SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1901.08149 · 2019

TransferTransfo: A Transfer Learning Approach for Neural Network Based Conversational Agents

arXiv · PDF · Open in the Atlas

Code that ran

18 of 26 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; 1 honoured the contract we drafted, 0 violated it and 10 have no contract verdict; 7 failures of our instrument; 8 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/1901.08149.json · badge: /api/ran/1901.08149.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
huggingface/transfer-learning-conv-ai canonical (classifier label)
classifier basis not recorded on the link
1 ran · 0 no recorded run
dladustn95/ENMODEL reimplementation (classifier label)
classifier basis: classifier_2025
4 ran · 0 no recorded run
noriyukipy/gptchat community (classifier label)
classifier basis not recorded on the link
3 ran · 3 no recorded run
cerebroai/AskIt community (classifier label)
classifier basis not recorded on the link
3 ran · 3 no recorded run
thu-coai/CDial-GPT extension (classifier label)
classifier basis: classifier_2025
3 ran · 0 no recorded run
liliaellouz/wabi-sabi community (classifier label)
classifier basis not recorded on the link
3 ran · 0 no recorded run
dladustn95/enLanguageModel reimplementation (classifier label)
classifier basis: classifier_2025
1 ran · 0 no recorded run
copy not recorded community
not in the paper's declared repositories
0 ran · 1 no recorded run
KhueNguyen312/Persona-Chatbot extension (classifier label)
classifier basis: classifier_2025
0 ran · 1 no recorded run
FunctionStatusWhere it lives
average_distributed_scalar Ran
ran, honoured the contract we drafted
dladustn95/ENMODEL/bertTran_train.py
pointer only (licence: NONE) · get_code("ebb24f4dc2a81286")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, dladustn95/ENMODEL 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.

build_input_from_segments Ran
ran; our contract draft was wrong, not the code
dladustn95/ENMODEL/bertTran_train.py
pointer only (licence: NONE) · get_code("0192b257eca590dd")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, dladustn95/ENMODEL 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.

build_input_from_segments Ran
ran; our contract draft was wrong, not the code
thu-coai/CDial-GPT/infer.py
code served (permissive licence) · get_code("9883f7c396f997ab")
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) 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 numpy as np
from transformers import PreTrainedTokenizer

class MockTokenizer:
    """Minimal tokenizer mock that implements convert_tokens_to_ids"""
    def __init__(self):
        self.token_to_id = {
            '[CLS]': 101,
            '[SEP]': 102,
            '[PAD]': 0,
            '[speaker1]': 103,
            '[speaker2]': 104,
        }
    
    def convert_tokens_to_ids(self, tokens):
        return [self.token_to_id[token] for token in tokens]

# Create mock tokenizer
tokenizer = MockTokenizer()

# Create sample history and reply as lists of token IDs
history = [
    [1, 2, 3],      # First turn
    [4, 5, 6, 7],   # Second turn
]

reply = [8, 9, 10]

ARGS = {
    'history': history,
    'reply': reply,
    'tokenizer': tokenizer,
    'with_eos': True,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, torch, transformers — 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
clean_output Ran
ran on a synthesized input
noriyukipy/gptchat/gptchat/chatlm/lib.py
code served (permissive licence) · get_code("8a6666cf079c15e4")
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:
# - ids: needs to be a list-like object with .index() method
# - sep_token_id: an integer token ID to search for
# - cls_token_id: an integer token ID to search for

# The function searches for sep_token_id in ids, then searches for cls_token_id
# in the slice after sep_token_id. It uses .index() method which works on lists.

ARGS = {
    "ids": [101, 2054, 2003, 102, 1045, 2572],  # Simulating token IDs (CLS, ?, ?, SEP, I, love)
    "sep_token_id": 102,  # SEP token position at index 3
    "cls_token_id": 1045,  # CLS token (different ID) at index 4 in the cleaned portion
}
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
cross_entropy_loss_with_padding Ran
ran on a synthesized input
noriyukipy/gptchat/gptchat/gpt2.py
code served (permissive licence) · get_code("60687ac481590a5d")
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 function. 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 tensorflow as tf
import numpy as np

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

# Define parameters
num_labels = 5
pad_token_id = 0

# Create small sample data for testing
# y_true: batch of token IDs with shape (batch_size, sequence_length)
batch_size = 2
sequence_length = 3

y_true = tf.constant([
    [1, 2, 0],  # second example has padding at end
    [3, 4, 1]
], dtype=tf.int64)

# y_pred: logits with shape (batch_size, sequence_length, num_labels)
y_pred = tf.random.normal((batch_size, sequence_length, num_labels), dtype=tf.float32)

ARGS = {
    "num_labels": num_labels,
    "pad_token_id": pad_token_id
}
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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_prepare_inputs_for_generation Ran
ran on a synthesized input
noriyukipy/gptchat/gptchat/chatlm/lib.py
code served (permissive licence) · get_code("e97eae5b76709a23")
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 function. 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 tensorflow as tf

