SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1907.11692 · arXiv.org · 2019

RoBERTa: A Robustly Optimized BERT Pretraining Approach

Myle Ott, Naman Goyal, Mike Lewis, Luke Zettlemoyer, Danqi Chen, Mandar Joshi, Omer Levy, Jingfei Du, Veselin Stoyanov, Yinhan Liu

arXiv · PDF · Open in the Atlas

Code that ran

37 of 48 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (11 of them constructed an object rather than computing a result; 3 honoured the contract we drafted, 0 violated it and 32 have no contract verdict; 2 failures of our instrument; 11 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/1907.11692.json · badge: /api/ran/1907.11692.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
GeorgeLuImmortal/Hierarchical-BERT-Model-with-Limited-Labelled-Data community
Papers with Code archive
8 ran · 2 no recorded run
traviscoan/cards community (classifier label)
classifier basis not recorded on the link
7 ran · 1 no recorded run
xiaoqian19940510/text-classification-surveys community
not in the paper's declared repositories
6 ran · 0 no recorded run
viethoang1512/kpa community (classifier label)
classifier basis not recorded on the link
4 ran · 0 no recorded run
ibm/vira-intent-discovery community (classifier label)
classifier basis not recorded on the link
3 ran · 0 no recorded run
lvyufeng/bert4ms community (classifier label)
classifier basis not recorded on the link
2 ran · 2 no recorded run
abdumaa/hiqualprop community
Papers with Code archive
2 ran · 0 no recorded run
wzzzd/LM_NER community
not in the paper's declared repositories
1 ran · 1 no recorded run
hkuds/easyrec community
not in the paper's declared repositories
1 ran · 1 no recorded run
salesforce/codet5 community
not in the paper's declared repositories
1 ran · 0 no recorded run
few-shot-NER-benchmark/BaselineCode community
not in the paper's declared repositories
1 ran · 0 no recorded run
pisalore/roberta_results community (classifier label)
classifier basis not recorded on the link
1 ran · 0 no recorded run
expertailab/spaceqa community
Papers with Code archive
0 ran · 1 no recorded run
kaushaltrivedi/fast-bert community
not in the paper's declared repositories
0 ran · 1 no recorded run
zaradana/Fast_BERT community
not in the paper's declared repositories
0 ran · 1 no recorded run
blawok/named-entity-recognition community
Papers with Code archive
0 ran · 1 no recorded run
FunctionStatusWhere it lives
BertAttention Ran
ran on a synthesized input · constructed an object; no result was computed
GeorgeLuImmortal/Hierarchical-BERT-Model-with-Limited-Labelled-Data/run_hbm.py
pointer only (licence: NONE) · get_code("ad7f6e57dedf08aa")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, GeorgeLuImmortal/Hierarchical-BERT-Model-with-Limited-Labelled-Data 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.

BertEmbeddings Ran
ran on a synthesized input
GeorgeLuImmortal/Hierarchical-BERT-Model-with-Limited-Labelled-Data/run_hbm.py
pointer only (licence: NONE) · get_code("c3af34b4f438eaea")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, GeorgeLuImmortal/Hierarchical-BERT-Model-with-Limited-Labelled-Data 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.

BertIntermediate Ran
ran on a synthesized input · constructed an object; no result was computed
GeorgeLuImmortal/Hierarchical-BERT-Model-with-Limited-Labelled-Data/run_hbm.py
pointer only (licence: NONE) · get_code("6e60ffe93c5ec5fa")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, GeorgeLuImmortal/Hierarchical-BERT-Model-with-Limited-Labelled-Data 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.

BertModel Ran
ran on a synthesized input
GeorgeLuImmortal/Hierarchical-BERT-Model-with-Limited-Labelled-Data/run_hbm.py
pointer only (licence: NONE) · get_code("f43b4161b7bdba24")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, GeorgeLuImmortal/Hierarchical-BERT-Model-with-Limited-Labelled-Data 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.

