SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1508.01211 · 2015

Listen, Attend and Spell

arXiv · PDF · Open in the Atlas

Code that ran

30 of 51 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, 1 violated it and 18 have no contract verdict; 10 failures of our instrument; 21 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/1508.01211.json · badge: /api/ran/1508.01211.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
foamliu/Listen-Attend-Spell community (classifier label)
classifier basis not recorded on the link
5 ran · 0 no recorded run
WindQAQ/listen-attend-and-spell community (classifier label)
classifier basis not recorded on the link
3 ran · 11 no recorded run
sciforce/phones-las community (classifier label)
classifier basis not recorded on the link
3 ran · 7 no recorded run
sooftware/Fairseq-Listen-Attend-Spell reimplementation (classifier label)
classifier basis: classifier_2025
3 ran · 0 no recorded run
colaprograms/speechify reimplementation (classifier label)
classifier basis: classifier_2025
3 ran · 0 no recorded run
msalhab96/Listen-Attend-and-Spell reimplementation (classifier label)
classifier basis: classifier_2025
3 ran · 0 no recorded run
custodio78/Speech_Recognition community (classifier label)
classifier basis not recorded on the link
3 ran · 0 no recorded run
switiz/las.pytorch community (classifier label)
classifier basis not recorded on the link
2 ran · 1 no recorded run
sanjanamallya/speech_to_text reimplementation (classifier label)
classifier basis: classifier_2025
2 ran · 1 no recorded run
anjandeepsahni/automatic_speech_recognition community (classifier label)
classifier basis not recorded on the link
2 ran · 0 no recorded run
jackaduma/LAS_Mandarin_PyTorch reimplementation (classifier label)
classifier basis: classifier_2025
1 ran · 0 no recorded run
foamliu/Listen-Attend-Spell-v2 community (classifier label)
classifier basis not recorded on the link
0 ran · 1 no recorded run
FunctionStatusWhere it lives
SpeechCollateFn Ran
ran on a synthesized input
anjandeepsahni/automatic_speech_recognition/Code/dataset.py
code served (permissive licence) · get_code("bc50e5bc9861a026")
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.utils.rnn as rnn

torch.manual_seed(0)

IGNORE_ID = -1

# Create sample data: list of (input, target) tuples
# Each input is a 1D tensor (sequence of audio features)
# Each target is a 1D tensor (sequence of token IDs)

# Sample 1: longer input sequence
input_1 = torch.randn(10, dtype=torch.float32)  # 10 time steps
target_1 = torch.tensor([1, 2, 3, 4], dtype=torch.long)  # 4 tokens

# Sample 2: shorter input sequence
input_2 = torch.randn(6, dtype=torch.float32)  # 6 time steps
target_2 = torch.tensor([5, 6], dtype=torch.long)  # 2 tokens

seq_list = [
    (input_1, target_1),
    (input_2, target_2),
]

ARGS = {
    "seq_list": seq_list,
}
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
accuracy Ran
ran on a synthesized input
foamliu/Listen-Attend-Spell/utils.py
code served (permissive licence) · get_code("5636e2d24b43d1a9")
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 small tensors for testing
batch_size = 2
num_classes = 4

# scores: prediction scores, shape (batch_size, num_classes)
# The function calls scores.topk(k, 1, True, True) and ind.eq(targets.view(-1, 1).expand_as(ind))
scores = torch.randn(batch_size, num_classes)

# targets: ground truth class indices, shape (batch_size,)
# The function calls targets.size(0) and targets.view(-1, 1)
targets = torch.tensor([1, 3])

# k: number of top predictions to consider
k = 1

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
add_asr_eval_argument Ran
ran; our contract draft was wrong, not the code
sooftware/Fairseq-Listen-Attend-Spell/fairseq_las/infer.py
code served (permissive licence) · get_code("af0229a45f77d3fb")
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 MockArgumentParser. 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 argparse

class MockArgumentParser:
    def __init__(self):
        self.arguments = {}
    
    def add_argument(self, *args, **kwargs):
        # Extract the parameter name from args (e.g., '--ctc' -> 'ctc')
        param_name = None
        for arg in args:
            if arg.startswith('--'):
                param_name = arg.lstrip('-')
                break
        
        if param_name:
            # Store default value if provided
            if 'default' in kwargs:
                self.arguments[param_name] = kwargs['default']
            elif kwargs.get('action') == 'store_true':
                self.arguments[param_name] = False
            else:
                self.arguments[param_name] = None
        
        return self

parser = MockArgumentParser()

