SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1502.03044 · ICML · 2015

Show, Attend and Tell: Neural Image Caption Generation with Visual Attention

arXiv · PDF · Open in the Atlas

Code that ran

58 of 84 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; 3 honoured the contract we drafted, 0 violated it and 41 have no contract verdict; 14 failures of our instrument; 26 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/1502.03044.json · badge: /api/ran/1502.03044.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
shaox192/NeuralGuidance community (classifier label)
classifier basis not recorded on the link
12 ran · 3 no recorded run
leob03/Image_captionning community (classifier label)
classifier basis not recorded on the link
10 ran · 2 no recorded run
yurayli/image_caption_pytorch community (classifier label)
classifier basis not recorded on the link
6 ran · 5 no recorded run
coldmanck/show-attend-and-tell community (classifier label)
classifier basis not recorded on the link
6 ran · 0 no recorded run
martincss/first-captioning community (classifier label)
classifier basis not recorded on the link
5 ran · 1 no recorded run
djain454/Show-Attend-and-Tell-Neural-Image-Caption-Generation-with-Visual-Attention community (classifier label)
classifier basis not recorded on the link
4 ran · 0 no recorded run
LinXueyuanStdio/LaTeX_OCR community (classifier label)
classifier basis not recorded on the link
3 ran · 1 no recorded run
Japanese-Image-Captioning/SAT-for-Japanese community (classifier label)
classifier basis not recorded on the link
2 ran · 1 no recorded run
Zhangyuewei98/arctic-captions-py3 reimplementation (classifier label)
classifier basis: classifier_2025
2 ran · 1 no recorded run
lvyufeng/image_captioning_mindspore community (classifier label)
classifier basis not recorded on the link
2 ran · 0 no recorded run
Aaatresh/Seeing-through-sound community (classifier label)
classifier basis not recorded on the link
1 ran · 1 no recorded run
ishritam/Image-captioning-with-visual-attention reimplementation (classifier label)
classifier basis: classifier_2025
1 ran · 0 no recorded run
krasserm/fairseq-image-captioning extension (classifier label)
classifier basis: classifier_2025
1 ran · 0 no recorded run
codekansas/keras-language-modeling community (classifier label)
classifier basis not recorded on the link
1 ran · 0 no recorded run
TheZino/Image_captioning_SATell reimplementation (classifier label)
classifier basis: classifier_2025
1 ran · 0 no recorded run
jaykshirsagar05/captionify reimplementation (classifier label)
classifier basis: classifier_2025
1 ran · 0 no recorded run
DeepRNN/image_captioning community (classifier label)
classifier basis not recorded on the link
0 ran · 3 no recorded run
maxy0524/image_captioning community (classifier label)
classifier basis not recorded on the link
0 ran · 3 no recorded run
fuqichen1998/eecs442-final-project-show-and-tell community (classifier label)
classifier basis not recorded on the link
0 ran · 2 no recorded run
NicholasKX/ShowAttendTell community (classifier label)
classifier basis not recorded on the link
0 ran · 1 no recorded run
Skumarr53/Image-Caption-Generation-using-Fastai reimplementation (classifier label)
classifier basis: classifier_2025
0 ran · 1 no recorded run
Anjaney1999/image-captioning-seqgan reimplementation (classifier label)
classifier basis: classifier_2025
0 ran · 1 no recorded run
FunctionStatusWhere it lives
BLEU_score Ran
ran on a synthesized input
yurayli/image_caption_pytorch/eval_metrics.py
code served (permissive licence) · get_code("4e8ebb8599fb73e6")
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 float, 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 nltk

# Download required NLTK data
try:
    nltk.data.find('tokenizers/punkt')
except LookupError:
    nltk.download('punkt', quiet=True)

# Create sample ground-truth and predicted captions
gt_caption = "a cat sitting on a mat <start> the cat is very cute <end>"
sample_caption = "a cat is sitting on the mat <start> very cute <end>"

ARGS = {
    "gt_caption": gt_caption,
    "sample_caption": sample_caption,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
nltk — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float, 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
Magnitude Ran
ran on a synthesized input
shaox192/NeuralGuidance/analysis/smoothness.py
code served (permissive licence) · get_code("88947ef0639ed6bb")
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

torch.manual_seed(0)

# g1 is a 2D tensor that will be squared, summed along dimension 1, and then meaned
# The function does: torch.sum(g1 ** 2, 1).mean() * 2 * 200
# g1 needs to be a tensor with shape (batch_size, features)
# Using batch_size=2, features=4 as specified

g1 = torch.randn(2, 4)

ARGS = {
    "g1": g1
}
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
accuracy Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
shaox192/NeuralGuidance/analysis/utils.py
code served (permissive licence) · get_code("b0f936d4d6ae3b8c")
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 list, 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 torch

torch.manual_seed(0)

# Create output tensor: batch_size=2, num_classes=4
# This represents logits from a model
output = torch.randn(2, 4)

# Create target tensor: batch_size=2
# These are class indices (0-3 for 4 classes)
target = torch.tensor([1, 3])

# topk parameter: tuple of k values to compute accuracy for
topk = (1, 2)

ARGS = {
    "output": output,
    "target": target,
    "topk": topk,
}
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 list, ndim 2, dtype family float — a list 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
accuracy Ran
ran; our contract draft was wrong, not the code
Japanese-Image-Captioning/SAT-for-Japanese/utils.py
code served (permissive licence) · get_code("53daad1d7b99fbdf")
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 float, 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

torch.manual_seed(0)

# Construct small tensors for the function
batch_size = 2
num_classes = 4
k = 2

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

# targets: true labels with shape (batch_size,)
targets = torch.randint(0, num_classes, (batch_size,))

ARGS = {
    "scores": scores,
    "targets": targets,
    "k": k,
}
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 float, 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
accuracy Ran
ran on a synthesized input
djain454/Show-Attend-and-Tell-Neural-Image-Caption-Generation-with-Visual-Attention/utils.py
code served (permissive licence) · get_code("fc7e660e3ad2dd0e")
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 float, 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 as nn

torch.manual_seed(0)

# Create predictions tensor: shape (batch_size=2, num_classes=4)
# topk will select k largest values along dimension 1
preds = torch.randn(2, 4)

# Create targets tensor: shape (batch_size=2,)
# Values should be class indices in range [0, num_classes)
targets = torch.tensor([0, 2])

# k for top-k accuracy
k = 2

ARGS = {
    "preds": preds,
    "targets": targets,
    "k": k
}
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 float, 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
cal_noise_ceiling_single_vox Ran
ran on a synthesized input
shaox192/NeuralGuidance/analysis/noise_ceiling.py
code served (permissive licence) · get_code("ecb092757c4db425")
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 sample data for the function
# The function expects data with shape (n_voxels, n_samples)
# We'll create a small dataset with 3 voxels and 5 samples
np.random.seed(0)
data = np.random.randn(3, 5)