BertOutput Ran
ran on a synthesized input
GeorgeLuImmortal/Hierarchical-BERT-Model-with-Limited-Labelled-Data/run_hbm.py
pointer only (licence: NONE) · get_code("5e5629cb8e93e004")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, GeorgeLuImmortal/Hierarchical-BERT-Model-with-Limited-Labelled-Data 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.

BertPooler Ran
ran on a synthesized input · constructed an object; no result was computed
GeorgeLuImmortal/Hierarchical-BERT-Model-with-Limited-Labelled-Data/run_hbm.py
pointer only (licence: NONE) · get_code("4e755eff48bf0752")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, GeorgeLuImmortal/Hierarchical-BERT-Model-with-Limited-Labelled-Data 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.

BertSelfAttention Ran
ran on a synthesized input
GeorgeLuImmortal/Hierarchical-BERT-Model-with-Limited-Labelled-Data/run_hbm.py
pointer only (licence: NONE) · get_code("df8363b5430892a0")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, GeorgeLuImmortal/Hierarchical-BERT-Model-with-Limited-Labelled-Data 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.

BertSelfOutput Ran
ran on a synthesized input · constructed an object; no result was computed
GeorgeLuImmortal/Hierarchical-BERT-Model-with-Limited-Labelled-Data/run_hbm.py
pointer only (licence: NONE) · get_code("fcf3cd282eb03efe")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, GeorgeLuImmortal/Hierarchical-BERT-Model-with-Limited-Labelled-Data 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.

CRF Ran
ran on a synthesized input
wzzzd/LM_NER/model/roberta_crf.py
pointer only (licence: NONE) · get_code("e65cfdc18cd8c48f")
How it ran

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

ClassificationHead Ran
ran on a synthesized input · constructed an object; no result was computed
abdumaa/hiqualprop/full/custom_sequence_classification.py
pointer only (licence: NONE) · get_code("7b8d60647337c2b1")
How it ran

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

CustomRobertaForSequenceClassification Ran
ran on a synthesized input
abdumaa/hiqualprop/full/custom_sequence_classification.py
pointer only (licence: NONE) · get_code("0196b04010ddfa00")
How it ran

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

Encoder Ran
ran on a synthesized input · constructed an object; no result was computed
xiaoqian19940510/text-classification-surveys/ChineseTextClassification/models/Transformer.py
pointer only (licence: NONE) · get_code("68faf3c1193d95e7")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, xiaoqian19940510/text-classification-surveys 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.

Model Ran
ran on a synthesized input · constructed an object; no result was computed
xiaoqian19940510/text-classification-surveys/ChineseTextClassification/models/Transformer.py
pointer only (licence: NONE) · get_code("9d6930c9ef7f273b")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, xiaoqian19940510/text-classification-surveys 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.

Multi_Head_Attention Ran
ran on a synthesized input · constructed an object; no result was computed
xiaoqian19940510/text-classification-surveys/ChineseTextClassification/models/Transformer.py
pointer only (licence: NONE) · get_code("717e1f9d4eae4141")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, xiaoqian19940510/text-classification-surveys 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.

Pooler Ran
ran on a synthesized input · constructed an object; no result was computed
hkuds/easyrec/model.py
pointer only (licence: NONE) · get_code("dba0b61c9d62f82f")
How it ran

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

Position_wise_Feed_Forward Ran
ran on a synthesized input · constructed an object; no result was computed
xiaoqian19940510/text-classification-surveys/ChineseTextClassification/models/Transformer.py
pointer only (licence: NONE) · get_code("c15e979783157fb0")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, xiaoqian19940510/text-classification-surveys 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.

Positional_Encoding Ran
ran on a synthesized input · constructed an object; no result was computed
xiaoqian19940510/text-classification-surveys/ChineseTextClassification/models/Transformer.py
pointer only (licence: NONE) · get_code("ce8217cf9489c436")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, xiaoqian19940510/text-classification-surveys 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.

