SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2005.00547 · ACL · 2020

GoEmotions: A Dataset of Fine-Grained Emotions

Alan Cowen, Dorottya Demszky, Sujith Ravi, Dana Movshovitz-Attias, Jeongwoo Ko, Gaurav Nemade

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 7 have no contract verdict; 2 failures of our instrument; 1 no recorded run), measured 2026-11-21; 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/2005.00547.json · badge: /api/ran/2005.00547.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
hlt-maia/emotion-transformer community (classifier label)
classifier basis not recorded on the link
4 ran · 0 no recorded run
nur-ag/emotion-classification community (classifier label)
classifier basis not recorded on the link
3 ran · 0 no recorded run
monologg/GoEmotions-pytorch community (classifier label)
classifier basis not recorded on the link
2 ran · 1 no recorded run
FunctionStatusWhere it lives
apply_to_sample Ran
ran on a synthesized input
hlt-maia/emotion-transformer/utils.py
code served (permissive licence) · get_code("da253f3773e31f0e")
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

torch.manual_seed(0)

# Define a simple function to apply to tensors
def simple_transform(x):
    """A simple transformation function that adds 1 to tensors"""
    return x + 1

# Create a sample with various nested structures
sample = {
    "tensor_field": torch.randn(2, 4),
    "nested_dict": {
        "inner_tensor": torch.randn(2, 3),
        "value": 42
    },
    "list_field": [
        torch.randn(2, 5),
        {"deep_tensor": torch.randn(2, 6)},
        "string_value"
    ],
    "scalar": 100
}

ARGS = {
    "f": simple_transform,
    "sample": sample
}
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 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
average_pooling Ran
ran; our contract draft was wrong, not the code
hlt-maia/emotion-transformer/model/utils.py
code served (permissive licence) · get_code("cdcab03abda9e8db")
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)

# Construct valid arguments for average_pooling
batch_size = 2
seq_length = 3
hidden_size = 4
padding_index = 0

# tokens: [batch_size x seq_length] with some padding tokens
tokens = torch.tensor([
    [1, 2, padding_index],
    [1, padding_index, padding_index]
], dtype=torch.long)

# embeddings: [batch_size x seq_length x hidden_size]
embeddings = torch.randn(batch_size, seq_length, hidden_size, dtype=torch.float32)

# mask: [batch_size x seq_length] - 1 for real tokens, 0 for padding
mask = torch.tensor([
    [1, 1, 0],
    [1, 0, 0]
], dtype=torch.long)

ARGS = {
    "tokens": tokens,
    "embeddings": embeddings,
    "mask": mask,
    "padding_index": padding_index
}
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
compute_metrics Ran
ran on a synthesized input
monologg/GoEmotions-pytorch/utils.py
code served (permissive licence) · get_code("9e482b3b0bc433ce")
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
from sklearn.metrics import precision_recall_fscore_support, accuracy_score

# Create small sample data for labels and predictions
# Both should be 1D arrays of class labels
np.random.seed(0)

# Create 10 samples with 3 possible classes (0, 1, 2)
labels = np.array([0, 1, 2, 0, 1, 2, 0, 1, 2, 0])
preds = np.array([0, 1, 2, 0, 1, 1, 0, 2, 2, 0])

ARGS = {
    'labels': labels,
    'preds': preds
}
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 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_examples_to_features Ran
ran on a synthesized input
monologg/GoEmotions-pytorch/data_loader.py
code served (permissive licence) · get_code("07f7c0967760a3c2")
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
import torch
import numpy as np
from unittest.mock import Mock

torch.manual_seed(0)

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

# Create label file
label_file_path = os.path.join(temp_dir, "labels.txt")
with open(label_file_path, 'w', encoding='utf-8') as f:
    f.write("admiration\n")
    f.write("amusement\n")
    f.write("anger\n")
    f.write("annoyance\n")

# Create train file
train_file_path = os.path.join(temp_dir, "train.txt")
with open(train_file_path, 'w', encoding='utf-8') as f:
    f.write("This is a great example\t0,1\n")
    f.write("I am very angry\t2,3\n")

# Create a minimal tokenizer mock
class MinimalTokenizer:
    def batch_encode_plus(self, text_pairs, max_length=None, pad_to_max_length=False):
        # Return minimal batch encoding with required keys
        batch_size = len(text_pairs)
        return {
            'input_ids': [[101] + [1000 + i for _ in range(max_length - 2)] + [102] for i in range(batch_size)],
            'attention_mask': [[1] * max_length for _ in range(batch_size)],
            'token_type_ids': [[0] * max_length for _ in range(batch_size)],
        }
    
    def tokenize(self, text):
        return text.split()

tokenizer = MinimalTokenizer()

# Create args object with required attributes
class Args:
    def __init__(self):
        self.data_dir = temp_dir
        self.label_file = "labels.txt"
        self.train_file = "train.txt"
        self.dev_file = "train.txt"
        self.test_file = "train.txt"

args = Args()

# Create minimal examples
examples = [
    Mock(guid='train-0', text_a="This is a great example", text_b=None, label=[0, 1]),
    Mock(guid='train-1', text_a="I am very angry", text_b=None, label=[2, 3]),
]

max_length = 8

ARGS = {
    'args': args,
    'examples': examples,
    'tokenizer': tokenizer,
    'max_length': max_length,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
loss_factory Ran
ran on a synthesized input
nur-ag/emotion-classification/src/models/neural.py
code served (permissive licence) · get_code("d0d218c6bdc76cdc")
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 CrossEntropyLoss. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# The function takes a single parameter: problem_type
# It returns different loss functions based on the string value
# We'll test with 'multiclass' as a valid problem type