ARGS = {
    'parser': parser
}
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 MockArgumentParser — a MockArgumentParser output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a MockArgumentParser 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
beam_decode Ran
ran; our contract draft was wrong, not the code
jackaduma/LAS_Mandarin_PyTorch/test_asr.py
code served (permissive licence) · get_code("a03edcfe518e7f8b")
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. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

class Hypothesis:
    def __init__(self, out_index):
        self.outIndex = out_index

class MockBeamDecodeModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(5, 5)
    
    def forward(self, feat, feat_len):
        # Return a list of Hypothesis objects (one per batch item)
        batch_size = feat.size(0)
        hyps = []
        for i in range(batch_size):
            # Each hypothesis has an outIndex which is a list of integers
            out_index = [1, 2, 3, 4, 0]  # Example decoded sequence
            hyps.append(Hypothesis(out_index))
        return hyps

# Create mock data
name = ["sample_utterance_001"]  # List with one name (batch size 1 effectively)
feat = torch.randn(2, 7, 5)  # (batch=2, time=7, feat_dim=5)
feat_len = torch.tensor([7, 5])  # Lengths for each item in batch
txt = torch.tensor([[1, 2, 3, 4, 0], [5, 6, 0, 0, 0]])  # (batch=2, max_txt_len=5), 0 is padding

model = MockBeamDecodeModel()
device = torch.device('cpu')

data = (name, feat, feat_len, txt)

ARGS = {
    'data': data,
    'model': model,
    'device': device
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type 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
beam_search Ran
ran on a synthesized input
anjandeepsahni/automatic_speech_recognition/Code/beamsearch.py
code served (permissive licence) · get_code("515dc628f1216f81")
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 numpy as np

def initial_state_function(X, batch_size):
    """Returns initial state and initial value for beam search"""
    # state can be any object, initial_value should be an integer (token id)
    initial_state = {"hidden": np.random.randn(batch_size, 4)}
    initial_value = 0  # START token
    return initial_state, initial_value

def generate_function(Y_tm1, state_tm1):
    """
    Generate next states and probabilities.
    
    Args:
        Y_tm1: list of previous token ids
        state_tm1: list of previous states
    
    Returns:
        state_t: list of new states
        p_t: numpy array of shape (batch_size, vocab_size) with probabilities
        extras_t: list of extra information per batch
    """
    batch_size = len(Y_tm1)
    vocab_size = 33
    
    # Generate new states
    state_t = [{"hidden": np.random.randn(4)} for _ in range(batch_size)]
    
    # Generate probability distributions over vocab
    p_t = np.random.dirichlet(np.ones(vocab_size), size=batch_size)
    
    # Generate extras (can be None or any auxiliary info)
    extras_t = [None for _ in range(batch_size)]
    
    return state_t, p_t, extras_t

# Input data
X = np.random.randn(2, 5)  # batch_size=2, feature_dim=5

ARGS = {
    "initial_state_function": initial_state_function,
    "generate_function": generate_function,
    "X": X,
    "end_id": 2,
    "batch_size": 2,
    "beam_width": 4,
    "num_hypotheses": 1,
    "max_length": 50,
    "vocab_size": 33,
}
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 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
build_LFR_features Ran
ran on a synthesized input
foamliu/Listen-Attend-Spell/data_gen.py
code served (permissive licence) · get_code("04f3eb7d85354749")
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 (5, 12) 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 simple input array: T x D
# T = 10 (time steps), D = 4 (features)
inputs = np.random.randn(10, 4).astype(np.float32)

