SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1903.08983 · 2019

SemEval-2019 Task 6: Identifying and Categorizing Offensive Language in Social Media (OffensEval)

arXiv · PDF · Open in the Atlas

Code that ran

9 of 10 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; 0 honoured the contract we drafted, 0 violated it and 9 have no contract verdict; 0 failures of our instrument; 1 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/1903.08983.json · badge: /api/ran/1903.08983.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
causalate-mitigates-bias/causal-ate-mitigates-bias community (classifier label)
classifier basis not recorded on the link
9 ran · 1 no recorded run
FunctionStatusWhere it lives
compute_perturbation_scores Ran
ran on a synthesized input
causalate-mitigates-bias/causal-ate-mitigates-bias/archived/train_ate_20240708.py
code served (permissive licence) · get_code("fe2766538003bcfe")
How it ran

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

import torch
import torch.nn as nn
from torch.utils.data import DataLoader

torch.manual_seed(0)

# Create a minimal model that accepts long tensor input and outputs scores
class SimpleModel(nn.Module):
    def __init__(self, vocab_size=100, output_dim=1):
        super().__init__()
        self.embedding = nn.Embedding(vocab_size, 8)
        self.linear = nn.Linear(8, output_dim)
    
    def forward(self, x):
        # x shape: (batch_size, seq_len)
        embedded = self.embedding(x)  # (batch_size, seq_len, 8)
        pooled = embedded.mean(dim=1)  # (batch_size, 8)
        output = self.linear(pooled)  # (batch_size, output_dim)
        return output

# Create model instance
model = SimpleModel(vocab_size=100, output_dim=1)

# Create perturbed_loader - list of (perturbed_batch, rejoined_batch) tuples
# Each batch is a tensor of shape (batch_size, seq_len) with long values
perturbed_loader = [
    (
        torch.randint(0, 100, (2, 4)),  # perturbed_batch: (2, 4)
        torch.randint(0, 100, (2, 3))   # rejoined_batch: (2, 3) - different seq_len
    ),
    (
        torch.randint(0, 100, (2, 5)),  # perturbed_batch: (2, 5)
        torch.randint(0, 100, (2, 4))   # rejoined_batch: (2, 4)
    )
]

device = 'cpu'
padding_value = 0

ARGS = {
    'model': model,
    'perturbed_loader': perturbed_loader,
    'device': device,
    'padding_value': padding_value
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, 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
compute_perturbation_scores_old Ran
ran on a synthesized input
causalate-mitigates-bias/causal-ate-mitigates-bias/archived/train_ate_202407082243.py
code served (permissive licence) · get_code("5461a6a98fcc9b1e")
How it ran

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

import torch
import torch.nn as nn
from torch.utils.data import DataLoader

torch.manual_seed(0)

# Define a minimal model that satisfies the function's usage
class SimpleModel(nn.Module):
    def __init__(self, vocab_size=100, output_size=1):
        super().__init__()
        self.embedding = nn.Embedding(vocab_size, 8)
        self.linear = nn.Linear(8, output_size)
    
    def forward(self, x):
        # x shape: (batch_size, seq_len)
        embedded = self.embedding(x)  # (batch_size, seq_len, 8)
        pooled = embedded.mean(dim=1)  # (batch_size, 8)
        output = self.linear(pooled)  # (batch_size, output_size)
        return output

# Create model instance
model = SimpleModel(vocab_size=100, output_size=1)

# Create perturbed_loader as a list of (perturbed_batch, rejoined_batch) tuples
# Each batch contains token indices (long tensors)
perturbed_loader = [
    (
        torch.randint(0, 100, (2, 4), dtype=torch.long),  # perturbed_batch: (batch_size=2, seq_len=4)
        torch.randint(0, 100, (2, 4), dtype=torch.long),  # rejoined_batch: (batch_size=2, seq_len=4)
    ),
    (
        torch.randint(0, 100, (2, 4), dtype=torch.long),
        torch.randint(0, 100, (2, 4), dtype=torch.long),
    ),
]

device = 'cpu'