ARGS = {
    "data": data
}
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
calc_max_length Ran
ran, honoured the contract we drafted
ishritam/Image-captioning-with-visual-attention/image_cap.py
pointer only (licence: NONE) · get_code("af5b85578e4e6f1b")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, ishritam/Image-captioning-with-visual-attention 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.

calculate_caption_lengths Ran
ran on a synthesized input
djain454/Show-Attend-and-Tell-Neural-Image-Caption-Generation-with-Visual-Attention/utils.py
code served (permissive licence) · get_code("9f843b78744f8406")
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 int, ndim 0, dtype family int, 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

torch.manual_seed(0)

# Create a word dictionary with special tokens
word_dict = {
    '<start>': 0,
    '<eos>': 1,
    '<pad>': 2,
    'hello': 3,
    'world': 4,
    'test': 5
}

# Create captions as a list of lists of token indices
# Each inner list represents tokens in a caption
captions = [
    [0, 3, 4, 1],      # [<start>, hello, world, <eos>]
    [0, 5, 3, 1, 2]    # [<start>, test, hello, <eos>, <pad>]
]

ARGS = {
    'word_dict': word_dict,
    'captions': captions
}
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 int, ndim 0, dtype family int, 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
caption_features Ran
ran on a synthesized input
martincss/first-captioning/preprocess_tokenize_captions.py
code served (permissive licence) · get_code("68d2533729c9efa1")
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 int, shape (2, 5) 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 tensorflow as tf

class MockTokenizer:
    def texts_to_sequences(self, texts):
        # Return sequences of different lengths for realistic padding test
        return [
            [1, 2, 3],
            [4, 5]
        ]

train_captions = [
    "a cat sitting on a mat",
    "a dog running"
]

tokenizer = MockTokenizer()

maxlen = 5

ARGS = {
    "train_captions": train_captions,
    "tokenizer": tokenizer,
    "maxlen": maxlen
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
tensorflow — 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 int, shape (2, 5) — shape (2, 5) 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
compute_numeric_gradient Ran
ran on a synthesized input
leob03/Image_captionning/utils/grad.py
code served (permissive licence) · get_code("112c9a161e488e7c")
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 2, dtype family float, shape (2, 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 torch.nn as nn

torch.manual_seed(0)

# Simple function that takes a tensor and returns a scalar
def simple_f(x):
    """A simple quadratic function: sum of x^2"""
    return (x ** 2).sum()

# Create a small input tensor
x = torch.randn(2, 3, requires_grad=False)

# Optional upstream gradient (None will default to ones)
dLdf = None

# Finite difference step size
h = 1e-07

ARGS = {
    "f": simple_f,
    "x": x,
    "dLdf": dLdf,
    "h": h,
}
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 2, dtype family float, shape (2, 3) — shape (2, 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
cook_refs Ran
ran; our contract draft was wrong, not the code
coldmanck/show-attend-and-tell/utils/coco/pycocoevalcap/cider/cider_scorer.py
pointer only (licence: MIT); the stored copy get_code serves, from wenhuchen/data-to-text-evaluation-metric, is pointer only (licence: NOASSERTION) · get_code("4c1b107103b7c856")
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 wenhuchen/data-to-text-evaluation-metric, 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.

cook_refs Ran
ran; our contract draft was wrong, not the code
coldmanck/show-attend-and-tell/utils/coco/pycocoevalcap/bleu/bleu_scorer.py
code served (permissive licence) · get_code("7586e1d32f7325f7")
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.

from collections import defaultdict

ARGS = {
    'refs': [
        'the quick brown fox jumps over the lazy dog',
        'a quick brown fox leaps over a lazy dog'
    ],
    'eff': 'shortest',
    'n': 4
}
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
cook_test Ran
ran; our contract draft was wrong, not the code
coldmanck/show-attend-and-tell/utils/coco/pycocoevalcap/cider/cider_scorer.py
pointer only (licence: MIT); the stored copy get_code serves, from wenhuchen/data-to-text-evaluation-metric, is pointer only (licence: NOASSERTION) · get_code("3573c4bc85f65dfd")
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 wenhuchen/data-to-text-evaluation-metric, 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.

cook_test Ran
ran on a synthesized input
coldmanck/show-attend-and-tell/utils/coco/pycocoevalcap/bleu/bleu_scorer.py
code served (permissive licence) · get_code("99e54843980299e2")
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.

from collections import defaultdict

def precook(s, n=4, out=False):
    """Takes a string as input and returns an object that can be given to
    either cook_refs or cook_test. This is optional: cook_refs and cook_test
    can take string arguments as well."""
    words = s.split()
    counts = defaultdict(int)
    for k in range(1, n + 1):
        for i in range(len(words) - k + 1):
            ngram = tuple(words[i:i + k])
            counts[ngram] += 1
    return (len(words), counts)

def cook_test(test, refs, eff=None, n=4):
    """Takes a test sentence and returns an object that
    encapsulates everything that BLEU needs to know about it."""
    reflen, refmaxcounts = refs
    testlen, counts = precook(test, n, True)
    result = {}
    if eff == 'closest':
        result['reflen'] = min(((abs(l - testlen), l) for l in reflen))[1]
    else:
        result['reflen'] = reflen
    result['testlen'] = testlen
    result['guess'] = [max(0, testlen - k + 1) for k in range(1, n + 1)]
    result['correct'] = [0] * n
    for ngram, count in counts.items():
        result['correct'][len(ngram) - 1] += min(refmaxcounts.get(ngram, 0), count)
    return result

# Build test arguments
test_string = "the quick brown fox jumps over the lazy dog"

# Create reference data using precook
ref_string_1 = "the quick brown fox jumps over the lazy dog"
ref_string_2 = "a fast brown fox leaps over the lazy dog"
reflen_1, refcounts_1 = precook(ref_string_1, n=4)
reflen_2, refcounts_2 = precook(ref_string_2, n=4)

# Merge reference counts (take max for each ngram)
refmaxcounts = defaultdict(int)
for ngram, count in refcounts_1.items():
    refmaxcounts[ngram] = max(refmaxcounts[ngram], count)
for ngram, count in refcounts_2.items():
    refmaxcounts[ngram] = max(refmaxcounts[ngram], count)

# Create refs tuple
refs = ([reflen_1, reflen_2], refmaxcounts)

ARGS = {
    'test': test_string,
    'refs': refs,
    'eff': None,
    'n': 4
}
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 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
correlation Ran
ran on a synthesized input
shaox192/NeuralGuidance/neural_guidance_training/model.py
code served (permissive licence) · get_code("e9b0f77f1b690e9a")
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 1, dtype family float, shape (2,) 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

torch.manual_seed(0)

# The function expects two tensors X and Y
# Looking at the einsum operations: 'ab,ab->a'
# This means:
# - X and Y should be 2D tensors with shape (batch_size, feature_dim)
# - The operation computes correlations along the batch dimension
# - batch_size=2, feature_dim=4 as per requirements