# m: number of frames to stack
m = 3

# n: number of frames to skip
n = 2

ARGS = {
    "inputs": inputs,
    "m": m,
    "n": n,
}
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 2, dtype family float, shape (5, 12) — shape (5, 12) 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
check_envirionment Ran
ran on a synthesized input
switiz/las.pytorch/utils.py
code served (permissive licence) · get_code("5cb79c0f243a5040")
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 device. 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 logging

torch.manual_seed(0)

# Configure logging to see output
logging.basicConfig(level=logging.DEBUG, format='%(levelname)s: %(message)s')

# The function takes one parameter: use_cuda (boolean)
# It checks if CUDA is available and returns a device
# The function doesn't actually use the return value in a way that requires specific setup
# It just needs a boolean flag for use_cuda

ARGS = {
    'use_cuda': 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 device — a device output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a device 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
collate_test Ran
ran; our contract draft was wrong, not the code
sanjanamallya/speech_to_text/speech_to_text.py
pointer only (licence: NONE) · get_code("593b181ec23986e4")
How it ran

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

collate_train Ran
ran; our contract draft was wrong, not the code
sanjanamallya/speech_to_text/speech_to_text.py
pointer only (licence: NONE) · get_code("86b35c8530964541")
How it ran

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

create_attention_images Ran
ran on a synthesized input
WindQAQ/listen-attend-and-spell/utils/image_utils.py
code served (permissive licence) · get_code("f820a5f4353c8d50")
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 EagerTensor, ndim 4, dtype family float, shape (2, 4, 5, 1) 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
import numpy as np

class MockAlignmentHistory:
    def __init__(self):
        # alignment_history.stack() is called, which should return a tensor
        # The function then does transpose([1, 2, 0]), so we need 3D tensor
        # Shape: [time_steps, batch_size, attention_size]
        self._stacked = tf.constant(np.random.rand(5, 2, 4).astype(np.float32))
    
    def stack(self):
        return self._stacked

class MockFinalContextState:
    def __init__(self):
        self.alignment_history = MockAlignmentHistory()

ARGS = {
    "final_context_state": MockFinalContextState()
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 4, dtype family float, shape (2, 4, 5, 1) — shape (2, 4, 5, 1) 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
detrend Ran
ran, violated the contract we drafted
colaprograms/speechify/util/spectrogram_generator.py
pointer only (licence: NONE) · get_code("4e5663ad841caf95")
How it ran

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

encode_text Ran
ran on a synthesized input
sciforce/phones-las/text_preprocess.py
code served (permissive licence) · get_code("24b568092a10eca5")
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 numpy as np

# Create a small vocabulary
vocab = ['hello', 'world', 'test', 'python']

# Create inputs dict with text field containing elements from vocab
inputs = {
    'text': ['hello', 'world', 'test']
}

ARGS = {
    'vocab': vocab,
    'inputs': inputs
}
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 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_dataset_itr Ran
ran; our contract draft was wrong, not the code
sooftware/Fairseq-Listen-Attend-Spell/fairseq_las/infer.py
code served (permissive licence) · get_code("3b409833fb8f8509")
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 EpochBatchIterator. 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 typing import Iterator, Dict, Any, List, Tuple

torch.manual_seed(0)

class SimpleDataset:
    def __init__(self):
        self.data = [
            {'id': 0, 'source': torch.randn(5, 4), 'target': torch.randn(6, 4)},
            {'id': 1, 'source': torch.randn(7, 4), 'target': torch.randn(8, 4)},
        ]
    
    def __len__(self):
        return len(self.data)
    
    def __getitem__(self, idx):
        return self.data[idx]

class EpochBatchIterator:
    def __init__(self, dataset, batches):
        self.dataset = dataset
        self.batches = batches
        self._cur_epoch_itr = None
    
    def next_epoch_itr(self, shuffle=False):
        self._cur_epoch_itr = iter(self.batches)
        return self
    
    def __iter__(self):
        return self
    
    def __next__(self):
        return next(self._cur_epoch_itr)

class BatchIterator:
    def __init__(self, dataset, batches):
        self.dataset = dataset
        self.batches = batches
    