ARGS = {
    'model': model,
    'perturbed_loader': perturbed_loader,
    'device': device,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, 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
generate_perturbed_dataloader Ran
ran on a synthesized input
causalate-mitigates-bias/causal-ate-mitigates-bias/utilities/generate_ate_data.py
code served (permissive licence) · get_code("9096f3e3c31357a5")
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 DataLoader. 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 random
from torch.utils.data import DataLoader

torch.manual_seed(0)
random.seed(0)

class SimpleVocab:
    def __init__(self, words):
        self.words = words
    
    def get_itos(self):
        return self.words

# Create a minimal vocab
vocab = SimpleVocab(["the", "cat", "dog", "sat", "on", "mat", "quick", "brown"])

# Create minimal training data: list of (text, label) tuples
train_data = [
    ("the cat sat on the mat", 0),
    ("the dog sat on the mat", 1),
    ("the quick brown cat", 0),
]

# Create a simple data loader
def collate_fn_train(batch):
    texts, labels = zip(*batch)
    return list(texts), list(labels)

train_loader = DataLoader(
    train_data,
    batch_size=2,
    shuffle=False,
    collate_fn=collate_fn_train
)

# Define all arguments for generate_perturbed_dataloader
ARGS = {
    "train_loader": train_loader,
    "vocab": vocab,
    "batch_size": 2,
    "perturbation_rate": 0.5,
    "num_perturbations": 2,
    "n_gram_length": 2,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, tqdm — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataLoader — a DataLoader output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataLoader 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
getATEScores Ran
ran on a synthesized input
causalate-mitigates-bias/causal-ate-mitigates-bias/experiments/compute_ate_scores.py
code served (permissive licence) · get_code("95bdecf0755bcbdc")
How it ran

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

import numpy as np
import pandas as pd
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.linear_model import LogisticRegression
import tempfile
import os

# Create a simple mock vectorizer that has a transform method
vectorizer = TfidfVectorizer(max_features=10)

# Create sample text data to fit the vectorizer
sample_texts = [
    "this is a test sentence",
    "another sample text here",
    "some more example data",
    "final test document"
]
vectorizer.fit(sample_texts)

# Create a simple mock classifier with predict_proba method
classifier = LogisticRegression(random_state=0)
X_dummy = vectorizer.transform(sample_texts)
y_dummy = np.array([0, 1, 0, 1])
classifier.fit(X_dummy, y_dummy)

# Create input dataframe with required columns
input_df = pd.DataFrame({
    'text': [
        'this is original text',
        'another original sentence'
    ],
    'unmasked_sentences': [
        'this is replaced text',
        'another replaced sentence'
    ],
    'masked_words': [
        'word1',
        'word2'
    ]
})

# Create temporary output directory
temp_dir = tempfile.mkdtemp()
output_file1 = os.path.join(temp_dir, 'unmasked_with_scores.csv')
output_file2 = os.path.join(temp_dir, 'scores.csv')

ARGS = {
    'input_unmasked_df': input_df,
    'input_classifier': classifier,
    'input_vectorizer': vectorizer,
    'output_unmasked_file_with_scores': output_file1,
    'output_ate_csv_file': output_file2
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas, sklearn — 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
perturb_sentence Ran
ran on a synthesized input
causalate-mitigates-bias/causal-ate-mitigates-bias/archived/generate_ate_data_20240710.py
code served (permissive licence) · get_code("7b4df577196dbcd5")
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 random

# Set seed for reproducibility
random.seed(0)

# Define the arguments
original_input = "the quick brown fox jumps over the lazy dog"

vocab = ["the", "a", "quick", "fast", "brown", "red", "fox", "dog", "cat", "jumps", "runs", "over", "under", "lazy", "sleepy"]

num_perturbations = 5

ARGS = {
    "original_input": original_input,
    "vocab": vocab,
    "num_perturbations": num_perturbations,
}
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
perturb_sentence Ran
ran on a synthesized input
causalate-mitigates-bias/causal-ate-mitigates-bias/archived/generate_ate_data_20240712_1316.py
code served (permissive licence) · get_code("8cbd7c91d1c7eb17")
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 random

# Set seed for reproducibility
random.seed(0)

# Define arguments for perturb_sentence function
original_input = "the quick brown fox jumps over the lazy dog"

vocab = ["cat", "dog", "bird", "fish", "elephant", "car", "house", "tree", "flower", "book"]

perturbation_rate = 0.5

num_perturbations = 10

n_gram_length = 5

ARGS = {
    "original_input": original_input,
    "vocab": vocab,
    "perturbation_rate": perturbation_rate,
    "num_perturbations": num_perturbations,
    "n_gram_length": n_gram_length
}
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
perturb_sentence Ran
ran on a synthesized input
causalate-mitigates-bias/causal-ate-mitigates-bias/archived/train_ate_20240708.py
code served (permissive licence) · get_code("77af5a0a462285b5")
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 random