RobertaClassificationHead Ran
ran on a synthesized input
salesforce/codet5/CodeT5/models.py
code served (permissive licence) · get_code("17243795c9ab1ec3")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned Tensor, ndim 2, dtype family float, shape (1, 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 as nn

class RobertaClassificationHead(nn.Module):
    """Head for sentence-level classification tasks."""

    def __init__(self, config):
        super().__init__()
        self.dense = nn.Linear(config.hidden_size * 2, config.hidden_size)
        self.out_proj = nn.Linear(config.hidden_size, 2)

    def forward(self, x, **kwargs):
        x = x.reshape(-1, x.size(-1) * 2)
        x = self.dense(x)
        x = torch.tanh(x)
        x = self.out_proj(x)
        return x


class Config:
    def __init__(self, hidden_size):
        self.hidden_size = hidden_size


INIT_ARGS = {
    'config': Config(hidden_size=4)
}

ARGS = {
    'x': torch.randn(2, 4)
}

if __name__ == '__main__':
    model = RobertaClassificationHead(**INIT_ARGS)
    output = model(**ARGS)
    print(f"Output shape: {output.shape}")
    print(f"Output: {output}")
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
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 (1, 2) — shape (1, 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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
RobertaNER Ran
ran on a synthesized input
few-shot-NER-benchmark/BaselineCode/src/meta_model.py
pointer only (licence: NONE) · get_code("152609ff366dfb9e")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, few-shot-NER-benchmark/BaselineCode 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.

Scaled_Dot_Product_Attention Ran
ran on a synthesized input
xiaoqian19940510/text-classification-surveys/ChineseTextClassification/models/Transformer.py
pointer only (licence: NONE) · get_code("efea4e3ee53c36c9")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, xiaoqian19940510/text-classification-surveys 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.

contrastive_loss Ran
ran on a synthesized input
viethoang1512/kpa/qs_kpa/baselines/tf_models.py
code served (permissive licence) · get_code("a5f2db749ea77f58")
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 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 tensorflow as tf
import numpy as np

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

# Create small tensors for contrastive loss
# y: binary labels (0 or 1) indicating whether pair is similar
# preds: predicted distances between pairs
batch_size = 2
pred_dim = 1

# Create y as binary labels (shape: batch_size,)
y = tf.constant([0.0, 1.0], dtype=tf.float32)

# Create preds as predicted distances (shape: batch_size,)
preds = tf.constant([0.5, 1.2], dtype=tf.float32)

# margin is a scalar hyperparameter
margin = 1.0

ARGS = {
    "y": y,
    "preds": preds,
    "margin": margin
}
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 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
convert_label_to_similarity Ran
ran; our contract draft was wrong, not the code
viethoang1512/kpa/qs_kpa/losses/circle_loss.py
code served (permissive licence) · get_code("6708df0bbbc13e34")
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)

# Create small normalized feature tensor (batch_size=2, feature_dim=4)
# Features should be normalized (L2 norm = 1) for similarity computation
normed_feature = torch.randn(2, 4)
normed_feature = normed_feature / torch.norm(normed_feature, dim=1, keepdim=True)

# Create label tensor (batch_size=2,)
label = torch.tensor([0, 1])

ARGS = {
    "normed_feature": normed_feature,
    "label": label,
}
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
convert_state_dict Ran
ran on a synthesized input
lvyufeng/bert4ms/cybertron/utils.py
code served (permissive licence) · get_code("8afaa985063f723e")
How it ran

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

import torch
import torch.nn as nn
from unittest.mock import Mock, patch
import sys

torch.manual_seed(0)

# Create a minimal torch.nn.Module
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 3)
        self.conv = nn.Conv1d(4, 2, 3)
    
    def forward(self, x):
        return self.linear(x)

# Create a torch module instance with state_dict
torch_module = SimpleModel()
torch_module.eval()

# Mock the cybertron.models conversion function
def mock_torch_to_mindspore(state_dict):
    """Mock conversion function that returns MindSpore-compatible checkpoint format"""
    result = []
    for name, param in state_dict.items():
        result.append({
            'name': name,
            'data': param.cpu().detach().numpy()
        })
    return result

# Patch the importlib to return our mock conversion function
def mock_import_module(module_path):
    mock_module = Mock()
    mock_module.torch_to_mindspore = mock_torch_to_mindspore
    return mock_module