    def next_epoch_itr(self, shuffle=False):
        return EpochBatchIterator(self.dataset, self.batches)

class Task:
    def __init__(self):
        self._dataset = SimpleDataset()
        self.batches = [
            [self._dataset[0]],
            [self._dataset[1]],
        ]
    
    def dataset(self, split):
        return self._dataset
    
    def get_batch_iterator(
        self,
        dataset,
        max_tokens=None,
        max_sentences=None,
        max_positions=None,
        ignore_invalid_inputs=False,
        required_batch_size_multiple=1,
        num_shards=1,
        shard_id=0,
        num_workers=0
    ):
        return BatchIterator(dataset, self.batches)

class Args:
    def __init__(self):
        self.gen_subset = 'valid'
        self.max_tokens = 4000
        self.max_sentences = 2
        self.skip_invalid_size_inputs_valid_test = False
        self.required_batch_size_multiple = 1
        self.num_shards = 1
        self.shard_id = 0
        self.num_workers = 0

args = Args()
task = Task()

ARGS = {
    'args': args,
    'task': task
}
Setup block drafted by
a language model, Claude Sonnet 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 EpochBatchIterator — a EpochBatchIterator output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a EpochBatchIterator 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_learning_rate Ran
ran on a synthesized input
foamliu/Listen-Attend-Spell/utils.py
code served (permissive licence) · get_code("591851e781d27201")
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
import torch.optim as optim

torch.manual_seed(0)

# Create a simple model
model = nn.Linear(4, 3)

# Create an optimizer with a learning rate
optimizer = optim.SGD(model.parameters(), lr=0.01)

# Construct arguments for get_learning_rate
ARGS = {
    'optimizer': optimizer
}
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
get_result Ran
ran; our contract draft was wrong, not the code
msalhab96/Listen-Attend-and-Spell/inference.py
pointer only (licence: NONE) · get_code("8dea3694156cfb07")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, msalhab96/Listen-Attend-and-Spell when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

get_top_candidates Ran
ran; our contract draft was wrong, not the code
msalhab96/Listen-Attend-and-Spell/inference.py
pointer only (licence: NONE) · get_code("9ac4e3e2b4a3f182")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, msalhab96/Listen-Attend-and-Spell 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.

keep_top_seq Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
msalhab96/Listen-Attend-and-Spell/inference.py
pointer only (licence: NONE) · get_code("cbf9a742d462dc80")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, msalhab96/Listen-Attend-and-Spell 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_dataset Ran
ran on a synthesized input
switiz/las.pytorch/utils.py
code served (permissive licence) · get_code("8210be28ca690ac0")
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 tempfile
import os

# Create a temporary file with tab-separated transcript data
temp_file = tempfile.NamedTemporaryFile(mode='w', delete=False, suffix='.txt')
temp_file.write("audio/file1.wav\t한글 텍스트 1\tEnglish transcript 1\n")
temp_file.write("audio/file2.wav\t한글 텍스트 2\tEnglish transcript 2\n")
temp_file.write("audio/file3.wav\t한글 텍스트 3\tEnglish transcript 3\n")
temp_file.close()

transcripts_path = temp_file.name

ARGS = {
    "transcripts_path": transcripts_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, 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
load_vocab Ran
ran on a synthesized input
WindQAQ/listen-attend-and-spell/utils/vocab_utils.py
code served (permissive licence) · get_code("75776719b114cdb1")
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 file with vocab entries
temp_file = tempfile.NamedTemporaryFile(mode='w', delete=False, suffix='.txt')
temp_file.write("apple\n")
temp_file.write("banana\n")
temp_file.write("cherry\n")
temp_file.close()

ARGS = {
    'filename': temp_file.name
}
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
pad_audio_sequences Ran
ran on a synthesized input
custodio78/Speech_Recognition/sr_model_utils.py
code served (permissive licence) · get_code("a1a21a20c140d0be")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Create small audio sequences with different lengths
# Each sequence is 2D: (time_steps, features)
seq1 = np.random.randn(4, 3).astype(np.float32)  # 4 time steps, 3 features
seq2 = np.random.randn(6, 3).astype(np.float32)  # 6 time steps, 3 features
seq3 = np.random.randn(5, 3).astype(np.float32)  # 5 time steps, 3 features

sequences = [seq1, seq2, seq3]

