SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1910.09387 · 2019

Clotho: An Audio Captioning Dataset

arXiv · PDF · Open in the Atlas

Code that ran

19 of 21 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 12 have no contract verdict; 6 failures of our instrument; 2 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/1910.09387.json · badge: /api/ran/1910.09387.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
richermans/AudioCaption canonical (classifier label)
classifier basis: census
6 ran · 1 no recorded run
wsntxxn/AudioCaption community (classifier label)
classifier basis not recorded on the link
13 ran · 1 no recorded run
FunctionStatusWhere it lives
NAME_TO_WIDTH Ran
ran on a synthesized input
wsntxxn/AudioCaption/captioning/models/eff_at_encoder.py
code served (permissive licence) · get_code("36732bf2fa117007")
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

def NAME_TO_WIDTH(name):
    mn_map = {'mn01': 0.1, 'mn02': 0.2, 'mn04': 0.4, 'mn05': 0.5, 'mn06': 0.6, 'mn08': 0.8, 'mn10': 1.0, 'mn12': 1.2, 'mn14': 1.4, 'mn16': 1.6, 'mn20': 2.0, 'mn30': 3.0, 'mn40': 4.0}
    dymn_map = {'dymn04': 0.4, 'dymn10': 1.0, 'dymn20': 2.0}
    try:
        if name.startswith('dymn'):
            w = dymn_map[name[:6]]
        else:
            w = mn_map[name[:4]]
    except:
        w = 1.0
    return w

ARGS = {
    'name': 'mn10'
}
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
cnn_out_size Ran
ran on a synthesized input
wsntxxn/AudioCaption/captioning/models/eff_at_encoder.py
code served (permissive licence) · get_code("96fd7d5df7f15091")
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 math

def cnn_out_size(in_size, padding, dilation, kernel, stride):
    s = in_size + 2 * padding - dilation * (kernel - 1) - 1
    return math.floor(s / stride + 1)

ARGS = {
    "in_size": 28,
    "padding": 1,
    "dilation": 1,
    "kernel": 3,
    "stride": 1
}
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 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
coco_score Ran
ran on a synthesized input
richermans/AudioCaption/utils/human_eval.py
code served (permissive licence) · get_code("42b965c67572da92")
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 1, dtype family float, shape (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

class MockScorer:
    def __init__(self, method_name='Bleu'):
        self._method_name = method_name
    
    def method(self):
        return self._method_name
    
    def compute_score(self, refs, res):
        if self._method_name == 'Bleu':
            score = [0.9, 0.8, 0.7, 0.6]
        else:
            score = 0.85
        return score, None

refs = {
    'audio_1': ['caption one', 'caption two', 'caption three'],
    'audio_2': ['another caption one', 'another caption two', 'another caption three']
}

scorer = MockScorer(method_name='Bleu')

ARGS = {
    'refs': refs,
    'scorer': scorer
}
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 ndarray, ndim 1, dtype family float, shape (4,) — shape (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
collate_fn Ran
ran on a synthesized input
richermans/AudioCaption/datasets/SJTUDataSet.py
code served (permissive licence) · get_code("47fe80f824147b4d")
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 function. 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)

# length_idxs is a list/set of indices indicating which data elements 
# should have their lengths tracked and returned
length_idxs = {0, 2}

# Create sample data batches
# Each batch item is a tuple containing multiple elements
# Some are tensors (with varying sequence lengths), some are scalars/other types

# Batch item 1: (tensor with seq_len=5, tensor with seq_len=3, scalar)
batch_item_1 = (
    torch.randn(5, 4),      # sequence tensor, length 5
    torch.randn(3, 4),      # sequence tensor, length 3
    torch.tensor(7)         # scalar tensor
)

# Batch item 2: (tensor with seq_len=4, tensor with seq_len=2, scalar)
batch_item_2 = (
    torch.randn(4, 4),      # sequence tensor, length 4
    torch.randn(2, 4),      # sequence tensor, length 2
    torch.tensor(9)         # scalar tensor
)

# Batch item 3: (tensor with seq_len=6, tensor with seq_len=4, scalar)
batch_item_3 = (
    torch.randn(6, 4),      # sequence tensor, length 6
    torch.randn(4, 4),      # sequence tensor, length 4
    torch.tensor(11)        # scalar tensor
)

data_batches = [batch_item_1, batch_item_2, batch_item_3]