X = torch.randn(2, 4)
Y = torch.randn(2, 4)

ARGS = {
    'X': X,
    'Y': Y
}
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 1, dtype family float, shape (2,) — shape (2,) 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
create_padding_mask Ran
ran; our contract draft was wrong, not the code
krasserm/fairseq-image-captioning/model/caption.py
code served (permissive licence) · get_code("f93be7ac25d7dac1")
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 Tensor, ndim 2, dtype family float, shape (2, 4) 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

torch.manual_seed(0)

# Create small tensors for testing
# src_tokens: shape (batch_size, seq_length) - token indices
batch_size = 2
seq_length = 4
src_tokens = torch.randint(0, 100, (batch_size, seq_length), dtype=torch.long)

# src_lengths: list/tensor of actual lengths for each sequence in batch
# These should be <= seq_length
src_lengths = [2, 3]

ARGS = {
    "src_tokens": src_tokens,
    "src_lengths": src_lengths,
}
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 Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) 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
extract_features Ran
ran on a synthesized input
yurayli/image_caption_pytorch/solver.py
code served (permissive licence) · get_code("92cdbcd5d83f0f6d")
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 2, dtype family float, shape (2, 6) 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 as nn

torch.manual_seed(0)

class SimpleNet(nn.Module):
    def __init__(self):
        super().__init__()
        self.layer1 = nn.Linear(4, 8)
        self.layer2 = nn.Linear(8, 6)
        self.layer3 = nn.Linear(6, 3)
    
    def forward(self, x):
        x = self.layer1(x)
        x = self.layer2(x)
        x = self.layer3(x)
        return x

net = SimpleNet()
xs = torch.randn(2, 4)

ARGS = {
    "net": net,
    "xs": xs
}
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 2, dtype family float, shape (2, 6) — shape (2, 6) 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
feed_forward_net Ran
ran on a synthesized input
yurayli/image_caption_pytorch/encode_image.py
code served (permissive licence) · get_code("aa4791e2bffdf413")
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 2, dtype family float, shape (2, 10) 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 as nn
import torch.nn.functional as F

torch.manual_seed(0)

# Create a minimal Inception V3-like net
class MinimalInceptionV3(nn.Module):
    def __init__(self):
        super().__init__()
        self.module1 = nn.Conv2d(3, 8, kernel_size=3, padding=1)
        self.module2 = nn.Conv2d(8, 8, kernel_size=3, padding=1)
        self.module3 = nn.Conv2d(8, 8, kernel_size=3, padding=1)
        self.module4 = nn.Conv2d(8, 16, kernel_size=3, padding=1)
        self.module5 = nn.Conv2d(16, 16, kernel_size=3, padding=1)
        self.module6 = nn.Conv2d(16, 32, kernel_size=3, padding=1)
        self.module7 = nn.Conv2d(32, 32, kernel_size=3, padding=1)
        self.module8 = nn.Conv2d(32, 32, kernel_size=3, padding=1)
        self.module9 = nn.Conv2d(32, 32, kernel_size=3, padding=1)
        self.module10 = nn.Conv2d(32, 16, kernel_size=1)
    
    def forward(self, x):
        x = self.module1(x)
        return x

# Create a minimal DenseNet161-like net
class MinimalDenseNet(nn.Module):
    def __init__(self):
        super().__init__()
        self.features = nn.Sequential(
            nn.Conv2d(3, 16, kernel_size=3, padding=1),
            nn.Conv2d(16, 32, kernel_size=3, padding=1)
        )
    
    def forward(self, x):
        return self.features(x)

# Create a minimal ResNet101-like net
class MinimalResNet(nn.Module):
    def __init__(self):
        super().__init__()
        self.conv1 = nn.Conv2d(3, 16, kernel_size=3, padding=1)
        self.layer1 = nn.Conv2d(16, 32, kernel_size=3, padding=1)
        self.layer2 = nn.Conv2d(32, 32, kernel_size=3, padding=1)
        self.avgpool = nn.AdaptiveAvgPool2d((1, 1))
        self.fc = nn.Linear(32, 10)
    
    def forward(self, x):
        x = self.conv1(x)
        return x

# Create a minimal VGG16-like net
class MinimalVGG(nn.Module):
    def __init__(self):
        super().__init__()
        self.features = nn.Sequential(
            nn.Conv2d(3, 16, kernel_size=3, padding=1),
            nn.Conv2d(16, 32, kernel_size=3, padding=1)
        )
        self.classifier = nn.Sequential(
            nn.Linear(32 * 4 * 4, 64),
            nn.Linear(64, 10)
        )
    
    def forward(self, x):
        x = self.features(x)
        x = x.view(x.size(0), -1)
        x = self.classifier(x)
        return x

# Small test tensor (batch=2, channels=3, height=4, width=4)
xs = torch.randn(2, 3, 4, 4)

ARGS = {
    'xs': xs,
    'net': MinimalVGG(),
    'pre_net': 'vgg16',
    'spatial': False
}
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 2, dtype family float, shape (2, 10) — shape (2, 10) 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
getFileNameToSave Ran
ran on a synthesized input
LinXueyuanStdio/LaTeX_OCR/visualize_attention.py
code served (permissive licence) · get_code("891b8f71cfafca9a")
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

path_to_save_attention = "/tmp/attention_output"
i = 0

ARGS = {
    "path_to_save_attention": path_to_save_attention,
    "i": i
}
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
getOutArray Ran
ran on a synthesized input
LinXueyuanStdio/LaTeX_OCR/visualize_attention.py
code served (permissive licence) · get_code("d7419b2156f83b93")
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, 4) 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
import torch

torch.manual_seed(0)

# Create a small attention vector with shape (1, height*width)
# Using att_h=3, att_w=4, so flattened size is 12
att_h = 3
att_w = 4
attention_size = att_h * att_w

# Create attention vector with values between 0 and 1
attentionVector = np.random.rand(1, attention_size).astype(np.float32)

ARGS = {
    "attentionVector": attentionVector,
    "att_w": att_w,
    "att_h": att_h
}
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, 4) — shape (3, 4) 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
getWH Ran
ran on a synthesized input
LinXueyuanStdio/LaTeX_OCR/visualize_attention.py
code served (permissive licence) · get_code("54e482c9c86eec94")
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 int; 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

ARGS = {
    'img_w': 640,
    'img_h': 480
}
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 int — 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
get_dist_matrix Ran
ran on a synthesized input
shaox192/NeuralGuidance/analysis/RSA.py
code served (permissive licence) · get_code("8a9b2d5877d086a3")
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 2, dtype family float, shape (2, 2) 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

torch.manual_seed(0)

