
code served (permissive licence) · get_code("033d70f48893755c")
How it ran

no invocation recorded.

get_dataset_embeddings Not yet run
unverified (harvested, no recorded run)
chingyaoc/debias_vl/discriminative/network/clip.py
code served (permissive licence) · get_code("b1b3243fc36c4849")
How it ran

no invocation recorded.

load Not yet run
unverified (harvested, no recorded run)
chingyaoc/debias_vl/discriminative/clip/clip.py
code served (permissive licence) · get_code("5397b32b2b3d700f")
How it ran

no invocation recorded.

load_celeba Not yet run
unverified (harvested, no recorded run)
chingyaoc/debias_vl/discriminative/datasets/celebA.py
code served (permissive licence) · get_code("279b050fc395872d")
How it ran

no invocation recorded.

load_dataloaders Not yet run
unverified (harvested, no recorded run)
chingyaoc/debias_vl/discriminative/datasets/celebA.py
code served (permissive licence) · get_code("c7a01bf9cb95620b")
How it ran

no invocation recorded.

Also present in other papers' code

Of 17 samples lifted from this paper's linked repositories, 17 carry a function identity and 0 do not. Of the 17 distinct functions with an identity, 5 are also present in other papers whose code we hold (5 with at least one paper dated earlier, 3 the same month, 5 later) and 12 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
get_pairs 258 other papers
44 dated earlier, 1 the same month, 205 later, 7 same year, order not determinable, 1 undated · 4 only as an assembly brick
1802.06474 (2018-02)
A Closed-form Solution to Photorealistic Image Stylization
1803.01128 (2018-03)
Seq2Sick: Evaluating the Robustness of Sequence-to-Sequence Models with Adversarial Examples
1809.00125 (2018-09)
Simple Fusion: Return of the Language Model
and 255 more not named here
whitespace_clean 231 other papers
29 dated earlier, 1 the same month, 195 later, 5 same year, order not determinable, 1 undated · 4 only as an assembly brick
1901.07441 (2019-01)
PadChest: A large chest x-ray image dataset with multi-label annotated reports
2107.06383 (2021-07)
How Much Can CLIP Benefit Vision-and-Language Tasks?
2108.10904 (2021-08)
SimVLM: Simple Visual Language Model Pretraining with Weak Supervision
and 228 more not named here
basic_clean 218 other papers
29 dated earlier, 1 the same month, 182 later, 5 same year, order not determinable, 1 undated · 4 only as an assembly brick
1901.07441 (2019-01)
PadChest: A large chest x-ray image dataset with multi-label annotated reports
2107.06383 (2021-07)
How Much Can CLIP Benefit Vision-and-Language Tasks?
2108.10904 (2021-08)
SimVLM: Simple Visual Language Model Pretraining with Weak Supervision
and 215 more not named here
build_model 119 other papers
24 dated earlier, 90 later, 3 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
load 17 other papers
1 dated earlier, 15 later, 1 same year, order not determinable
2211.16582 (2022-11)
SinDDM: A Single Image Denoising Diffusion Model
2305.17100 (2023-05)
BiomedGPT: A Generalist Vision-Language Foundation Model for Diverse Biomedical Tasks
and 14 more not named here

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

Machine learning models have been shown to inherit biases from their training datasets. This can be particularly problematic for vision-language foundation models trained on uncurated datasets scraped from the internet. The biases can be amplified and propagated to downstream applications like zero-shot classifiers and text-to-image generative models. In this study, we propose a general approach for debiasing vision-language foundation models by projecting out biased directions in the text embedding. In particular, we show that debiasing only the text embedding with a calibrated projection matrix suffices to yield robust classifiers and fair generative models. The proposed closed-form solution enables easy integration into large-scale pipelines, and empirical results demonstrate that our approach effectively reduces social bias and spurious correlation in both discriminative and generative vision-language models without the need for additional data or training.

For agents

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

get_harvested_code_for_paper("2302.00070")
get_code_for_paper("2302.00070")
have("2302.00070")

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

curl https://syntology.ai/api/ran/2302.00070.json

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

[![Syntology run record](https://syntology.ai/api/ran/2302.00070.svg)](https://syntology.ai/paper/2302.00070)

Connect an agent — have() is free.