# Create arguments for the function
ARGS = {
    'module_or_pth_file': torch_module,
    'model': 'test_model'
}

# Setup the patch for importlib and mindspore
import importlib as real_importlib
original_import = real_importlib.import_module

def patched_import_module(name):
    if 'cybertron.models' in name:
        mock_module = Mock()
        mock_module.torch_to_mindspore = mock_torch_to_mindspore
        return mock_module
    return original_import(name)

real_importlib.import_module = patched_import_module

# Mock MindSpore imports
sys.modules['mindspore'] = Mock()
sys.modules['mindspore.train'] = Mock()
sys.modules['mindspore.train.serialization'] = Mock()

from unittest.mock import MagicMock
mock_mindspore = MagicMock()
sys.modules['mindspore'] = mock_mindspore
sys.modules['mindspore.train'] = MagicMock()
sys.modules['mindspore.train.serialization'] = MagicMock()

sys.modules['mindspore'].Tensor = MagicMock()
sys.modules['mindspore'].Parameter = MagicMock(side_effect=lambda x, y: {'name': y, 'data': x})
sys.modules['mindspore.train.serialization'].save_checkpoint = MagicMock()

ARGS = {
    'module_or_pth_file': torch_module,
    'model': 'test_model'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
mindspore, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
convert_to_single_none_row_df Ran
ran on a synthesized input
ibm/vira-intent-discovery/trim.py
code served (permissive licence) · get_code("e06455e72782e6d0")
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 DataFrame. 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 pandas as pd

# The function expects a DataFrame with 'size' and 'ids' columns
# that can be aggregated with .agg() and summed

none_df = pd.DataFrame({
    'size': [10, 20, 30],
    'ids': [[1, 2], [3, 4], [5, 6]]
})

ARGS = {
    'none_df': none_df
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 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
cos_sim Ran
ran, honoured the contract we drafted
viethoang1512/kpa/qs_kpa/losses/multiple_negatives_ranking_loss.py
code served (permissive licence) · get_code("8ca90d5388372be5")
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)

# Create two small tensors with distinct dimensions
# a: batch_size=2, embedding_dim=4
a = torch.randn(2, 4)

# b: batch_size=3, embedding_dim=4 (distinct first dim to probe shapes)
b = torch.randn(3, 4)

ARGS = {
    "a": a,
    "b": b,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type 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
dot_score Ran
ran, honoured the contract we drafted
viethoang1512/kpa/qs_kpa/losses/multiple_negatives_ranking_loss.py
code served (permissive licence) · get_code("afa18317f44e63ad")
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)

# Create two small tensors for the dot_score function
# a: shape (2, 4) - 2 vectors of dimension 4
# b: shape (3, 4) - 3 vectors of dimension 4
# This will produce a (2, 3) output matrix of dot products

a = torch.randn(2, 4)
b = torch.randn(3, 4)

ARGS = {
    "a": a,
    "b": b,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type 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
drop_duplicates Ran
ran on a synthesized input
traviscoan/cards/utils.py
code served (permissive licence) · get_code("163e38293198fa6a")
How it ran

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

import torch
import numpy as np

# Analyze the function: drop_duplicates takes an iterable (seq) and returns a list
# with duplicates removed. The function uses: iteration, set membership testing,
# and set.add() method. It needs an iterable that can contain hashable elements.

# Create test data - a simple list with some duplicates
seq = [1, 2, 3, 2, 4, 1, 5, 3]

ARGS = {
    "seq": seq
}
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 int — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
estimate_ensemble Ran
ran on a synthesized input
traviscoan/cards/fit/ensemble.py
code served (permissive licence) · get_code("9f11fd760e57188c")
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
from sklearn.preprocessing import LabelEncoder

# Create sample probability arrays
np.random.seed(0)
probs1 = np.random.rand(2, 4)  # 2 samples, 4 classes
probs1 = probs1 / probs1.sum(axis=1, keepdims=True)  # normalize to valid probabilities

probs2 = np.random.rand(2, 4)  # 2 samples, 4 classes
probs2 = probs2 / probs2.sum(axis=1, keepdims=True)  # normalize to valid probabilities