# Create a small data matrix: 2 samples, 4 features
data_mat = torch.randn(2, 4)

# Test with 'corr' method
ARGS = {
    'data_mat': data_mat,
    'method': 'corr'
}
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 2, dtype family float, shape (2, 2) — shape (2, 2) 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_hypotheses Ran
ran on a synthesized input
lvyufeng/image_captioning_mindspore/src/utils.py
code served (permissive licence) · get_code("2d423582a2c44126")
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 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 numpy as np

torch.manual_seed(0)

# Create a small predictions tensor (batch_size=2, sequence_length=4)
# The function calls .tolist() on predictions and indexes with cap_lens
predictions = torch.tensor([
    [1, 2, 3, 4],
    [5, 6, 7, 8]
], dtype=torch.long)

# Create cap_lens list with different lengths for each batch element
# These values indicate how many elements to keep from each prediction
cap_lens = [2, 3]

ARGS = {
    "predictions": predictions,
    "cap_lens": cap_lens
}
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 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
get_model_corr_mat Ran
ran on a synthesized input
shaox192/NeuralGuidance/analysis/RSA.py
code served (permissive licence) · get_code("70e878dc080d6418")
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 2, dtype family float, shape (4, 4) 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

torch.manual_seed(0)

# The function expects lbs (labels) to be a 1D tensor of shape (batch_size,)
# It performs element-wise comparison to create a same_label_matrix
# Then uses torch.where to create a model_matrix

lbs = torch.tensor([0, 1, 0, 1], dtype=torch.long)

ARGS = {
    "lbs": lbs
}
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 2, dtype family float, shape (4, 4) — shape (4, 4) 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_model_weights Ran
ran on a synthesized input
shaox192/NeuralGuidance/manifold_guidance/man_cotrain.py
code served (permissive licence) · get_code("29c867d79b8c8f8e")
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 ResNet18_Weights. 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.models as models

torch.manual_seed(0)

ARGS = {
    'arch': 'resnet18'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, torchvision — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ResNet18_Weights — a ResNet18_Weights output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ResNet18_Weights 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_references Ran
ran on a synthesized input
lvyufeng/image_captioning_mindspore/src/utils.py
code served (permissive licence) · get_code("9ee8094f76ff694d")
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 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 numpy as np

torch.manual_seed(0)

# Build word_map with required special tokens
word_map = {
    '<start>': 0,
    '<pad>': 1,
    'hello': 2,
    'world': 3,
    'test': 4,
}

# Build all_captions as a 2D tensor where each row contains encoded captions
# Shape: (num_images, max_captions_per_image, max_caption_length)
# But looking at the function, it treats all_captions[j] as a list of captions
# So we need shape (num_images, num_captions, caption_length)
all_captions = torch.tensor([
    [[0, 2, 3, 1], [0, 4, 1, 1]],  # Image 0: 2 captions of length 4
    [[0, 2, 4, 1], [0, 3, 1, 1]],  # Image 1: 2 captions of length 4
], dtype=torch.long)

ARGS = {
    'all_captions': all_captions,
    'word_map': word_map,
}
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 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
get_repeated_image Ran
ran on a synthesized input
shaox192/NeuralGuidance/analysis/noise_ceiling.py
code served (permissive licence) · get_code("cd8511add99a8df2")
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 numpy as np

torch.manual_seed(0)

# Analyze the function:
# - full_beta_data: not used in the function body, can be any object
# - ordering: used with len() and indexing [trial_id], should be a list-like of image IDs

# Create minimal valid arguments
full_beta_data = None  # Not used in function body

# ordering: list of image IDs, where some images appear exactly 3 times
# Create an ordering where:
# - image_id 0 appears at trials 0, 1, 2 (3 times)
# - image_id 1 appears at trials 3, 4, 5 (3 times)
# - image_id 2 appears at trials 6, 7 (2 times - should be filtered out)
# - image_id 3 appears at trial 8 (1 time - should be filtered out)
ordering = [0, 0, 0, 1, 1, 1, 2, 2, 3]

ARGS = {
    "full_beta_data": full_beta_data,
    "ordering": ordering
}
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 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
get_toy_data Ran
ran on a synthesized input
leob03/Image_captionning/main_func.py
code served (permissive licence) · get_code("c5e0eb463e9afc32")
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 json
import tempfile
import os

# Create a temporary JSON file with toy data
temp_dir = tempfile.mkdtemp()
temp_file = os.path.join(temp_dir, 'final_data.json')

toy_data = {
    "samples": [
        {"id": 1, "value": "test1"},
        {"id": 2, "value": "test2"}
    ],
    "metadata": {"version": "1.0"}
}

with open(temp_file, 'w') as f:
    json.dump(toy_data, f)

ARGS = {
    'path': temp_file
}
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 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
image_fnames_captions Ran
ran on a synthesized input
martincss/first-captioning/coco_utils.py
code served (permissive licence) · get_code("c369482485646bd1")
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 json
import tempfile
from pathlib import Path

# Create a temporary directory and files for testing
temp_dir = tempfile.TemporaryDirectory()
temp_path = Path(temp_dir.name)

# Create a mock captions file
captions_data = {
    "annotations": [
        {
            "image_id": 1,
            "caption": "a dog running in the park"
        },
        {
            "image_id": 2,
            "caption": "a cat sitting on a mat"
        },
        {
            "image_id": 123456789012,
            "caption": "a bird flying in the sky"
        }
    ]
}

captions_file = temp_path / "captions.json"
with open(captions_file, 'w') as f:
    json.dump(captions_data, f)

# Create images directory
images_dir = temp_path / "images"
images_dir.mkdir(exist_ok=True)

# Create dummy image files
for image_id in [1, 2, 123456789012]:
    img_path = images_dir / ('COCO_train2014_{:012d}.jpg'.format(image_id))
    img_path.touch()

ARGS = {
    "captions_file": str(captions_file),
    "images_dir": images_dir,
    "partition": "train"
}
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 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
itemlist Ran
ran; our contract draft was wrong, not the code
Zhangyuewei98/arctic-captions-py3/capgen.py
pointer only (licence: NONE) · get_code("8569db484e9d39b6")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, Zhangyuewei98/arctic-captions-py3 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.

load Ran
ran on a synthesized input
codekansas/keras-language-modeling/generate_insurance_qa_embeddings.py
code served (permissive licence) · get_code("eccb788cbbf8f688")
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 os
import pickle
import tempfile

# Create a temporary directory to use as the path
temp_dir = tempfile.mkdtemp()

# Create a sample object to pickle
sample_data = {
    'model_weights': [1.0, 2.0, 3.0],
    'config': {'learning_rate': 0.001, 'epochs': 10},
    'results': [0.95, 0.96, 0.97]
}

# Pickle the sample data to a file
filename = 'test_data.pkl'
filepath = os.path.join(temp_dir, filename)
with open(filepath, 'wb') as f:
    pickle.dump(sample_data, f)

# Define the arguments for the load function
ARGS = {
    'path': temp_dir,
    'name': filename
}
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 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
load_coco_captions Ran
ran on a synthesized input
leob03/Image_captionning/main_func.py
code served (permissive licence) · get_code("088be7e4343bda87")
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 tempfile
import os

torch.manual_seed(0)