ARGS = {
    "sequences": sequences,
    "tail": True,
}
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 — 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
pad_list Ran
ran on a synthesized input
foamliu/Listen-Attend-Spell/utils.py
code served (permissive licence) · get_code("ce10a394eb6dd5f4")
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 3, dtype family float, shape (2, 5, 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 list of tensors with different lengths
# The function expects xs to be a list of tensors
# It calls .size(0) to get the first dimension (sequence length)
# It uses .new() to create a new tensor with the same type

xs = [
    torch.randn(3, 4),  # First tensor: length 3, feature dim 4
    torch.randn(5, 4),  # Second tensor: length 5, feature dim 4
]

pad_value = 0.0

ARGS = {
    "xs": xs,
    "pad_value": pad_value,
}
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 3, dtype family float, shape (2, 5, 4) — shape (2, 5, 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
pad_txt_sequences Ran
ran on a synthesized input
custodio78/Speech_Recognition/sr_model_utils.py
code served (permissive licence) · get_code("be04197d5ade9e2f")
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

torch.manual_seed(0)

# sequences: a list of sequences (each sequence is iterable, like a list or string)
# Each sequence will be padded to the same length
sequences = [
    [1, 2, 3],
    [4, 5],
    [6, 7, 8, 9]
]

# pad_tok: the token/value to use for padding
pad_tok = 0

ARGS = {
    "sequences": sequences,
    "pad_tok": pad_tok
}
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
parse_trans Ran
ran on a synthesized input
WindQAQ/listen-attend-and-spell/vctk/preprocess.py
code served (permissive licence) · get_code("465de61d8fb45c5a")
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 string
import tempfile
import os

# Create a temporary text file with valid content
temp_dir = tempfile.mkdtemp()
test_file = os.path.join(temp_dir, "test.txt")

with open(test_file, 'w') as f:
    f.write("Hello, World! This is a test.\n")

ARGS = {
    'filename': test_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 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
prepare_result_files Ran
ran; our contract draft was wrong, not the code
sooftware/Fairseq-Listen-Attend-Spell/fairseq_las/infer.py
code served (permissive licence) · get_code("d6980854a04a00a8")
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 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 tempfile

class Args:
    def __init__(self):
        # Create a temporary directory for results
        self.results_path = tempfile.mkdtemp()
        # Create a temporary file path (just a string, not an actual file)
        self.path = "/some/model/path/model_checkpoint.pt"
        self.gen_subset = "test"

args = Args()

ARGS = {"args": args}
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 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
process_line Ran
ran on a synthesized input
sciforce/phones-las/text_preprocess.py
code served (permissive licence) · get_code("78f8ed248b3655ea")
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 re

class Args:
    def __init__(self):
        self.delimiter = '|'
        self.strip_stress = True

ARGS = {
    'args': Args(),
    'line': 'hello world|HEL0 WOR1D'
}
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
serving_input_factory Ran
ran on a synthesized input
sciforce/phones-las/export.py
code served (permissive licence) · get_code("26b39a8d6aa48289")
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 tensorflow as tf

ARGS = {
    'num_channels': 4
}
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 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
signal_to_spectrogram Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
colaprograms/speechify/util/spectrogram_generator.py
pointer only (licence: NONE) · get_code("4d8d193470ad0970")
How it ran

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

spec_augment Ran
ran on a synthesized input
foamliu/Listen-Attend-Spell/data_gen.py
code served (permissive licence) · get_code("7dc944e5f14aedce")
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 (8, 12) 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 random
import numpy as np