# Create a LabelEncoder with classes
le = LabelEncoder()
le.fit(['cat', 'dog', 'bird', 'fish'])

ARGS = {
    'probs1': probs1,
    'probs2': probs2,
    'le': le
}
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 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
filter_all_data Ran
ran on a synthesized input
ibm/vira-intent-discovery/prepare.py
code served (permissive licence) · get_code("d4db5e37c5831f1a")
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 torch
import numpy as np

torch.manual_seed(0)

# Create a valid row dictionary that passes all filter conditions
row = {
    'text': 'This is a valid sample text with proper content',
    'is_feedback': False
}

# Verify the row passes the filter
MASKED_TOKENS = {'PERSON', 'DATE', 'LOCATION', 'EMAIL', 'PHONE_NUMBER', 'TOXIC'}

text = row['text']
alpha_numeric_chars = len([c for c in text if c.isalnum() or c == ' '])
non_masked_tokens_count = len([w for w in text.split() if w not in MASKED_TOKENS])

# Check conditions:
# 1. 0 < len(text) < 250: text length is 45, passes
# 2. alpha_numeric_chars / len(text) > 0.75: should be high ratio
# 3. not row['is_feedback']: is_feedback is False, passes
# 4. non_masked_tokens_count > 1: should have multiple non-masked words

ARGS = {
    'row': row
}
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 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
filter_oracle_train Ran
ran on a synthesized input
ibm/vira-intent-discovery/prepare.py
code served (permissive licence) · get_code("57b444571a95cb8a")
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 torch

torch.manual_seed(0)

# The function expects a dict-like object (a row) with keys 'dialog_act' and 'is_profanity'
row = {
    'dialog_act': 'inform',  # A dialog act that is NOT in the filtered list
    'is_profanity': False    # Not profanity
}

ARGS = {
    'row': row
}
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 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
flatten Ran
ran, honoured the contract we drafted
traviscoan/cards/utils.py
code served (permissive licence) · get_code("954d7736e2d968f3")
How it ran

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

import torch

torch.manual_seed(0)

# The function takes a list of lists and flattens it
# We need to create a simple list of lists with different sublists

list_of_lists = [
    [1, 2, 3],
    [4, 5],
    [6, 7, 8, 9],
    [10]
]

ARGS = {
    "list_of_lists": list_of_lists
}
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 list, ndim 1, dtype family int — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_tasks Ran
ran; our contract draft was wrong, not the code
pisalore/roberta_results/glue_tasks/download_glue_data.py
pointer only (licence: MIT); the stored copy get_code serves, from nyu-mll/GLUE-baselines, is pointer only (licence: NONE) · get_code("3caf26b311c98658")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from nyu-mll/GLUE-baselines, 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.

invert_mask Ran
ran on a synthesized input
lvyufeng/bert4ms/cybertron/models/bart.py
code served (permissive licence) · get_code("e8ae42ea576c0232")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family float, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Create a 2D attention mask tensor with shape (batch_size=2, seq_length=4)
# Using integers 0 and 1 to represent padding and non-padding tokens
attention_mask = torch.tensor([
    [1, 1, 1, 0],
    [1, 1, 0, 0]
], dtype=torch.long)

ARGS = {
    "attention_mask": attention_mask
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
predict_claim Ran
ran on a synthesized input
traviscoan/cards/fit/ensemble.py
code served (permissive licence) · get_code("712ccc81e62f75b9")
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

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

# probs: numpy array of shape (batch_size, num_classes)
# We'll use batch_size=2 and num_classes=4
probs = np.random.rand(2, 4)
# Normalize to valid probabilities
probs = probs / probs.sum(axis=1, keepdims=True)

# labels: list of label strings
labels = ["cat", "dog", "bird", "fish"]

ARGS = {
    "probs": probs,
    "labels": labels,
}
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
read_csv Ran
ran on a synthesized input
traviscoan/cards/utils.py
code served (permissive licence) · get_code("df853d8826e7a476")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 2, dtype family 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 csv
import tempfile
import os