# Create a minimal COCO dataset dictionary and save it to a temporary file
temp_dir = tempfile.mkdtemp()
temp_path = os.path.join(temp_dir, 'coco.pt')

# Create minimal data dictionary matching the expected structure
data_dict = {
    'train_images': torch.randn(2, 3, 112, 112),  # 2 training images, 3 channels, 112x112
    'val_images': torch.randn(3, 3, 112, 112),    # 3 validation images, 3 channels, 112x112
    'train_captions': torch.randint(0, 100, (2, 15)),  # 2 captions, max length 15
    'val_captions': torch.randint(0, 100, (3, 15)),    # 3 captions, max length 15
    'vocab': {
        'idx_to_token': ['<pad>', '<unk>', 'the', 'a', 'cat', 'dog', 'sits', 'on', 'mat'],
        'token_to_idx': {
            '<pad>': 0,
            '<unk>': 1,
            'the': 2,
            'a': 3,
            'cat': 4,
            'dog': 5,
            'sits': 6,
            'on': 7,
            'mat': 8
        }
    }
}

# Save the data dictionary to the temporary path
torch.save(data_dict, temp_path)

# Define ARGS with the path to the temporary file
ARGS = {
    'path': temp_path
}
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
load_embeddings Ran
ran; our contract draft was wrong, not the code
Japanese-Image-Captioning/SAT-for-Japanese/utils.py
code served (permissive licence) · get_code("4cc9097b30b644e0")
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 torch
import numpy as np
import tempfile
import os

torch.manual_seed(0)

# Create a temporary embedding file in GloVe format
temp_dir = tempfile.mkdtemp()
emb_file = os.path.join(temp_dir, "embeddings.txt")

# Create a small GloVe-format embedding file
# Format: word dim1 dim2 dim3 dim4
with open(emb_file, 'w') as f:
    f.write("apple 0.1 0.2 0.3 0.4\n")
    f.write("banana 0.5 0.6 0.7 0.8\n")
    f.write("cherry 0.9 1.0 1.1 1.2\n")
    f.write("date 1.3 1.4 1.5 1.6\n")

# Create a word_map that maps words to indices
word_map = {
    "apple": 0,
    "banana": 1,
    "cherry": 2,
    "date": 3,
    "elderberry": 4,
}

ARGS = {
    "emb_file": emb_file,
    "word_map": word_map,
}
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
load_image Ran
ran on a synthesized input
martincss/first-captioning/preprocess_encode_images.py
code served (permissive licence) · get_code("b0bb3778f118bff1")
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 tensorflow as tf
import tempfile
import os
from PIL import Image
import numpy as np

# Create a temporary JPEG image file for testing
temp_dir = tempfile.mkdtemp()
img_array = np.random.randint(0, 256, (224, 224, 3), dtype=np.uint8)
img = Image.fromarray(img_array)
image_path = os.path.join(temp_dir, "test_image.jpg")
img.save(image_path)

ARGS = {
    "image_path": image_path,
    "cnn_encoder": "InceptionV3"
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy, tensorflow — 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
load_img_folder_ls Ran
ran on a synthesized input
shaox192/NeuralGuidance/manifold_guidance/man_data_loader.py
code served (permissive licence) · get_code("b8d0de38fa1b3c8b")
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 list, 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 tempfile
import os

# Create a temporary text file with folder paths
with tempfile.NamedTemporaryFile(mode='w', delete=False, suffix='.txt') as f:
    f.write("/path/to/folder1: some metadata\n")
    f.write("/path/to/folder2: some metadata\n")
    f.write("/path/to/folder3: some metadata\n")
    txt_file_path = f.name

try:
    ARGS = {
        'txt_pth': txt_file_path
    }
finally:
    # Clean up the temporary file after defining ARGS
    pass
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 list, ndim 1, dtype family float — a list 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_textfiles Ran
ran; our contract draft was wrong, not the code
Zhangyuewei98/arctic-captions-py3/metrics.py
pointer only (licence: NONE) · get_code("05ec0ad6e8114d01")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, Zhangyuewei98/arctic-captions-py3 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.

make_tokenizer Ran
ran on a synthesized input
martincss/first-captioning/preprocess_tokenize_captions.py
code served (permissive licence) · get_code("6954f2198f23dbee")
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 Tokenizer. 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 tensorflow as tf

# Create small sample captions for training the tokenizer
train_captions = [
    "a cat sitting on the mat",
    "a dog running in the park",
    "a bird flying in the sky",
    "the cat is sleeping",
    "dogs love to play fetch"
]

# Set top_k to a reasonable value for vocabulary size
top_k = 20

ARGS = {
    "train_captions": train_captions,
    "top_k": top_k
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tokenizer — a Tokenizer output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Tokenizer 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
pickle_load Ran
ran on a synthesized input
shaox192/NeuralGuidance/analysis/utils.py
code served (permissive licence) · get_code("198d90979fd0346e")
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 pickle as pkl
import tempfile
import os

# Create a temporary file and pickle some data into it
temp_dir = tempfile.mkdtemp()
temp_file = os.path.join(temp_dir, "test_data.pkl")

# Create sample data to pickle
sample_data = {
    "key1": "value1",
    "key2": [1, 2, 3, 4, 5],
    "key3": {"nested": "dict"}
}

# Write the pickled data to the temporary file
with open(temp_file, 'wb') as f:
    pkl.dump(sample_data, f)

# Define ARGS
ARGS = {
    "fpth": temp_file
}
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 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
pil_loader Ran
ran, honoured the contract we drafted
shaox192/NeuralGuidance/manifold_guidance/man_data_loader.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from zhentingwang/diagnosis, is pointer only (licence: NONE) · get_code("f6be833ba409926f")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (Apache-2.0), but get_code serves the stored copy, first harvested from zhentingwang/diagnosis, 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.

pil_loader Ran
ran, honoured the contract we drafted
djain454/Show-Attend-and-Tell-Neural-Image-Caption-Generation-with-Visual-Attention/dataset.py
code served (permissive licence) · get_code("f321f54723433661")
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 Image. 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 numpy as np
from PIL import Image
import io
import os
import tempfile

# Create a temporary directory for test files
temp_dir = tempfile.mkdtemp()

# Create a simple RGB image and save it to a temporary file
img = Image.new('RGB', (10, 10), color=(73, 109, 137))
test_image_path = os.path.join(temp_dir, 'test_image.jpg')
img.save(test_image_path)

ARGS = {
    'path': test_image_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
PIL, numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
precook Ran
ran; our contract draft was wrong, not the code
coldmanck/show-attend-and-tell/utils/coco/pycocoevalcap/bleu/bleu_scorer.py
code served (permissive licence) · get_code("ec4448373d72d636")
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 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.