# Set seed for reproducibility
random.seed(0)
np.random.seed(0)

# Create a small spectrogram array (frequency bins x time frames)
spec = np.random.randn(8, 12).astype(np.float32)

ARGS = {
    "spec": spec,
    "num_mask": 2,
    "freq_masking": 0.15,
    "time_masking": 0.2,
    "value": 0
}
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 2, dtype family float, shape (8, 12) — shape (8, 12) 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
split_txts Ran
ran on a synthesized input
custodio78/Speech_Recognition/sr_data_utils.py
code served (permissive licence) · get_code("04400535de0c770b")
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)

ARGS = {
    "txts": [
        "hello world",
        "python code",
        "test string"
    ]
}
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
unitlinear Ran
ran, honoured the contract we drafted
colaprograms/speechify/util/spectrogram_generator.py
pointer only (licence: NONE) · get_code("b79342d36def2329")
How it ran

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

bilstm Not yet run
unverified (harvested, no recorded run)
WindQAQ/listen-attend-and-spell/las/ops.py
code served (permissive licence) · get_code("da03745c4d1dc159")
How it ran

no invocation recorded.

bilstm Not yet run
unverified (harvested, no recorded run)
sciforce/phones-las/las/ops.py
code served (permissive licence) · get_code("5e4b9b4da2357031")
How it ran

no invocation recorded.

cer Not yet run
unverified (harvested, no recorded run)
switiz/las.pytorch/utils.py
code served (permissive licence) · get_code("5eeba5cf42081f13")
How it ran

no invocation recorded.

compute_loss_sigmoid Not yet run
unverified (harvested, no recorded run)
sciforce/phones-las/model_helper.py
code served (permissive licence) · get_code("3eaa88271428a793")
How it ran

no invocation recorded.

create_vocab_table Not yet run
unverified (harvested, no recorded run)
WindQAQ/listen-attend-and-spell/utils/vocab_utils.py
code served (permissive licence) · get_code("6ea3a110248fc75e")
How it ran

no invocation recorded.

dense_to_sparse Not yet run
unverified (harvested, no recorded run)
WindQAQ/listen-attend-and-spell/utils/metrics_utils.py
code served (permissive licence) · get_code("5c2d13de9a6666e7")
How it ran

no invocation recorded.

edit_distance Not yet run
unverified (harvested, no recorded run)
WindQAQ/listen-attend-and-spell/utils/metrics_utils.py
code served (permissive licence) · get_code("49847ff466218c84")
How it ran

no invocation recorded.

extract_mfcc Not yet run
unverified (harvested, no recorded run)
WindQAQ/listen-attend-and-spell/vctk/preprocess.py
code served (permissive licence) · get_code("4e09b2d42a5a90a2")
How it ran

no invocation recorded.

get_iterator Not yet run
unverified (harvested, no recorded run)
WindQAQ/listen-attend-and-spell/utils/iterator_utils.py
code served (permissive licence) · get_code("bb990c536558a8da")
How it ran

no invocation recorded.

lstm_cell Not yet run
unverified (harvested, no recorded run)
WindQAQ/listen-attend-and-spell/las/ops.py
code served (permissive licence) · get_code("4fc2dd53c063d576")
How it ran

no invocation recorded.

lstm_cell Not yet run
unverified (harvested, no recorded run)
sciforce/phones-las/las/ops.py
code served (permissive licence) · get_code("8beb2dfe1d9a90a4")
How it ran

no invocation recorded.

make_example Not yet run
unverified (harvested, no recorded run)
sciforce/phones-las/preprocess_all.py
code served (permissive licence) · get_code("fbf7b91f4fcb63fb")
How it ran

no invocation recorded.

pad_collate Not yet run
unverified (harvested, no recorded run)
foamliu/Listen-Attend-Spell-v2/data_gen.py
code served (permissive licence) · get_code("7b22196dd6d0fb21")
How it ran

no invocation recorded.

phrase_to_binfs Not yet run
unverified (harvested, no recorded run)
sciforce/phones-las/infer.py
code served (permissive licence) · get_code("bf95c99f28cf2c1d")
How it ran

no invocation recorded.

process_dataset Not yet run
unverified (harvested, no recorded run)
WindQAQ/listen-attend-and-spell/utils/dataset_utils.py
code served (permissive licence) · get_code("00ec8a3f07cfb9b9")
How it ran

no invocation recorded.