# Create a temporary CSV file for testing
temp_dir = tempfile.mkdtemp()
csv_path = os.path.join(temp_dir, "test.csv")

# Write sample CSV data
with open(csv_path, 'w', newline='', encoding='utf-8') as f:
    writer = csv.writer(f)
    writer.writerow(["header1", "header2", "header3"])
    writer.writerow(["value1", "value2", "value3"])
    writer.writerow(["value4", "value5", "value6"])

ARGS = {
    "path": csv_path,
    "remove_header": False
}
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 2, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
remove_between_square_brackets Ran
ran on a synthesized input
traviscoan/cards/preprocess.py
code served (permissive licence) · get_code("cd2c2c1263e0882d")
How it ran

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

import re

def remove_between_square_brackets(text):
    """
    Removes text between square brackets from a string.

    Args:
        text (str): Text to clean
    
    Returns:
        str: Cleaned text.
    """
    return re.sub('\\[[^]]*\\]', '', text)


ARGS = {
    'text': 'Hello [world] this is [a test] string with [brackets]'
}
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 str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
remove_non_ascii Ran
ran on a synthesized input
traviscoan/cards/preprocess.py
code served (permissive licence) · get_code("a9831bf0d050a65c")
How it ran

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

import unicodedata

# The function takes a single parameter: text (str)
# It uses unicodedata.normalize() and string encoding/decoding methods
# We need to provide a string that contains non-ASCII characters to demonstrate the function

ARGS = {
    "text": "Héllo Wörld! 你好 مرحبا Ñoño"
}
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 str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
BertEncoder Not yet run
unverified (harvested, no recorded run)
GeorgeLuImmortal/Hierarchical-BERT-Model-with-Limited-Labelled-Data/run_hbm.py
pointer only (licence: NONE) · get_code("00ea7705507db813")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, GeorgeLuImmortal/Hierarchical-BERT-Model-with-Limited-Labelled-Data 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.

BertLayer Not yet run
unverified (harvested, no recorded run)
GeorgeLuImmortal/Hierarchical-BERT-Model-with-Limited-Labelled-Data/run_hbm.py
pointer only (licence: NONE) · get_code("f2ab856479b90e8a")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, GeorgeLuImmortal/Hierarchical-BERT-Model-with-Limited-Labelled-Data 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.

Easyrec Not yet run
unverified (harvested, no recorded run)
hkuds/easyrec/model.py
pointer only (licence: NONE) · get_code("b2bd29f8a81790e2")
How it ran

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

QuestionAnswering Not yet run
unverified (harvested, no recorded run)
expertailab/spaceqa/question_answering/src/question_answering/question_answering.py
code served (permissive licence) · get_code("da835c893505251b")
How it ran

no invocation recorded.

RoBertaCRF Not yet run
unverified (harvested, no recorded run)
wzzzd/LM_NER/model/roberta_crf.py
pointer only (licence: NONE) · get_code("7d40208cd5d0da23")
How it ran

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

RobertaForMultiLabelSequenceClassification Not yet run
unverified (harvested, no recorded run)
kaushaltrivedi/fast-bert/fast_bert/modeling.py
code served (permissive licence) · get_code("b3e0c37e1e2b9503")
How it ran

no invocation recorded.

RobertaForMultiLabelSequenceClassification Not yet run
unverified (harvested, no recorded run)
zaradana/Fast_BERT/0_fast_bert/modeling.py
code served (permissive licence) · get_code("6788010bd752077c")
How it ran

no invocation recorded.

fit_logistic_classifier Not yet run
unverified (harvested, no recorded run)
traviscoan/cards/fit/logistic.py
code served (permissive licence) · get_code("53e20baefb5f4cd1")
How it ran

no invocation recorded.

load_from_cache Not yet run
unverified (harvested, no recorded run)
lvyufeng/bert4ms/cybertron/utils.py
code served (permissive licence) · get_code("e462fa293bb314aa")
How it ran

no invocation recorded.

shift_tokens_right Not yet run
unverified (harvested, no recorded run)
lvyufeng/bert4ms/cybertron/models/bart.py
code served (permissive licence) · get_code("8f0494a0f9a303b8")
How it ran

no invocation recorded.

train Not yet run
unverified (harvested, no recorded run)
blawok/named-entity-recognition/src/source_roberta/train_roberta.py
code served (permissive licence) · get_code("92e80bcaa1f41216")
How it ran

no invocation recorded.