ARGS = {
    "s": "the quick brown fox jumps over the lazy dog",
    "n": 4,
    "out": False
}
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 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
precook Ran
ran; our contract draft was wrong, not the code
coldmanck/show-attend-and-tell/utils/coco/pycocoevalcap/cider/cider_scorer.py
pointer only (licence: MIT); the stored copy get_code serves, from wenhuchen/data-to-text-evaluation-metric, is pointer only (licence: NOASSERTION) · get_code("9560e3e3e41bc1b2")
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 wenhuchen/data-to-text-evaluation-metric, 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.

process_caption_tokens Ran
ran on a synthesized input
djain454/Show-Attend-and-Tell-Neural-Image-Caption-Generation-with-Visual-Attention/generate_json_data.py
code served (permissive licence) · get_code("369a98c031b8ad9d")
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 list, ndim 2, dtype family int; 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 numpy as np

torch.manual_seed(0)

# Build a minimal word_dict with required special tokens and some sample words
word_dict = {
    '<start>': 0,
    '<eos>': 1,
    '<pad>': 2,
    '<unk>': 3,
    'cat': 4,
    'dog': 5,
    'sits': 6,
    'runs': 7,
}

# Build caption_tokens as a list of token sequences
caption_tokens = [
    ['cat', 'sits'],
    ['dog', 'runs', 'fast'],
]

# Set max_length to accommodate the longest caption plus start and eos tokens
max_length = 5

ARGS = {
    'caption_tokens': caption_tokens,
    'word_dict': word_dict,
    'max_length': max_length,
}
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 list, ndim 2, dtype family int — a list 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
rel_error Ran
ran on a synthesized input
leob03/Image_captionning/utils/grad.py
code served (permissive licence) · get_code("0a60ca96331ba158")
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 float, 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

torch.manual_seed(0)

# Create two small tensors of the same shape
x = torch.randn(2, 4)
y = torch.randn(2, 4)

ARGS = {
    "x": x,
    "y": y,
    "eps": 1e-10
}
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 float, 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
remove_special_tokens Ran
ran; our contract draft was wrong, not the code
yurayli/image_caption_pytorch/eval_metrics.py
code served (permissive licence) · get_code("9bfeeae84da4c830")
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 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.

ARGS = {
    "caption_text": "a <start> photo of a cat <end> sitting on <unk> the mat"
}
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 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
rnn_forward Ran
ran on a synthesized input
leob03/Image_captionning/rnn_lstm_captioning.py
code served (permissive licence) · get_code("818e7744734a9021")
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 torch

torch.manual_seed(0)

# Define dimensions
N = 2  # batch size
T = 3  # sequence length
D = 4  # input dimension
H = 5  # hidden dimension

# Create input tensor of shape (N, T, D)
x = torch.randn(N, T, D)

# Create initial hidden state of shape (N, H)
h0 = torch.randn(N, H)

# Create weight matrix for input-to-hidden of shape (D, H)
Wx = torch.randn(D, H)

# Create weight matrix for hidden-to-hidden of shape (H, H)
Wh = torch.randn(H, H)

# Create bias vector of shape (H,)
b = torch.randn(H)

ARGS = {
    'x': x,
    'h0': h0,
    'Wx': Wx,
    'Wh': Wh,
    'b': b
}
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 — 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
rnn_step_backward Ran
ran on a synthesized input
leob03/Image_captionning/rnn_lstm_captioning.py
code served (permissive licence) · get_code("7459c63f19bd1a20")
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 torch

torch.manual_seed(0)

# Define dimensions
N = 2  # batch size
D = 4  # input dimension
H = 3  # hidden dimension

# Create cache tuple containing (x, prev_h, Wx, Wh, b)
x = torch.randn(N, D)
prev_h = torch.randn(N, H)
Wx = torch.randn(D, H)
Wh = torch.randn(H, H)
b = torch.randn(H)

cache = (x, prev_h, Wx, Wh, b)

# Create gradient of loss with respect to next hidden state
dnext_h = torch.randn(N, H)

ARGS = {
    "dnext_h": dnext_h,
    "cache": cache
}
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 — 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
rnn_step_forward Ran
ran on a synthesized input
leob03/Image_captionning/rnn_lstm_captioning.py
code served (permissive licence) · get_code("a4639ecce23676b1")
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 torch

torch.manual_seed(0)

# Define dimensions
N = 2  # batch size
D = 4  # input dimension
H = 3  # hidden dimension

# Create input data for this timestep
x = torch.randn(N, D)

# Create hidden state from previous timestep
prev_h = torch.randn(N, H)

# Create weight matrices
Wx = torch.randn(D, H)
Wh = torch.randn(H, H)

# Create bias vector
b = torch.randn(H)

ARGS = {
    'x': x,
    'prev_h': prev_h,
    'Wx': Wx,
    'Wh': Wh,
    'b': b
}
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 — 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
run_clean Ran
ran on a synthesized input
shaox192/NeuralGuidance/analysis/RSA.py
code served (permissive licence) · get_code("e2a84722a01bf8a0")
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

torch.manual_seed(0)

# Create a minimal model that takes images and returns neural outputs
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 3)
    
    def forward(self, x):
        # x shape: (batch_size, 4)
        return self.linear(x)

# Create a minimal data loader as a list of (images, labels) tuples
val_loader = [
    (torch.randn(2, 4), torch.randint(0, 3, (2,))),
    (torch.randn(2, 4), torch.randint(0, 3, (2,))),
    (torch.randn(2, 4), torch.randint(0, 3, (2,))),
    (torch.randn(2, 4), torch.randint(0, 3, (2,))),
]

# Create the model
model = SimpleModel()

ARGS = {
    'val_loader': val_loader,
    'model': model,
}
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 — 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
split_captions Ran
ran on a synthesized input
yurayli/image_caption_pytorch/data_utils.py
code served (permissive licence) · get_code("5ce371a11110cca1")
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 tempfile
import os

# Create a temporary directory and files for testing
temp_dir = tempfile.mkdtemp()
path = temp_dir + "/"

# Create the Flickr8k.token.txt file with sample data
token_file_content = """image1.jpg#0	a dog is running in the park
image1.jpg#1	the dog runs fast
image1.jpg#2	a brown dog playing
image2.jpg#0	cats sitting on the fence
image2.jpg#1	two cats on wood
image3.jpg#0	a bird flying high
image3.jpg#1	the bird soars through sky"""

with open(path + 'Flickr8k.token.txt', 'w') as f:
    f.write(token_file_content)

# Define training and test file lists (image IDs without the #N suffix)
trn_files = ['image1.jpg', 'image2.jpg']
test_files = ['image3.jpg']

ARGS = {
    'path': path,
    'trn_files': trn_files,
    'test_files': test_files
}
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
split_hparams Ran
ran on a synthesized input
martincss/first-captioning/grid_search.py
code served (permissive licence) · get_code("8b4a39f185ba405c")
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 numpy as np