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

# Create a small stand-in class for self
class MockModel:
    pass

# Create the arguments
sep_token_id = 5

# Generate the prepare_inputs_for_generation function
def generate_prepare_inputs_for_generation(sep_token_id):
    def prepare_inputs_for_generation(self, inputs, **kwargs):
        batch_size, seq_len = inputs.shape
        sep_idx = tf.where(inputs[0] == sep_token_id).numpy().max()
        context_len = sep_idx
        response_len = seq_len - sep_idx
        token_type_ids = np.concatenate([np.zeros((batch_size, context_len)), np.ones((batch_size, response_len))], axis=1)
        attention_mask = np.concatenate([np.ones((batch_size, context_len)), np.ones((batch_size, response_len))], axis=1)
        return {'inputs': inputs, 'token_type_ids': tf.convert_to_tensor(token_type_ids, dtype=tf.int32), 'attention_mask': tf.convert_to_tensor(attention_mask, dtype=tf.int32)}
    return prepare_inputs_for_generation

# Test by creating sample inputs
# Create inputs tensor with sep_token_id at position 2 in first row
batch_size = 2
seq_len = 4
inputs = tf.constant([
    [1, 2, 5, 3],  # sep_token_id at index 2
    [1, 2, 5, 3]
], dtype=tf.int32)

ARGS = {
    'sep_token_id': sep_token_id
}
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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_response Ran
ran on a synthesized input
cerebroai/AskIt/decoder.py
code served (permissive licence) · get_code("cbad4369484f0733")
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. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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
import numpy as np
import torch
import torch.nn as nn
import torch.nn.functional as F
from configparser import ConfigParser
from io import StringIO

torch.manual_seed(0)

class StandInTokenizer:
    """Minimal tokenizer for testing"""
    def __init__(self):
        self.eos_token = "[EOS]"
    
    def encode(self, text):
        # Return a list of token ids
        return [1, 2, 3]
    
    def decode(self, token_ids, clean_up_tokenization_spaces=True):
        # Return a string representation
        return "generated text [EOS]"

class StandInModel(nn.Module):
    """Minimal language model for testing"""
    def __init__(self, vocab_size=100, hidden_size=4):
        super().__init__()
        self.embedding = nn.Embedding(vocab_size, hidden_size)
        self.linear = nn.Linear(hidden_size, vocab_size)
    
    def forward(self, input_ids):
        # input_ids shape: (batch_size, seq_len)
        embedded = self.embedding(input_ids)
        # Output logits shape: (batch_size, seq_len, vocab_size)
        logits = self.linear(embedded)
        return (logits,)

def create_config():
    """Create a minimal config object"""
    config_str = """
[model]
no_cuda = True

[decoder]
num_samples = 2
length = 3
temperature = 0.7
top_k = 5
top_p = 0.9
seed = 42
"""
    config = ConfigParser()
    config.read_string(config_str)
    return config

model = StandInModel(vocab_size=100, hidden_size=4)
tokenizer = StandInTokenizer()
context = "Hello world"
config = create_config()

ARGS = {
    "model": model,
    "tokenizer": tokenizer,
    "context": context,
    "config": config,
}
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
make_logdir Ran
ran on a synthesized input
huggingface/transfer-learning-conv-ai/utils.py
code served (permissive licence) · get_code("10f1aa81cc87b047")
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.

from datetime import datetime
import os
import socket

ARGS = {
    'model_name': 'gpt2'
}
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
pad_dataset Ran
ran; our contract draft was wrong, not the code
dladustn95/ENMODEL/bertTran_train.py
pointer only (licence: NONE) · get_code("87d44b2808b48aed")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, dladustn95/ENMODEL 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.

rmv_apostrophe Ran
ran on a synthesized input
liliaellouz/wabi-sabi/interact.py
code served (permissive licence) · get_code("755e75365f12f49e")
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 rmv_apostrophe(string):
    phrase = re.sub("won\\'t", 'will not', string)
    phrase = re.sub("can\\'t", 'can not', phrase)
    phrase = re.sub("\\'cause'", 'because', phrase)
    phrase = re.sub('let\\s', 'let us', phrase)
    phrase = re.sub("ma\\'am", 'madam', phrase)
    phrase = re.sub("y\\'all'", 'you all', phrase)
    phrase = re.sub("o\\'clock", 'of the clock', phrase)
    phrase = re.sub("n\\'t", ' not', phrase)
    phrase = re.sub("\\'re", ' are', phrase)
    phrase = re.sub("\\'s", ' is', phrase)
    phrase = re.sub("\\'d", ' would', phrase)
    phrase = re.sub("\\'ll", ' will', phrase)
    phrase = re.sub("\\'t", ' not', phrase)
    phrase = re.sub("\\'ve", ' have', phrase)
    phrase = re.sub("\\'m", ' am', phrase)
    phrase = re.sub("\\'", ' ', phrase)
    return phrase

