SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1805.03294 · 2018

Improved training of end-to-end attention models for speech recognition

arXiv · PDF · Open in the Atlas

Code that ran

12 of 12 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; 2 honoured the contract we drafted, 1 violated it and 8 have no contract verdict; 1 failure of our instrument; 0 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/1805.03294.json · badge: /api/ran/1805.03294.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
bobchennan/espnet community (classifier label)
classifier basis not recorded on the link
7 ran · 0 no recorded run
creatorscan/espnet community (classifier label)
classifier basis not recorded on the link
5 ran · 0 no recorded run
FunctionStatusWhere it lives
add_results_to_json Ran
ran on a synthesized input
creatorscan/espnet/espnet/mt/mt_utils.py
code served (permissive licence) · get_code("bbe192baf1d74dd1")
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 logging

# Set up logging to avoid warnings
logging.basicConfig(level=logging.INFO)

# Define char_list - a list of characters including special tokens
char_list = ['<blank>', '<space>', 'a', 'b', 'c', 'd', 'e', 'f', 'g', 'h']

# Define nbest_hyps - list of hypothesis dictionaries
# Each hypothesis has 'yseq' (token indices) and 'score'
nbest_hyps = [
    {
        'yseq': [0, 3, 2, 4, 1, 5],  # First element (0) is skipped in parsing
        'score': -2.5
    },
    {
        'yseq': [0, 3, 2, 5, 1, 6],
        'score': -3.1
    }
]

# Define js - the groundtruth utterance dict
js = {
    'utt2spk': 'speaker_001',
    'output': [
        {
            'name': 'ref',
            'text': 'reference text'
        },
        {
            'text': 'source text',
            'token': 's o u r c e',
            'tokenid': '0 1 2 3 4 5'
        }
    ]
}

ARGS = {
    'js': js,
    'nbest_hyps': nbest_hyps,
    'char_list': char_list
}
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
count_tokens Ran
ran on a synthesized input
bobchennan/espnet/espnet/lm/lm_utils.py
code served (permissive licence) · get_code("ff899e7de6121891")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Create sample token ID sequences
# The function iterates over data (list of sequences)
# For each sequence, it calls len() and np.count_nonzero()
# So we need a list of numpy arrays containing integer token IDs

sequence_1 = np.array([1, 2, 3, 0, 2, 0, 5], dtype=np.int32)
sequence_2 = np.array([4, 0, 1, 2, 0, 0, 3], dtype=np.int32)
sequence_3 = np.array([0, 0, 2, 1], dtype=np.int32)

data = [sequence_1, sequence_2, sequence_3]

# unk_id is the ID of the unknown token to count
# The function uses it in: np.count_nonzero(sentence == unk_id)
# So we need an integer ID that appears in our sequences
unk_id = 0

ARGS = {
    "data": data,
    "unk_id": unk_id
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family int — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
count_tokens Ran
ran on a synthesized input
creatorscan/espnet/espnet/lm/lm_utils.py
code served (permissive licence) · get_code("e2309fb95e3f748c")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Create sample token ID sequences
sequence1 = np.array([1, 2, 3, 0, 5, 0, 7], dtype=np.int32)
sequence2 = np.array([2, 0, 4, 6, 0, 8], dtype=np.int32)
sequence3 = np.array([1, 1, 0, 3, 5], dtype=np.int32)

data = [sequence1, sequence2, sequence3]
unk_id = 0

ARGS = {
    "data": data,
    "unk_id": unk_id,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family int — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
end_detect Ran
ran, violated the contract we drafted
bobchennan/espnet/espnet/nets/e2e_asr_common.py
code served (permissive licence) · get_code("db10d7f5f75d9faa")
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 bool, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Construct ended_hyps: a list of hypothesis dictionaries
# Each hypothesis needs 'score' and 'yseq' keys based on function usage
ended_hyps = [
    {'score': -5.0, 'yseq': [1, 2, 3, 4, 5]},
    {'score': -4.5, 'yseq': [1, 2, 3, 4]},
    {'score': -6.0, 'yseq': [1, 2, 3]},
    {'score': -4.8, 'yseq': [1, 2, 3, 4, 5]},
    {'score': -5.5, 'yseq': [1, 2]},
]

# i: current step/position (integer)
i = 5

# M: number of iterations to check (default 3)
M = 3

# D_end: threshold for end detection (default is log(1 * exp(-10)))
D_end = np.log(1 * np.exp(-10))

ARGS = {
    'ended_hyps': ended_hyps,
    'i': i,
    'M': M,
    'D_end': D_end,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, six — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type bool, 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_vgg2l_odim Ran
ran, honoured the contract we drafted
bobchennan/espnet/espnet/nets/e2e_asr_common.py
code served (permissive licence) · get_code("34e1e172156ca868")
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 numpy as np

def get_vgg2l_odim(idim, in_channel=3, out_channel=128):
    idim = idim / in_channel
    idim = np.ceil(np.array(idim, dtype=np.float32) / 2)
    idim = np.ceil(np.array(idim, dtype=np.float32) / 2)
    return int(idim) * out_channel

