SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2109.00181 · EMNLP · 2021

CTAL: Pre-training Cross-modal Transformer for Audio-and-Language Representations

Hang Li, Zitao Liu, Tianqiao Liu, Yu Kang, Wenbiao Ding, Zhongqin Wu

arXiv · PDF · Open in the Atlas

Code that ran

15 of 19 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (2 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 14 have no contract verdict; 0 failures of our instrument; 4 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/2109.00181.json · badge: /api/ran/2109.00181.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
ydkwim/ctal canonical (classifier label)
classifier basis: census
13 ran · 1 no recorded run
tal-ai/CTAL_EMNLP2021 community
Papers with Code archive
2 ran · 3 no recorded run
FunctionStatusWhere it lives
AcousticInputRepresentations Ran
ran on a synthesized input · constructed an object; no result was computed
tal-ai/CTAL_EMNLP2021/code/m2p_model.py
pointer only (licence: GPL-3.0) · get_code("822f2b1124d338bb")
How it ran

The body of this sample is withheld: GPL-3.0, as GitHub detects it on this paper's repository, tal-ai/CTAL_EMNLP2021, is not in the permissive set Syntology serves inline. 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.

RobertaAMHead Ran
ran on a synthesized input · constructed an object; no result was computed
tal-ai/CTAL_EMNLP2021/code/m2p_model.py
pointer only (licence: GPL-3.0) · get_code("bc9ad7176f8222c8")
How it ran

The body of this sample is withheld: GPL-3.0, as GitHub detects it on this paper's repository, tal-ai/CTAL_EMNLP2021, is not in the permissive set Syntology serves inline. 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 Ran
ran on a synthesized input
ydkwim/ctal/code/run_s-iemocap.py
code served (permissive licence) · get_code("d3d9401fa2444bc8")
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
from unittest.mock import Mock

torch.manual_seed(0)

# Create a minimal tokenizer stand-in that supports the .pad() method
class MinimalTokenizer:
    def pad(self, batch_dict, return_tensors=None):
        # batch_dict has 'input_ids' and 'attention_mask' as lists
        input_ids_list = batch_dict['input_ids']
        attention_mask_list = batch_dict['attention_mask']
        
        # Find max length for padding
        max_len = max(len(ids) for ids in input_ids_list)
        
        # Pad sequences
        padded_input_ids = []
        padded_attention_mask = []
        for ids, mask in zip(input_ids_list, attention_mask_list):
            pad_len = max_len - len(ids)
            padded_input_ids.append(ids + [0] * pad_len)
            padded_attention_mask.append(mask + [0] * pad_len)
        
        if return_tensors == 'pt':
            return {
                'input_ids': torch.tensor(padded_input_ids, dtype=torch.long),
                'attention_mask': torch.tensor(padded_attention_mask, dtype=torch.long)
            }
        return {
            'input_ids': padded_input_ids,
            'attention_mask': padded_attention_mask
        }

tokenizer = MinimalTokenizer()

# Create sample_list with 2 samples
sample_list = [
    {
        'text_input': {
            'input_ids': [101, 2054, 2003],
            'attention_mask': [1, 1, 1]
        },
        'label': 0,
        'audio_name': 'audio_001.wav'
    },
    {
        'text_input': {
            'input_ids': [101, 2054],
            'attention_mask': [1, 1]
        },
        'label': 1,
        'audio_name': 'audio_002.wav'
    }
]

ARGS = {
    'sample_list': sample_list,
    'tokenizer': tokenizer
}
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
collate Ran
ran on a synthesized input
ydkwim/ctal/code/run_s-reg-mosei.py
code served (permissive licence) · get_code("6ceec151acbbc145")
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)

class MockTokenizer:
    def pad(self, batch_dict, return_tensors=None):
        input_ids = batch_dict['input_ids']
        attention_mask = batch_dict['attention_mask']
        
        max_len = max(len(ids) for ids in input_ids)
        
        padded_ids = []
        padded_masks = []
        
        for ids, mask in zip(input_ids, attention_mask):
            pad_len = max_len - len(ids)
            padded_ids.append(ids + [0] * pad_len)
            padded_masks.append(mask + [0] * pad_len)
        