Also present in other papers' code

Of 48 samples lifted from this paper's linked repositories, 48 carry a function identity and 0 do not. Of the 47 distinct functions with an identity, 11 are also present in other papers whose code we hold (6 with at least one paper dated earlier, 0 the same month, 6 later) and 36 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
get_tasks 13 other papers
4 dated earlier, 9 later
1711.05225 (2017-11)
CheXNet: Radiologist-Level Pneumonia Detection on Chest X-Rays with Deep Learning
1804.07461 (2018-04)
GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding
1810.04805 (2018-10)
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
and 10 more not named here
cos_sim 7 other papers
7 later
2104.08663 (2021-04)
BEIR: A Heterogeneous Benchmark for Zero-shot Evaluation of Information Retrieval Models
2310.20501 (2023-10)
Neural Retrievers are Biased Towards LLM-Generated Content
2024.findings-acl.421 (2024 (year only))
Cocktail: A Comprehensive Information Retrieval Benchmark with LLM-Generated Documents Integration
and 4 more not named here
dot_score 6 other papers
6 later
2104.08663 (2021-04)
BEIR: A Heterogeneous Benchmark for Zero-shot Evaluation of Information Retrieval Models
2310.20501 (2023-10)
Neural Retrievers are Biased Towards LLM-Generated Content
2024.findings-acl.421 (2024 (year only))
Cocktail: A Comprehensive Information Retrieval Benchmark with LLM-Generated Documents Integration
and 3 more not named here
convert_label_to_similarity 2 other papers
2 later
2002.10857 (2020-02)
Circle Loss: A Unified Perspective of Pair Similarity Optimization
2204.08381 (2022-04)
Multiple-environment Self-adaptive Network for Aerial-view Geo-localization
CRF
brick of this paper's assembly: model
1 other paper
1 dated earlier · 1 only as an assembly brick
1810.04805 (2018-10)
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Pooler
brick of this paper's assembly: model
1 other paper
1 later · 1 only as an assembly brick
1908.10084 (2019-08)
Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks
convert_state_dict 1 other paper
1 dated earlier
1810.04805 (2018-10)
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
flatten 1 other paper
1 later
2502.01051 (2025-02)
Diffusion Model as a Noise-Aware Latent Reward Model for Step-Level Preference Optimization
invert_mask 1 other paper
1 dated earlier
1810.04805 (2018-10)
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
load_from_cache 1 other paper
1 dated earlier
1810.04805 (2018-10)
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
shift_tokens_right 1 other paper
1 dated earlier
1810.04805 (2018-10)
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

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

Language model pretraining has led to significant performance gains but careful comparison between different approaches is challenging. Training is computationally expensive, often done on private datasets of different sizes, and, as we will show, hyperparameter choices have significant impact on the final results. We present a replication study of BERT pretraining (Devlin et al., 2019) that carefully measures the impact of many key hyperparameters and training data size. We find that BERT was significantly undertrained, and can match or exceed the performance of every model published after it. Our best model achieves state-of-the-art results on GLUE, RACE and SQuAD. These results highlight the importance of previously overlooked design choices, and raise questions about the source of recently reported improvements. We release our models and code. 1

For agents

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

get_harvested_code_for_paper("1907.11692")
get_code_for_paper("1907.11692")
have("1907.11692")

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

curl https://syntology.ai/api/ran/1907.11692.json

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

[![Syntology run record](https://syntology.ai/api/ran/1907.11692.svg)](https://syntology.ai/paper/1907.11692)

Connect an agent — have() is free.