ARGS = {
    "idim": 240,
    "in_channel": 3,
    "out_channel": 128
}
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 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
label_smoothing_dist Ran
ran, honoured the contract we drafted
bobchennan/espnet/espnet/nets/e2e_asr_common.py
code served (permissive licence) · get_code("0d9f822390e7d551")
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 (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 json
import tempfile
import numpy as np
import os

# Create a temporary transcript file for testing
transcript_data = {
    'utts': {
        'utt1': {
            'output': [
                {
                    'tokenid': '1 2 3 1 2'
                }
            ]
        },
        'utt2': {
            'output': [
                {
                    'tokenid': '2 3 4'
                }
            ]
        },
        'utt3': {
            'output': [
                {
                    'tokenid': '1 1 1'
                }
            ]
        }
    }
}

# Create a temporary file with the transcript data
temp_file = tempfile.NamedTemporaryFile(mode='w', suffix='.json', delete=False)
json.dump(transcript_data, temp_file)
temp_file.close()
transcript_path = temp_file.name

ARGS = {
    'odim': 5,
    'lsm_type': 'unigram',
    'transcript': transcript_path,
    'blank': 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 1, dtype family float, shape (5,) — shape (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
load_dataset Ran
ran on a synthesized input
creatorscan/espnet/espnet/lm/lm_utils.py
code served (permissive licence) · get_code("826840e1c3869a70")
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 tempfile
import os
import numpy as np

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

# Create a temporary input text file with sample tokens
input_file = os.path.join(temp_dir, "test_tokens.txt")
with open(input_file, 'w', encoding='utf-8') as f:
    f.write("hello world foo\n")
    f.write("bar baz hello\n")
    f.write("unknown token here\n")

# Create a label dictionary mapping tokens to IDs
label_dict = {
    '<unk>': 0,
    'hello': 1,
    'world': 2,
    'foo': 3,
    'bar': 4,
    'baz': 5,
}

# Create output directory
outdir = os.path.join(temp_dir, "output")

ARGS = {
    'path': input_file,
    'label_dict': label_dict,
    'outdir': outdir,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
h5py, numpy, tqdm — 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
make_batchset Ran
ran on a synthesized input
bobchennan/espnet/espnet/asr/asr_mix_utils.py
code served (permissive licence) · get_code("6de087d747415428")
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 3, 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 logging
import numpy as np

# Set up logging
logging.basicConfig(level=logging.INFO)

# Create sample data dictionary matching the expected structure
# The function expects data to be Dict[str, Dict[str, Any]]
# where each entry has 'input' and 'output' keys with 'shape' information

data = {
    "utt_0": {
        "input": [{"shape": [100]}],  # input length 100
        "output": [{"shape": [50]}]   # output length 50
    },
    "utt_1": {
        "input": [{"shape": [80]}],   # input length 80
        "output": [{"shape": [40]}]   # output length 40
    },
    "utt_2": {
        "input": [{"shape": [120]}],  # input length 120
        "output": [{"shape": [60]}]   # output length 60
    },
    "utt_3": {
        "input": [{"shape": [90]}],   # input length 90
        "output": [{"shape": [45]}]   # output length 45
    },
    "utt_4": {
        "input": [{"shape": [110]}],  # input length 110
        "output": [{"shape": [55]}]   # output length 55
    },
    "utt_5": {
        "input": [{"shape": [70]}],   # input length 70
        "output": [{"shape": [35]}]   # output length 35
    }
}

ARGS = {
    "data": data,
    "batch_size": 32,
    "max_length_in": 100,
    "max_length_out": 50,
    "num_batches": 0,
    "min_batch_size": 1
}
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 3, 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
make_lexical_tree Ran
ran; our contract draft was wrong, not the code
bobchennan/espnet/espnet/lm/lm_utils.py
code served (permissive licence) · get_code("2956cc98fe367b81")
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 torch
import numpy as np

# Create a small word dictionary with word IDs
word_dict = {
    "hello": 1,
    "world": 2,
    "cat": 3,
    "dog": 4,
}

# Create a small subword (character) dictionary
subword_dict = {
    "h": 1,
    "e": 2,
    "l": 3,
    "o": 4,
    "w": 5,
    "r": 6,
    "d": 7,
    "c": 8,
    "a": 9,
    "t": 10,
    "g": 11,
}