        if return_tensors == 'pt':
            return {
                'input_ids': torch.tensor(padded_ids, dtype=torch.long),
                'attention_mask': torch.tensor(padded_masks, dtype=torch.long)
            }
        
        return {
            'input_ids': padded_ids,
            'attention_mask': padded_masks
        }

sample_list = [
    {
        'text_input': {
            'input_ids': [101, 2054],
            'attention_mask': [1, 1]
        },
        'label': 0.5,
        'audio_name': 'audio_001.wav'
    },
    {
        'text_input': {
            'input_ids': [101, 2054, 2003],
            'attention_mask': [1, 1, 1]
        },
        'label': 0.8,
        'audio_name': 'audio_002.wav'
    }
]

tokenizer = MockTokenizer()

ARGS = {
    'sample_list': sample_list,
    'tokenizer': tokenizer
}
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
down_sample_frames Ran
ran on a synthesized input
ydkwim/ctal/code/m2p_mask.py
code served (permissive licence) · get_code("1cec2fe099d913d4")
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, 4, 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 torch

torch.manual_seed(0)

# Analyze the function:
# - spec: a tensor with shape (batch, time_steps, features)
#   - Uses: spec.shape[1], spec.shape[2], spec.shape[0]
#   - Operations: indexing with :-left_over, view() reshaping
# - dr: a downsampling rate (integer divisor)
#   - Uses: modulo operation (%), integer division (//)

# Create a small spec tensor
# Shape: (batch=2, time_steps=12, features=4)
# This allows dr values like 2, 3, 4, 6 to divide 12 evenly
spec = torch.randn(2, 12, 4)

# Downsampling rate
dr = 3

ARGS = {
    "spec": spec,
    "dr": dr,
}
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, 4, 12) — shape (2, 4, 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
eval_helper Ran
ran on a synthesized input
ydkwim/ctal/code/run_s-reg-mosei.py
code served (permissive licence) · get_code("402f48cf719e8ff0")
How it ran

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

import numpy as np

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

# Create small sample data for testing eval_helper
# pred_label and true_label should be 1D arrays of floats
# The function expects them to have the same length
pred_label = np.array([0.5, 1.2, -0.8, 0.1, 2.5, -1.5, 0.0, 1.8], dtype=np.float32)
true_label = np.array([0.3, 1.0, -0.5, 0.2, 2.2, -1.3, 0.0, 2.0], dtype=np.float32)

ARGS = {
    "pred_label": pred_label,
    "true_label": true_label,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, sklearn — 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
get_centroids Ran
ran on a synthesized input
ydkwim/ctal/code/calculate_eer.py
code served (permissive licence) · get_code("5c03130d658096c8")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family float, shape (2, 3) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# The function takes embeddings and computes mean across dim=1
# embeddings needs to be a tensor with at least 2 dimensions
# dim=1 reduction suggests shape like (batch_size, sequence_length, embedding_dim)
# or similar where we average over dimension 1

embeddings = torch.randn(2, 4, 3)  # (batch=2, seq_len=4, embed_dim=3)

ARGS = {
    "embeddings": embeddings
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 3) — shape (2, 3) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_cossim Ran
ran on a synthesized input
ydkwim/ctal/code/calculate_eer.py
code served (permissive licence) · get_code("ddd85402bca56b00")
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, 3, 2) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn.functional as F

torch.manual_seed(0)

# Create small tensors with distinct dimensions
# embeddings shape: (speaker_ct, utterance_per_speaker_ct, embedding_size)
# Let's use: 2 speakers, 3 utterances per speaker, 4 embedding dimension
speaker_ct = 2
utterance_per_speaker_ct = 3
embedding_size = 4

embeddings = torch.randn(speaker_ct, utterance_per_speaker_ct, embedding_size)

# centroids shape: [speaker_ct, embedding_size]
centroids = torch.randn(speaker_ct, embedding_size)

ARGS = {
    "embeddings": embeddings,
    "centroids": centroids,
}
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, 3, 2) — shape (2, 3, 2) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_utterance_centroids Ran
ran on a synthesized input
ydkwim/ctal/code/calculate_eer.py
code served (permissive licence) · get_code("a0c7c003d46d7536")
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, 3, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# The function expects embeddings with shape (speaker_ct, utterance_per_speaker_ct, embedding_size)
# It performs operations: sum(dim=1), reshape, subtraction, division
# All of these work on tensors, so we need a valid tensor