ARGS = {
    "length_idxs": length_idxs
}
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
conv_conv_block Ran
ran on a synthesized input
wsntxxn/AudioCaption/captioning/models/crnn_trm_encoder.py
code served (permissive licence) · get_code("7c55804fff5d4cc0")
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 Sequential. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# The function takes two parameters: in_channel and out_channel
# These are integers representing the number of input and output channels for the Conv2d layers

ARGS = {
    "in_channel": 3,
    "out_channel": 64,
}
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 Sequential — a Sequential output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Sequential output the recorded ndim 1 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
cs_interspeech Ran
ran on a synthesized input
wsntxxn/AudioCaption/captioning/models/conv_filter_pruning.py
code served (permissive licence) · get_code("8ebc0d04df3640f5")
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 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
from scipy.spatial import distance

# Read the function to understand what it needs:
# cs_interspeech(Z) expects Z to be a 4D array
# It unpacks as: d, c, a, b = np.shape(Z)
# Then reshapes to (d, c, -1) and transposes to (a*b, c, d)
# For each i in range(d), it processes A[:, :, i] which is shape (a*b, c)
# The rank1_apporx function does SVD on (a*b, c) matrix

# Let's create a small 4D array with distinct dimensions
# d=3, c=2, a=2, b=2 means shape (3, 2, 2, 2)
Z = np.random.randn(3, 2, 2, 2)

ARGS = {
    "Z": Z
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — 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 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
generate_length_mask Ran
ran on a synthesized input
wsntxxn/AudioCaption/captioning/models/flops_counting_model.py
code served (permissive licence) · get_code("99b3ce923e424db3")
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 (3, 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 torch

torch.manual_seed(0)

# lens: can be a list, tuple, or tensor of sequence lengths
# max_length: optional int or None
lens = [3, 5, 2]
max_length = None

ARGS = {
    "lens": lens,
    "max_length": max_length
}
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 (3, 5) — shape (3, 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
gumbel_softmax_sample Ran
ran on a synthesized input
wsntxxn/AudioCaption/captioning/losses/loss.py
code served (permissive licence) · get_code("e698f855f3588740")
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, 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 a small logit tensor (batch_size=2, num_classes=4)
logit = torch.randn(2, 4)

# Create a temperature scalar
temperature = 0.5

ARGS = {
    "logit": logit,
    "temperature": temperature,
}
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, 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
make_divisible Ran
ran, honoured the contract we drafted
wsntxxn/AudioCaption/captioning/models/eff_at_encoder.py
code served (permissive licence) · get_code("9e10f323aeefe252")
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
import numpy as np

# Analyze the function:
# - v: float - a numeric value to be made divisible
# - divisor: int - the divisor to make v divisible by
# - min_value: Optional[int] - minimum value constraint (defaults to divisor if None)
# The function performs arithmetic operations and returns an int

ARGS = {
    "v": 47.5,
    "divisor": 8,
    "min_value": None,
}
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 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
open_or_fd Ran
ran; our contract draft was wrong, not the code
richermans/AudioCaption/utils/kaldi_io.py
pointer only (licence: MIT); the stored copy get_code serves, from cffan/corp, is pointer only (licence: NONE) · get_code("df07b4faa7609d90")
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 cffan/corp, 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.

operator_norm_pruning Ran
ran on a synthesized input
wsntxxn/AudioCaption/captioning/models/conv_filter_pruning.py
code served (permissive licence) · get_code("18d5889c7ea06109")
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 1, dtype family int, shape (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 numpy as np
import torch

# Analyze the function:
# - Takes a single parameter: weight
# - weight is expected to have shape (c_out, c_in, h, w)
# - The function reshapes it to (c_out, c_in, h*w)
# - It performs SVD operations and matrix operations
# - Returns an index array

# Create a small weight tensor with distinct dimensions
# Shape: (c_out=3, c_in=2, h=4, w=5) -> reshapes to (3, 2, 20)
torch.manual_seed(0)

weight = np.random.randn(3, 2, 4, 5).astype(np.float32)