# word_unk is the ID for unknown words
word_unk = 0

ARGS = {
    "word_dict": word_dict,
    "subword_dict": subword_dict,
    "word_unk": word_unk,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 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
parse_hypothesis Ran
ran on a synthesized input
creatorscan/espnet/espnet/mt/mt_utils.py
code served (permissive licence) · get_code("48cef9e2f067bc77")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Construct char_list - a list of characters/tokens
char_list = ['<space>', 'a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i']

# Construct hyp - a hypothesis dict with 'yseq' and 'score' keys
# yseq should be a list where first element is ignored (sliced [1:])
# and remaining elements are indices into char_list
hyp = {
    'yseq': [0, 1, 2, 3, 4, 1, 5],  # First 0 will be ignored, rest are indices
    'score': -10.5
}

ARGS = {
    'hyp': hyp,
    'char_list': char_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, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
read_tokens Ran
ran on a synthesized input
bobchennan/espnet/espnet/lm/lm_utils.py
code served (permissive licence) · get_code("b9cca15d6b089537")
How it ran

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

import numpy as np
import tempfile
import os

# Create a temporary file with sample token data
temp_file = tempfile.NamedTemporaryFile(mode='w', delete=False, suffix='.txt')
temp_file.write("cat dog bird\n")
temp_file.write("dog cat\n")
temp_file.write("bird bird cat dog\n")
temp_file.close()

# Create a label dictionary mapping token strings to IDs
label_dict = {
    'cat': 0,
    'dog': 1,
    'bird': 2,
    '<unk>': 3
}

ARGS = {
    'filename': temp_file.name,
    'label_dict': label_dict
}
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 — a list output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
read_tokens Ran
ran on a synthesized input
creatorscan/espnet/espnet/lm/lm_utils.py
code served (permissive licence) · get_code("dc5fc4e9e58970cb")
How it ran

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

import tempfile
import os
import numpy as np

# Create a temporary file with sample token data
temp_file = tempfile.NamedTemporaryFile(mode='w', delete=False, suffix='.txt', encoding='utf-8')
temp_filename = temp_file.name

# Write sample data: sequences of tokens
temp_file.write("apple banana cherry\n")
temp_file.write("dog cat bird\n")
temp_file.write("the quick brown fox\n")
temp_file.close()

# Create label_dict mapping token strings to IDs
label_dict = {
    'apple': 1,
    'banana': 2,
    'cherry': 3,
    'dog': 4,
    'cat': 5,
    'bird': 6,
    'the': 7,
    'quick': 8,
    'brown': 9,
    'fox': 10,
    '<unk>': 0
}

ARGS = {
    'filename': temp_filename,
    'label_dict': label_dict
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tqdm — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list — a list output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block

Also present in other papers' code

Of 12 samples lifted from this paper's linked repositories, 12 carry a function identity and 0 do not. Of the 11 distinct functions with an identity, 4 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 4 later) and 7 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
end_detect 2 other papers
2 later
2005.14327 (2020-05)
On the Comparison of Popular End-to-End Models for Large Scale Speech Recognition
2404.02098 (2024-04)
BRAVEn: Improving Self-Supervised Pre-training for Visual and Auditory Speech Recognition
label_smoothing_dist 2 other papers
2 later
2005.14327 (2020-05)
On the Comparison of Popular End-to-End Models for Large Scale Speech Recognition
2404.02098 (2024-04)
BRAVEn: Improving Self-Supervised Pre-training for Visual and Auditory Speech Recognition
get_vgg2l_odim 1 other paper
1 later
2404.02098 (2024-04)
BRAVEn: Improving Self-Supervised Pre-training for Visual and Auditory Speech Recognition
make_lexical_tree 1 other paper
1 later
2212.04356 (2022-12)
Robust Speech Recognition via Large-Scale Weak Supervision

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

Sequence-to-sequence attention-based models on subword units allow simple open-vocabulary end-to-end speech recognition. In this work, we show that such models can achieve competitive results on the Switchboard 300h and LibriSpeech 1000h tasks. In particular, we report the state-of-the-art word error rates (WER) of 3.54% on the dev-clean and 3.82% on the test-clean evaluation subsets of LibriSpeech. We introduce a new pretraining scheme by starting with a high time reduction factor and lowering it during training, which is crucial both for convergence and final performance. In some experiments, we also use an auxiliary CTC loss function to help the convergence. In addition, we train long short-term memory (LSTM) language models on subword units. By shallow fusion, we report up to 27% relative improvements in WER over the attention baseline without a language model.

For agents

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

get_harvested_code_for_paper("1805.03294")
get_code_for_paper("1805.03294")
have("1805.03294")

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

curl https://syntology.ai/api/ran/1805.03294.json

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

[![Syntology run record](https://syntology.ai/api/ran/1805.03294.svg)](https://syntology.ai/paper/1805.03294)

Connect an agent — have() is free.