# Set seed for reproducibility
random.seed(42)

# Define a simple sentence
sentence = "the quick brown fox jumps over the lazy dog"

# Define a vocabulary as a list of words
vocab = ["the", "a", "quick", "slow", "brown", "red", "fox", "dog", "cat", "jumps", "runs", "walks", "over", "under", "on"]

# Define the number of perturbations
num_perturbations = 5

# Construct the ARGS dictionary
ARGS = {
    "sentence": sentence,
    "vocab": vocab,
    "num_perturbations": num_perturbations
}
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
perturb_sentence Ran
ran on a synthesized input
causalate-mitigates-bias/causal-ate-mitigates-bias/utilities/generate_ate_data.py
code served (permissive licence) · get_code("5c31fb2c09b40f18")
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 random

# Set seed for reproducibility
random.seed(0)

# Create a small vocabulary list
vocab = ["cat", "dog", "bird", "fish", "tree", "house", "car", "book", "apple", "run", "jump", "sleep", "eat", "play", "walk"]

# Create a sample input sentence (space-separated tokens)
original_input = "the cat sat on the mat and looked at the tree"

# Create an original label
original_label = "positive"

# Define all arguments for the function
ARGS = {
    "original_input": original_input,
    "original_label": original_label,
    "vocab": vocab,
    "perturbation_rate": 0.5,
    "num_perturbations": 10,
    "n_gram_length": 5
}
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
train_sklearn_ate_model Ran
ran on a synthesized input
causalate-mitigates-bias/causal-ate-mitigates-bias/utilities/train_ate.py
code served (permissive licence) · get_code("dd709e49ce602b82")
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 LinearRegression. 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
from sklearn.linear_model import LinearRegression

# Create a small sklearn model that has fit() and predict() methods
ate_model = LinearRegression()

# Create small training data
# 5 samples, 3 features
new_batch_inputs = [[1.0, 2.0, 3.0], 
                     [2.0, 3.0, 4.0],
                     [3.0, 4.0, 5.0],
                     [4.0, 5.0, 6.0],
                     [5.0, 6.0, 7.0]]

# 5 samples, 1 output (regression targets)
new_batch_outputs = [10.0, 15.0, 20.0, 25.0, 30.0]

# Number of epochs
num_epochs = 3

ARGS = {
    'ate_model': ate_model,
    'new_batch_inputs': new_batch_inputs,
    'new_batch_outputs': new_batch_outputs,
    'num_epochs': num_epochs
}
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 LinearRegression — a LinearRegression output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a LinearRegression 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
generate_perturbed_dataloader Not yet run
unverified (harvested, no recorded run)
causalate-mitigates-bias/causal-ate-mitigates-bias/archived/generate_ate_data_20240712_1316.py
code served (permissive licence) · get_code("55cca739ac4df98f")
How it ran

no invocation recorded.

Also present in other papers' code

Of 10 samples lifted from this paper's linked repositories, 10 carry a function identity and 0 do not. Of the 10 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 10 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.

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

We present the results and the main findings of SemEval-2019 Task 6 on Identifying and Categorizing Offensive Language in Social Media (OffensEval). The task was based on a new dataset, the Offensive Language Identification Dataset (OLID), which contains over 14,000 English tweets. It featured three sub-tasks. In sub-task A, the goal was to discriminate between offensive and non-offensive posts. In sub-task B, the focus was on the type of offensive content in the post. Finally, in sub-task C, systems had to detect the target of the offensive posts. OffensEval attracted a large number of participants and it was one of the most popular tasks in SemEval-2019. In total, about 800 teams signed up to participate in the task, and 115 of them submitted results, which we present and analyze in this report.

For agents

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

get_harvested_code_for_paper("1903.08983")
get_code_for_paper("1903.08983")
have("1903.08983")

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

curl https://syntology.ai/api/ran/1903.08983.json

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

[![Syntology run record](https://syntology.ai/api/ran/1903.08983.svg)](https://syntology.ai/paper/1903.08983)

Connect an agent — have() is free.