ARGS = {
    'problem_type': 'multiclass'
}
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 CrossEntropyLoss — a CrossEntropyLoss output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a CrossEntropyLoss 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
mask_fill Ran
ran on a synthesized input
hlt-maia/emotion-transformer/model/utils.py
code served (permissive licence) · get_code("899bda55086c1bdb")
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)

# Small batch size (2), sequence length (3), hidden dimension (4)
bsz = 2
seq_len = 3
hidden_dim = 4
padding_index = 0

# Create token sequences with some padding tokens
# tokens shape: [bsz x seq_len]
tokens = torch.tensor([
    [1, 2, 0],  # sequence with padding at end
    [3, 0, 0]   # sequence with padding at positions 1, 2
], dtype=torch.long)

# Create embeddings
# embeddings shape: [bsz x seq_len x hidden_dim]
embeddings = torch.randn(bsz, seq_len, hidden_dim, dtype=torch.float32)

# Fill value for masked positions
fill_value = 0.0

ARGS = {
    "fill_value": fill_value,
    "tokens": tokens,
    "embeddings": embeddings,
    "padding_index": padding_index
}
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
max_pooling Ran
ran; our contract draft was wrong, not the code
hlt-maia/emotion-transformer/model/utils.py
code served (permissive licence) · get_code("f01a46c098a727cf")
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)

# Define tensors for max_pooling
batch_size = 2
seq_length = 3
hidden_size = 4
padding_index = 0

# tokens: [batch_size x seq_length] containing token ids
# Some tokens are padding (value 0), others are non-padding
tokens = torch.tensor([
    [1, 2, 0],      # sequence with padding at end
    [3, 0, 0]       # sequence with padding
], dtype=torch.long)

# embeddings: [batch_size x seq_length x hidden_size]
embeddings = torch.randn(batch_size, seq_length, hidden_size, dtype=torch.float32)

ARGS = {
    "tokens": tokens,
    "embeddings": embeddings,
    "padding_index": padding_index
}
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
nop Ran
ran on a synthesized input
nur-ag/emotion-classification/src/serve.py
code served (permissive licence) · get_code("cbcad0088e9c4680")
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
import torch.nn as nn

torch.manual_seed(0)

ARGS = {
    "it": torch.randn(2, 4),
}
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
reformat_tensor Ran
ran on a synthesized input
nur-ag/emotion-classification/src/models/partial.py
code served (permissive licence) · get_code("e28a925a19f55cde")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 2, dtype family float, shape (2, 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
import numpy as np

torch.manual_seed(0)

# The function takes a tensor parameter and converts it to a numpy array,
# then reshapes it to (batch_size, -1).
# We'll test with a torch tensor that can be converted to numpy.

# Create a small tensor with batch_size=2 and some features
tensor_input = torch.randn(2, 4, 3)  # batch_size=2, will reshape to (2, 12)

ARGS = {
    "tensor": tensor_input
}
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 ndarray, ndim 2, dtype family float, shape (2, 12) — shape (2, 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
load_and_cache_examples Not yet run
unverified (harvested, no recorded run)
monologg/GoEmotions-pytorch/data_loader.py
code served (permissive licence) · get_code("5b789dfece5f5d6f")
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, 2 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 2 later) and 8 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
max_pooling 5 other papers
5 later
2109.06352 (2021-09)
Uncertainty-Aware Machine Translation Evaluation
2023.acl-long.47 (2023 (year only))
Test-time Adaptation for Machine Translation Evaluation by Uncertainty Minimization
2402.18091 (2024-02)
Polos: Multimodal Metric Learning from Human Feedback for Image Captioning
and 2 more not named here
average_pooling 4 other papers
4 later
2109.06352 (2021-09)
Uncertainty-Aware Machine Translation Evaluation
2023.acl-long.47 (2023 (year only))
Test-time Adaptation for Machine Translation Evaluation by Uncertainty Minimization
2402.18091 (2024-02)
Polos: Multimodal Metric Learning from Human Feedback for Image Captioning
and 1 more not named here

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 2026-11-21.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Understanding emotion expressed in language has a wide range of applications, from building empathetic chatbots to detecting harmful online behavior. Advancement in this area can be improved using large-scale datasets with a fine-grained typology, adaptable to multiple downstream tasks. We introduce GoEmotions, the largest manually annotated dataset of 58k English Reddit comments, labeled for 27 emotion categories or Neutral. We demonstrate the high quality of the annotations via Principal Preserved Component Analysis. We conduct transfer learning experiments with existing emotion benchmarks to show that our dataset generalizes well to other domains and different emotion taxonomies. Our BERTbased model achieves an average F1-score of .46 across our proposed taxonomy, leaving much room for improvement. 1

For agents

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

get_harvested_code_for_paper("2005.00547")
get_code_for_paper("2005.00547")
have("2005.00547")

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

curl https://syntology.ai/api/ran/2005.00547.json

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

[![Syntology run record](https://syntology.ai/api/ran/2005.00547.svg)](https://syntology.ai/paper/2005.00547)

Connect an agent — have() is free.