# Create a small tensor with distinct dimension sizes:
# speaker_ct=2, utterance_per_speaker_ct=3, embedding_size=4
embeddings = torch.randn(2, 3, 4)

ARGS = {
    "embeddings": embeddings
}
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, 3, 4) — shape (2, 3, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
load_acoustic_data Ran
ran on a synthesized input
ydkwim/ctal/code/m2p_dataloader.py
code served (permissive licence) · get_code("e5abd0422f0b516c")
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 os
import torch
import numpy as np
import tempfile

# Create a temporary directory and a dummy .npy file
temp_dir = tempfile.mkdtemp()
dummy_data = np.random.randn(2, 4).astype(np.float32)
npy_filename = "acoustic_data.npy"
npy_filepath = os.path.join(temp_dir, npy_filename)
np.save(npy_filepath, dummy_data)

ARGS = {
    "npy_path": npy_filename,
    "npy_root": temp_dir,
}
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 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
load_semantic_data Ran
ran on a synthesized input
ydkwim/ctal/code/m2p_dataloader.py
code served (permissive licence) · get_code("7d4578febace3ca2")
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 os
import tempfile

class SimpleTokenizer:
    def __call__(self, text):
        return text.split()

# Create a temporary directory and file for testing
temp_dir = tempfile.mkdtemp()
temp_file = os.path.join(temp_dir, "test.txt")

# Write sample data in the expected CSV format (with comma-separated values)
# Format: field1,field2,field3 where field3 is the text we want
with open(temp_file, 'w') as f:
    f.write("id1,label1,hello world\n")
    f.write("id2,label2,test data\n")
    f.write("id3,label3,sample text\n")

# Extract just the filename to use as txt_path
txt_filename = "test.txt"

ARGS = {
    "txt_path": txt_filename,
    "txt_root": temp_dir,
    "tokenizer": SimpleTokenizer()
}
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
multiclass_acc Ran
ran, honoured the contract we drafted
ydkwim/ctal/code/run_s-reg-mosei.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from mc-eiu/mc-eiu, is pointer only (licence: NONE) · get_code("142e017e37f584c6")
How it ran

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

process_train_MLM_data Ran
ran on a synthesized input
ydkwim/ctal/code/m2p_mask.py
code served (permissive licence) · get_code("27ab2a9027ec745a")
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 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 torch

torch.manual_seed(0)

# Create a small spec tensor
# spec has shape (batch_size, seq_len, feature_dim)
# The function accesses spec[:, :, -1], so we need at least 3 dimensions
# Let's make it (batch=2, seq=4, features=3)
spec = torch.randint(0, 5, (2, 4, 3)).float()
# Set some values to 0 to test the masking logic
spec[0, 2:, -1] = 0
spec[1, 3:, -1] = 0

# Create a config dict with the expected keys
config = {
    'downsample_rate': 1,
    'hidden_size': 768
}

ARGS = {
    'spec': spec,
    'config': config
}
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 3, 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
ua_helper Ran
ran on a synthesized input
ydkwim/ctal/code/run_s-iemocap.py
code served (permissive licence) · get_code("f291cb1801e3b2b1")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned float64, ndim 0, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Analyze the function:
# - pred_label: used as index into np.eye(class_num), so should be array of integers in range [0, class_num)
# - true_label: used as index into np.eye(class_num), so should be array of integers in range [0, class_num)
# - class_num: integer, default 4

# Create small sample data
np.random.seed(0)

# Create predicted labels: array of integers in range [0, 4)
pred_label = np.array([0, 1, 2, 3, 0, 1])

# Create true labels: array of integers in range [0, 4)
true_label = np.array([0, 1, 1, 3, 2, 1])

# Number of classes
class_num = 4

ARGS = {
    "pred_label": pred_label,
    "true_label": true_label,
    "class_num": class_num,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
wa_helper Ran
ran on a synthesized input
ydkwim/ctal/code/run_s-iemocap.py
code served (permissive licence) · get_code("38621d5c1eb92ad6")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned float64, ndim 0, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
import torch