process_speakers Not yet run
unverified (harvested, no recorded run)
WindQAQ/listen-attend-and-spell/vctk/preprocess.py
code served (permissive licence) · get_code("05d046c3a10a8bb4")
How it ran

no invocation recorded.

pyramidal_stack Not yet run
unverified (harvested, no recorded run)
WindQAQ/listen-attend-and-spell/las/ops.py
code served (permissive licence) · get_code("5eefaae1dce38d90")
How it ran

no invocation recorded.

read_dataset Not yet run
unverified (harvested, no recorded run)
WindQAQ/listen-attend-and-spell/utils/dataset_utils.py
code served (permissive licence) · get_code("5c1b9de46d48a986")
How it ran

no invocation recorded.

sequence_loss_sigmoid Not yet run
unverified (harvested, no recorded run)
sciforce/phones-las/model_helper.py
code served (permissive licence) · get_code("5b0af40833a50caf")
How it ran

no invocation recorded.

soft2hard_attention Not yet run
unverified (harvested, no recorded run)
sciforce/phones-las/infer.py
code served (permissive licence) · get_code("3f5f00347c16f25b")
How it ran

no invocation recorded.

train Not yet run
unverified (harvested, no recorded run)
sanjanamallya/speech_to_text/speech_to_text.py
pointer only (licence: NONE) · get_code("226df6a3394d1c2d")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, sanjanamallya/speech_to_text 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. No invocation recorded.

Also present in other papers' code

Of 51 samples lifted from this paper's linked repositories, 51 carry a function identity and 0 do not. Of the 51 distinct functions with an identity, 6 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 6 later) and 45 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 3 other papers
3 later
1512.01882 (2015-12)
THCHS-30 : A Free Chinese Speech Corpus
1703.03872 (2017-03)
Deep Image Matting
1804.07573 (2018-04)
MobileFaceNets: Efficient CNNs for Accurate Real-Time Face Verification on Mobile Devices
pad_list 3 other papers
3 later
1512.01882 (2015-12)
THCHS-30 : A Free Chinese Speech Corpus
1512.02595 (2015-12)
Deep Speech 2: End-to-End Speech Recognition in English and Mandarin
2212.04559 (2022-12)
SpeechLMScore: Evaluating speech generation using speech language model
cer 1 other paper
1 later
1512.02595 (2015-12)
Deep Speech 2: End-to-End Speech Recognition in English and Mandarin
check_envirionment 1 other paper
1 later
1512.02595 (2015-12)
Deep Speech 2: End-to-End Speech Recognition in English and Mandarin
load_dataset 1 other paper
1 later
1512.02595 (2015-12)
Deep Speech 2: End-to-End Speech Recognition in English and Mandarin
pad_collate 1 other paper
1 later
1512.01882 (2015-12)
THCHS-30 : A Free Chinese Speech Corpus

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

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

We present Listen, Attend and Spell (LAS), a neural network that learns to transcribe speech utterances to characters. Unlike traditional DNN-HMM models, this model learns all the components of a speech recognizer jointly. Our system has two components: a listener and a speller. The listener is a pyramidal recurrent network encoder that accepts filter bank spectra as inputs. The speller is an attention-based recurrent network decoder that emits characters as outputs. The network produces character sequences without making any independence assumptions between the characters. This is the key improvement of LAS over previous end-to-end CTC models. On a subset of the Google voice search task, LAS achieves a word error rate (WER) of 14.1% without a dictionary or a language model, and 10.3% with language model rescoring over the top 32 beams. By comparison, the state-of-the-art CLDNN-HMM model achieves a WER of 8.0%.

For agents

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

get_harvested_code_for_paper("1508.01211")
get_code_for_paper("1508.01211")
have("1508.01211")

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

curl https://syntology.ai/api/ran/1508.01211.json

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

[![Syntology run record](https://syntology.ai/api/ran/1508.01211.svg)](https://syntology.ai/paper/1508.01211)

Connect an agent — have() is free.