ARGS = {
    'hparams': {
        'embedding_dim': 64,
        'units': 128,
        'lstm_units': 256,
        'n_layers_init': 2,
        'n_layers_att': 3,
        'init_dropout': 0.1,
        'attn_dropout': 0.2,
        'lstm_dropout': 0.15,
        'logit_dropout': 0.05,
        'l1_reg': 0.001,
        'l2_reg': 0.002,
        'lambda_reg': 0.5,
        'optimizer': 'adam',
        'learning_rate': 0.001
    }
}
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 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
split_image_files Ran
ran on a synthesized input
yurayli/image_caption_pytorch/data_utils.py
code served (permissive licence) · get_code("b6424d334cb20b0b")
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 os
import tempfile
import shutil

# Create a temporary directory structure to mock the Flickr8k dataset
temp_dir = tempfile.mkdtemp()
path = temp_dir + '/'

# Create the Flicker8k_Dataset directory with some image files
dataset_dir = os.path.join(temp_dir, 'Flicker8k_Dataset')
os.makedirs(dataset_dir, exist_ok=True)

# Create some dummy image files
image_files = ['image1.jpg', 'image2.jpg', 'image3.jpg', 'image4.jpg', 'image5.jpg']
for img in image_files:
    open(os.path.join(dataset_dir, img), 'w').close()

# Create train images file
train_images = ['image1.jpg', 'image2.jpg']
with open(os.path.join(temp_dir, 'Flickr_8k.trainImages.txt'), 'w') as f:
    f.write('\n'.join(train_images))

# Create dev images file
dev_images = ['image3.jpg']
with open(os.path.join(temp_dir, 'Flickr_8k.devImages.txt'), 'w') as f:
    f.write('\n'.join(dev_images))

# Create test images file
test_images = ['image4.jpg', 'image5.jpg']
with open(os.path.join(temp_dir, 'Flickr_8k.testImages.txt'), 'w') as f:
    f.write('\n'.join(test_images))

ARGS = {
    'path': path
}
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 — 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
tensor_to_image Ran
ran on a synthesized input
leob03/Image_captionning/utils/utils.py
code served (permissive licence) · get_code("bc260615679a81dd")
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 3, dtype family int, shape (4, 5, 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)

# Create a valid tensor input: shape (3, H, W) with elements in [0, 1]
H, W = 4, 5
tensor = torch.rand(3, H, W)

ARGS = {
    'tensor': tensor
}
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 3, dtype family int, shape (4, 5, 3) — shape (4, 5, 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
to_var Ran
ran; our contract draft was wrong, not the code
TheZino/Image_captioning_SATell/models.py
pointer only (licence: NONE) · get_code("502c50812b7f31ab")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, TheZino/Image_captioning_SATell 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.

train_captioner Ran
ran on a synthesized input
leob03/Image_captionning/main_func.py
code served (permissive licence) · get_code("0a5ea4832aa6d5f3")
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 torch
import torch.nn as nn
import math

torch.manual_seed(0)

class SimpleCaptioner(nn.Module):
    """Minimal captioner model that accepts images and captions and returns a loss."""
    def __init__(self, image_dim, caption_dim):
        super().__init__()
        self.fc = nn.Linear(image_dim, caption_dim)
        self.loss_fn = nn.MSELoss()
    
    def forward(self, images, captions):
        # images: (batch_size, image_dim)
        # captions: (batch_size, caption_dim)
        # Return a scalar loss
        output = self.fc(images)
        loss = self.loss_fn(output, captions)
        return loss

# Create minimal tensors
image_dim = 4
caption_dim = 3
batch_size = 2
num_samples = 6

# Image data: (num_samples, image_dim)
image_data = torch.randn(num_samples, image_dim)

# Caption data: (num_samples, caption_dim)
caption_data = torch.randn(num_samples, caption_dim)

# Create model
model = SimpleCaptioner(image_dim, caption_dim)

ARGS = {
    'model': model,
    'image_data': image_data,
    'caption_data': caption_data,
    'num_epochs': 2,
    'batch_size': batch_size,
    'learning_rate': 0.01,
    'lr_decay': 0.95,
    'device': torch.device('cpu')
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
matplotlib, 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
unpickle Ran
ran on a synthesized input
Aaatresh/Seeing-through-sound/vOICe/data_generators/data_generator_cifar10.py
code served (permissive licence) · get_code("bfc501c46e8ad1f9")
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 pickle
import tempfile
import os

# Create a temporary pickle file with sample data
temp_dir = tempfile.mkdtemp()
temp_file = os.path.join(temp_dir, "test_batch")

# Create sample data matching CIFAR-10 batch format
sample_data = {
    b'batch_label': b'training batch 1',
    b'labels': [0, 1, 2, 3, 4],
    b'data': bytes(range(256)) * 50,  # Mock image data
    b'filenames': [b'file1.png', b'file2.png', b'file3.png', b'file4.png', b'file5.png']
}

# Pickle the sample data
with open(temp_file, 'wb') as f:
    pickle.dump(sample_data, f)

ARGS = {
    'file': temp_file
}
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 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
visualize_dataset Ran
ran on a synthesized input
leob03/Image_captionning/utils/utils.py
code served (permissive licence) · get_code("f534fbe3633b3692")
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 3, dtype family int, shape (20, 14, 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)

# Create small tensors for X_data: [batch, 3, width, height]
# batch=10, 3 channels, 4x4 spatial dimensions
X_data = torch.rand(10, 3, 4, 4)

# Create corresponding labels: [batch]
# Labels for 3 classes (0, 1, 2)
y_data = torch.tensor([0, 0, 0, 0, 1, 1, 1, 2, 2, 2])

# Number of samples per class to visualize
samples_per_class = 2

# List of class names
class_list = ['class_0', 'class_1', 'class_2']

ARGS = {
    'X_data': X_data,
    'y_data': y_data,
    'samples_per_class': samples_per_class,
    'class_list': class_list
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
matplotlib, numpy, torch, torchvision — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 3, dtype family int, shape (20, 14, 3) — shape (20, 14, 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
word_for_id Ran
ran; our contract draft was wrong, not the code
jaykshirsagar05/captionify/fastapi/caption.py
code served (permissive licence) · get_code("8a04d895237560a5")
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 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
import numpy as np

class StandInTokenizer:
    def __init__(self):
        self.word_index = {
            'hello': 1,
            'world': 2,
            'python': 3,
            'code': 4
        }

tokenizer = StandInTokenizer()
integer = 2

ARGS = {
    'integer': integer,
    'tokenizer': tokenizer
}
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 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
accuracy Not yet run
unverified (harvested, no recorded run)
NicholasKX/ShowAttendTell/utils.py
code served (permissive licence) · get_code("95216f51978817ce")
How it ran

no invocation recorded.

attention_visualizer Not yet run
unverified (harvested, no recorded run)
leob03/Image_captionning/utils/utils.py
code served (permissive licence) · get_code("754d759415cbca01")
How it ran

no invocation recorded.