ARGS = {
    "weight": weight
}
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 1, dtype family int, shape (3,) — shape (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
pad_unsort_packed_sequence Ran
ran; our contract draft was wrong, not the code
wsntxxn/AudioCaption/captioning/models/hf_wrapper.py
code served (permissive licence) · get_code("bfac58a04b6835f5")
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 3, dtype family float, shape (2, 4, 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
from torch.nn.utils.rnn import pack_padded_sequence

torch.manual_seed(0)

# Create a simple packed sequence
# We need: batch_size=2, sequence_length=4, input_size=3
batch_size = 2
seq_length = 4
input_size = 3

# Create input tensor (seq_length, batch_size, input_size)
data = torch.randn(seq_length, batch_size, input_size)

# Create lengths for each sequence in the batch
lengths = torch.tensor([4, 3], dtype=torch.long)

# Pack the sequence
packed_input = pack_padded_sequence(data, lengths, enforce_sorted=False)

# Create inverse indices for unsorting
# pack_padded_sequence with enforce_sorted=False returns sorted_indices
# We need the inverse permutation to unsort
sorted_indices = packed_input.sorted_indices
inv_ix = torch.argsort(sorted_indices)

ARGS = {
    "input": packed_input,
    "inv_ix": inv_ix
}
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 3, dtype family float, shape (2, 4, 3) — shape (2, 4, 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
parse_poolingfunction Ran
ran on a synthesized input
richermans/AudioCaption/models/encoder.py
code served (permissive licence) · get_code("c26450c99bacd017")
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 AttentionPool. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# Test case 1: mean pooling
ARGS_mean = {
    'poolingfunction_name': 'mean'
}

# Test case 2: linear pooling
ARGS_linear = {
    'poolingfunction_name': 'linear'
}

# Test case 3: attention pooling
ARGS_attention = {
    'poolingfunction_name': 'attention',
    'inputdim': 4,
    'outputdim': 3
}

# Use one of the test cases as the primary ARGS
ARGS = ARGS_attention
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 AttentionPool — a AttentionPool output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a AttentionPool 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
popen Ran
ran; our contract draft was wrong, not the code
richermans/AudioCaption/utils/kaldi_io.py
pointer only (licence: MIT); the stored copy get_code serves, from cffan/corp, is pointer only (licence: NONE) · get_code("d9a4862872e34aad")
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 cffan/corp, 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.

rank1_apporx Ran
ran on a synthesized input
wsntxxn/AudioCaption/captioning/models/conv_filter_pruning.py
code served (permissive licence) · get_code("0fac7426325be231")
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 1, dtype family float, shape (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

# Create a small test matrix for rank1_apporx
# The function performs SVD and extracts a rank-1 approximation
# data should be a 2D array (matrix)
data = np.random.randn(4, 3)

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 ndarray, ndim 1, dtype family float, shape (4,) — shape (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
read_key Ran
ran; our contract draft was wrong, not the code
richermans/AudioCaption/utils/kaldi_io.py
pointer only (licence: MIT); the stored copy get_code serves, from cffan/corp, is pointer only (licence: NONE) · get_code("aa10e48fc468d9e0")
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 cffan/corp, 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.

reparameterize_argmax Ran
ran on a synthesized input
wsntxxn/AudioCaption/captioning/losses/loss.py
code served (permissive licence) · get_code("5ead97ccfc2dee7b")
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, 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 takes a logit tensor and performs a reparameterization trick
# It needs a tensor that supports .size(), .max(), .view(), .scatter_(), and .detach()

logit = torch.randn(2, 4)  # batch_size=2, num_classes=4
dim = -1

ARGS = {
    "logit": logit,
    "dim": dim,
}
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, 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
sample_gumbel Ran
ran; our contract draft was wrong, not the code
wsntxxn/AudioCaption/captioning/losses/loss.py
pointer only (licence: MIT); the stored copy get_code serves, from xidongwu/AutoTrainOnce, is pointer only (licence: NONE) · get_code("2908a8bdfec77e29")
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 xidongwu/AutoTrainOnce, 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.

sort_pack_padded_sequence Ran
ran; our contract draft was wrong, not the code
wsntxxn/AudioCaption/captioning/models/hf_wrapper.py
pointer only (licence: MIT); the stored copy get_code serves, from 98lingchen/MICCAI2026, is pointer only (licence: NONE) · get_code("86839e01585b85d8")
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 98lingchen/MICCAI2026, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

build_vocab Not yet run
unverified (harvested, no recorded run)
richermans/AudioCaption/utils/build_vocab.py
code served (permissive licence) · get_code("ef7046e8db3256f6")
How it ran

no invocation recorded.

pack_wrapper Not yet run
unverified (harvested, no recorded run)
wsntxxn/AudioCaption/captioning/models/hf_wrapper.py
code served (permissive licence) · get_code("471b42e75608e3d8")
How it ran

no invocation recorded.