# Set seed for reproducibility
torch.manual_seed(0)

# Create small numpy arrays for predicted and true labels
# Using small integer arrays to represent class predictions
pred_label = np.array([0, 1, 2, 1, 0, 2])
true_label = np.array([0, 1, 1, 1, 0, 2])

ARGS = {
    "pred_label": pred_label,
    "true_label": true_label,
}
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 float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
AcousticModel Not yet run
unverified (harvested, no recorded run)
tal-ai/CTAL_EMNLP2021/code/m2p_model.py
pointer only (licence: GPL-3.0) · get_code("3a0fda00078a444e")
How it ran

The body of this sample is withheld: GPL-3.0, as GitHub detects it on this paper's repository, tal-ai/CTAL_EMNLP2021, is not in the permissive set Syntology serves inline. 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.

RobertaForMaskedAM Not yet run
unverified (harvested, no recorded run)
tal-ai/CTAL_EMNLP2021/code/m2p_model.py
pointer only (licence: GPL-3.0) · get_code("76c38b19897f7c7b")
How it ran

The body of this sample is withheld: GPL-3.0, as GitHub detects it on this paper's repository, tal-ai/CTAL_EMNLP2021, is not in the permissive set Syntology serves inline. 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.

RobertaM2Model Not yet run
unverified (harvested, no recorded run)
tal-ai/CTAL_EMNLP2021/code/m2p_model.py
pointer only (licence: GPL-3.0) · get_code("85af9ffb207dd2b2")
How it ran

The body of this sample is withheld: GPL-3.0, as GitHub detects it on this paper's repository, tal-ai/CTAL_EMNLP2021, is not in the permissive set Syntology serves inline. 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.

process_train_MAM_data Not yet run
unverified (harvested, no recorded run)
ydkwim/ctal/code/m2p_mask.py
code served (permissive licence) · get_code("fa1982cd98fbaf9d")
How it ran

no invocation recorded.

Also present in other papers' code

Of 19 samples lifted from this paper's linked repositories, 19 carry a function identity and 0 do not. Of the 19 distinct functions with an identity, 2 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 0 the same month, 1 later) and 17 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
multiclass_acc 5 other papers
1 dated earlier, 4 later
1906.00295 (2019-06)
Multimodal Transformer for Unaligned Multimodal Language Sequences
2407.02751 (2024-07)
Emotion and Intent Joint Understanding in Multimodal Conversation: A Benchmarking Dataset
2411.01409 (2024-11)
Classifier-guided Gradient Modulation for Enhanced Multimodal Learning
and 2 more not named here
get_centroids 2 other papers
2 dated earlier
1710.10467 (2017-10)
Generalized End-to-End Loss for Speaker Verification
2010.11607 (2020-10)
Backdoor Attack against Speaker Verification

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

Existing approaches for audio-language taskspecific prediction focus on building complicated late-fusion mechanisms. However, these models face challenges of overfitting with limited labels and poor generalization. In this paper, we present a Cross-modal Transformer for Audio-and-Language, i.e., CTAL, which aims to learn the intra-and inter-modalities connections between audio and language through two proxy tasks from a large number of audio-and-language pairs: masked language modeling and masked cross-modal acoustic modeling. After fine-tuning our CTAL model on multiple downstream audioand-language tasks, we observe significant improvements on different tasks, including emotion classification, sentiment analysis, and speaker verification. Furthermore, we design a fusion mechanism in the fine-tuning phase, which allows CTAL to achieve better performance. Lastly, we conduct detailed ablation studies to demonstrate that both our novel cross-modality fusion component and audiolanguage pre-training methods contribute to the promising results. The code and pretrained models are available at https:// github.com/tal-ai/CTAL_EMNLP2021.

For agents

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

get_harvested_code_for_paper("2109.00181")
get_code_for_paper("2109.00181")
have("2109.00181")

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

curl https://syntology.ai/api/ran/2109.00181.json

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

[![Syntology run record](https://syntology.ai/api/ran/2109.00181.svg)](https://syntology.ai/paper/2109.00181)

Connect an agent — have() is free.