Showing the first 60. An agent gets every one from get_harvested_code_for_paper("1502.03044").

Also present in other papers' code

Of 84 samples lifted from this paper's linked repositories, 84 carry a function identity and 0 do not. Of the 82 distinct functions with an identity, 18 are also present in other papers whose code we hold (6 with at least one paper dated earlier, 0 the same month, 17 later) and 64 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
116 later · 8 only as an assembly brick
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
1906.05849 (2019-06)
Contrastive Multiview Coding
and 113 more not named here
pil_loader 66 other papers
66 later
1710.10196 (2017-10)
Progressive Growing of GANs for Improved Quality, Stability, and Variation
1711.02512 (2017-11)
Fine-tuning CNN Image Retrieval with No Human Annotation
1902.09212 (2019-02)
Deep High-Resolution Representation Learning for Human Pose Estimation
and 63 more not named here
cook_refs 25 other papers
1 dated earlier, 24 later
1411.5726 (2014-11)
CIDEr: Consensus-based Image Description Evaluation
1503.01817 (2015-03)
YFCC100M: The New Data in Multimedia Research
1506.08909 (2015-06)
The Ubuntu Dialogue Corpus: A Large Dataset for Research in Unstructured Multi-Turn Dialogue Systems
and 22 more not named here
cook_test 22 other papers
1 dated earlier, 21 later
1411.5726 (2014-11)
CIDEr: Consensus-based Image Description Evaluation
1503.01817 (2015-03)
YFCC100M: The New Data in Multimedia Research
1506.08909 (2015-06)
The Ubuntu Dialogue Corpus: A Large Dataset for Research in Unstructured Multi-Turn Dialogue Systems
and 19 more not named here
precook 22 other papers
1 dated earlier, 21 later
1411.4555 (2014-11)
Show and Tell: A Neural Image Caption Generator
1611.09268 (2016-11)
MS MARCO: A Human Generated MAchine Reading COmprehension Dataset
1611.09571 (2016-11)
Predicting Human Eye Fixations via an LSTM-based Saliency Attentive Model
and 19 more not named here
cook_refs 21 other papers
21 later
1611.09268 (2016-11)
MS MARCO: A Human Generated MAchine Reading COmprehension Dataset
1611.09571 (2016-11)
Predicting Human Eye Fixations via an LSTM-based Saliency Attentive Model
1806.00807 (2018-06)
Learning Semantic Sentence Embeddings using Sequential Pair-wise Discriminator
and 18 more not named here
precook 14 other papers
1 dated earlier, 13 later
1411.5726 (2014-11)
CIDEr: Consensus-based Image Description Evaluation
1806.00807 (2018-06)
Learning Semantic Sentence Embeddings using Sequential Pair-wise Discriminator
1809.01797 (2018-09)
Describing a Knowledge Base
and 11 more not named here
pil_loader 11 other papers
11 later
2111.15667 (2021-11)
Adaptive Token Sampling For Efficient Vision Transformers
2206.02967 (2022-06)
Masked Unsupervised Self-training for Label-free Image Classification
2207.01145 (2022-07)
Memory Population in Continual Learning via Outlier Elimination
and 8 more not named here
to_var 9 other papers
9 later · 1 only as an assembly brick
1511.06349 (2015-11)
Generating Sentences from a Continuous Space
1704.03162 (2017-04)
Show, Ask, Attend, and Answer: A Strong Baseline For Visual Question Answering
1905.01639 (2019-05)
Deep Video Inpainting
and 6 more not named here
accuracy 8 other papers
8 later
1606.02960 (2016-06)
Sequence-to-Sequence Learning as Beam-Search Optimization
2305.11692 (2023-05)
Surgical-VQLA: Transformer with Gated Vision-Language Embedding for Visual Question Localized-Answering in Robotic Surgery
2307.05182 (2023-07)
CAT-ViL: Co-Attention Gated Vision-Language Embedding for Visual Question Localized-Answering in Robotic Surgery
and 5 more not named here
precook 8 other papers
1 dated earlier, 7 later
1411.5726 (2014-11)
CIDEr: Consensus-based Image Description Evaluation
1503.01817 (2015-03)
YFCC100M: The New Data in Multimedia Research
1506.08909 (2015-06)
The Ubuntu Dialogue Corpus: A Large Dataset for Research in Unstructured Multi-Turn Dialogue Systems
and 5 more not named here
cook_test 2 other papers
2 later
1806.00807 (2018-06)
Learning Semantic Sentence Embeddings using Sequential Pair-wise Discriminator
2111.12085 (2021-11)
UniTAB: Unifying Text and Box Outputs for Grounded Vision-Language Modeling
itemlist 2 other papers
2 later
1603.06147 (2016-03)
A Character-Level Decoder without Explicit Segmentation for Neural Machine Translation
1610.03017 (2016-10)
Fully Character-Level Neural Machine Translation without Explicit Segmentation
load_embeddings 2 other papers
2 later
1606.02960 (2016-06)
Sequence-to-Sequence Learning as Beam-Search Optimization
2311.10774 (2023-11)
MMC: Advancing Multimodal Chart Understanding with Large-scale Instruction Tuning
unpickle 2 other papers
2 later
1806.00451 (2018-06)
Do CIFAR-10 Classifiers Generalize to CIFAR-10?
1904.12043 (2019-04)
Dynamic Mini-batch SGD for Elastic Distributed Training: Learning in the Limbo of Resources
unzip 2 other papers
2 later
1603.06147 (2016-03)
A Character-Level Decoder without Explicit Segmentation for Neural Machine Translation
1610.03017 (2016-10)
Fully Character-Level Neural Machine Translation without Explicit Segmentation
remove_special_tokens 1 other paper
1 dated earlier
1411.4555 (2014-11)
Show and Tell: A Neural Image Caption Generator
sample_from_start 1 other paper
1 later
1804.00861 (2018-04)
Generating Diverse and Accurate Visual Captions by Comparative Adversarial 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 2026-11-21.

Repositories linked to this paper

More are linked than shown here.

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

Abstract

Inspired by recent work in machine translation and object detection, we introduce an attention based model that automatically learns to describe the content of images. We describe how we can train this model in a deterministic manner using standard backpropagation techniques and stochastically by maximizing a variational lower bound. We also show through visualization how the model is able to automatically learn to fix its gaze on salient objects while generating the corresponding words in the output sequence. We validate the use of attention with state-of-the-art performance on three benchmark datasets: Flickr8k, Flickr30k and MS COCO.

For agents

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

get_harvested_code_for_paper("1502.03044")
get_code_for_paper("1502.03044")
have("1502.03044")

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

curl https://syntology.ai/api/ran/1502.03044.json

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

[![Syntology run record](https://syntology.ai/api/ran/1502.03044.svg)](https://syntology.ai/paper/1502.03044)

Connect an agent — have() is free.