Also present in other papers' code

Of 21 samples lifted from this paper's linked repositories, 21 carry a function identity and 0 do not. Of the 21 distinct functions with an identity, 7 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 0 the same month, 6 later) and 14 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
pad_unsort_packed_sequence 14 other papers
4 dated earlier, 10 later · 2 only as an assembly brick
1612.00563 (2016-12)
Self-critical Sequence Training for Image Captioning
1707.07998 (2017-07)
Bottom-Up and Top-Down Attention for Image Captioning and Visual Question Answering
1906.05963 (2019-06)
Image Captioning: Transforming Objects into Words
and 11 more not named here
sample_gumbel 3 other papers
3 later · 1 only as an assembly brick
aaai_29990 (2024 (year only))
GOALNET: Interleaving Neural Goal Predicate Inference with Classical Planning for Generalization in Robot Instruction Following
2403.14729 (2024-03)
Auto-Train-Once: Controller Network Guided Automatic Network Pruning from Scratch
2410.11988 (2024-10)
DISP-LLM: Dimension-Independent Structural Pruning for Large Language Models
sort_pack_padded_sequence 3 other papers
3 later
2021.acl-long.459 (2021 (year only))
Cross-modal Memory Networks for Radiology Report Generation
aaai_29097 (2024 (year only))
Noise-Aware Image Captioning with Progressively Exploring Mismatched Words
2606.21447 (2026-06)
Precision Recall Controllable Radiology Report Generation via Hybrid Natural Language and Clinical Reward Learning
open_or_fd 2 other papers
2 later
2203.15455 (2022-03)
WeNet 2.0: More Productive End-to-End Speech Recognition Toolkit
2311.03611 (2023-11)
Plug-and-Play Stability for Intracortical Brain-Computer Interfaces: A One-Year Demonstration of Seamless Brain-to-Text Communication
popen 2 other papers
2 later
2203.15455 (2022-03)
WeNet 2.0: More Productive End-to-End Speech Recognition Toolkit
2311.03611 (2023-11)
Plug-and-Play Stability for Intracortical Brain-Computer Interfaces: A One-Year Demonstration of Seamless Brain-to-Text Communication
read_key 2 other papers
2 later
2203.15455 (2022-03)
WeNet 2.0: More Productive End-to-End Speech Recognition Toolkit
2311.03611 (2023-11)
Plug-and-Play Stability for Intracortical Brain-Computer Interfaces: A One-Year Demonstration of Seamless Brain-to-Text Communication
make_divisible 1 other paper
1 dated earlier
1804.02767 (2018-04)
YOLOv3: An Incremental Improvement

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

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

Abstract

Audio captioning is the novel task of general audio content description using free text. It is an intermodal translation task (not speech-to-text), where a system accepts as an input an audio signal and outputs the textual description (i.e. the caption) of that signal. In this paper we present Clotho, a dataset for audio captioning consisting of 4981 audio samples of 15 to 30 seconds duration and 24 905 captions of eight to 20 words length, and a baseline method to provide initial results. Clotho is built with focus on audio content and caption diversity, and the splits of the data are not hampering the training or evaluation of methods. All sounds are from the Freesound platform, and captions are crowdsourced using Amazon Mechanical Turk and annotators from English speaking countries. Unique words, named entities, and speech transcription are removed with post-processing. Clotho is freely available online (https://zenodo.org/record/3490684).

For agents

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

get_harvested_code_for_paper("1910.09387")
get_code_for_paper("1910.09387")
have("1910.09387")

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

curl https://syntology.ai/api/ran/1910.09387.json

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

[![Syntology run record](https://syntology.ai/api/ran/1910.09387.svg)](https://syntology.ai/paper/1910.09387)

Connect an agent — have() is free.