ARGS = {
    "string": "won't can't let's ma'am y'all o'clock isn't that's I'd we'll haven't I'm it's"
}
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
sample_sequence Ran
ran on a synthesized input
cerebroai/AskIt/decoder.py
code served (permissive licence) · get_code("d769737d0223bb6e")
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 int, shape (2, 6) as a later run recorded it. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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
import torch.nn.functional as F
from configparser import ConfigParser
from io import StringIO

torch.manual_seed(0)

# Build a minimal model that returns logits in the expected shape
class MinimalLanguageModel(nn.Module):
    def __init__(self, vocab_size=8, hidden_dim=4):
        super().__init__()
        self.embedding = nn.Embedding(vocab_size, hidden_dim)
        self.linear = nn.Linear(hidden_dim, vocab_size)
    
    def forward(self, input_ids):
        # input_ids shape: (batch_size, seq_len)
        embeddings = self.embedding(input_ids)  # (batch_size, seq_len, hidden_dim)
        logits = self.linear(embeddings)  # (batch_size, seq_len, vocab_size)
        return (logits,)  # Return as tuple to match the unpacking in sample_sequence

# Build a minimal config object
config_str = """
[model]
no_cuda = True

[decoder]
num_samples = 2
length = 3
temperature = 0.7
top_k = 5
top_p = 0.9
"""

config = ConfigParser()
config.read_string(config_str)

# Create model
model = MinimalLanguageModel(vocab_size=8, hidden_dim=4)

# Create context (list of token indices)
context = [1, 2, 3]

ARGS = {
    'model': model,
    'context': context,
    'config': config
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family int, shape (2, 6) — shape (2, 6) 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
string_transformer Ran
ran on a synthesized input
liliaellouz/wabi-sabi/tokenize_2.py
code served (permissive licence) · get_code("325cbcf13635d567")
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 re

def rmv_apostrophe(string):
    phrase = re.sub("won\\'t", 'will not', string)
    phrase = re.sub("can\\'t", 'can not', phrase)
    phrase = re.sub("\\'cause'", 'because', phrase)
    phrase = re.sub('let\\s', 'let us', phrase)
    phrase = re.sub("ma\\'am", 'madam', phrase)
    phrase = re.sub("y\\'all'", 'you all', phrase)
    phrase = re.sub("o\\'clock", 'of the clock', phrase)
    phrase = re.sub("n\\'t", ' not', phrase)
    phrase = re.sub("\\'re", ' are', phrase)
    phrase = re.sub("\\'s", ' is', phrase)
    phrase = re.sub("\\'d", ' would', phrase)
    phrase = re.sub("\\'ll", ' will', phrase)
    phrase = re.sub("\\'t", ' not', phrase)
    phrase = re.sub("\\'ve", ' have', phrase)
    phrase = re.sub("\\'m", ' am', phrase)
    phrase = re.sub("\\'", ' ', phrase)
    return phrase

def string_transformer(string, tokenizer):
    phrase = string.lower()
    phrase = rmv_apostrophe(phrase)
    sentence_list = [sentence.strip() for sentence in phrase.strip().split('.')]
    sentence_list = [sentence for sentence in sentence_list if sentence != '']
    sentence_list = [sentence + ' .' for sentence in sentence_list]
    tokenized = [tokenizer.encode(sentence) for sentence in sentence_list]
    return [item for sublist in tokenized for item in sublist]


class SimpleTokenizer:
    def __init__(self):
        self.vocab = {
            'hello': 1,
            'world': 2,
            'this': 3,
            'is': 4,
            'a': 5,
            'test': 6,
            '.': 7,
            'won': 8,
            't': 9,
            'will': 10,
            'not': 11,
            'can': 12,
            'do': 13,
        }
    
    def encode(self, sentence):
        tokens = []
        words = sentence.lower().split()
        for word in words:
            tokens.append(self.vocab.get(word, 0))
        return tokens


test_string = "Hello world. This is a test. Won't you help."
test_tokenizer = SimpleTokenizer()

ARGS = {
    "string": test_string,
    "tokenizer": test_tokenizer,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family 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
test_data Ran
ran; our contract draft was wrong, not the code
thu-coai/CDial-GPT/infer.py
code served (permissive licence) · get_code("692fe8bc3c032edb")
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) 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 json
import tempfile
import os

# Create a temporary JSON file with test data
temp_dir = tempfile.mkdtemp()
test_file = os.path.join(temp_dir, "test_data.json")

# Write test data as a dict with 'test' key
test_data_content = {
    "test": [
        {"id": 1, "text": "sample text 1"},
        {"id": 2, "text": "sample text 2"}
    ]
}

with open(test_file, 'w', encoding='utf-8') as f:
    json.dump(test_data_content, f)

# Create a simple args object
class Args:
    def __init__(self, datapath):
        self.datapath = datapath

ARGS = {
    "args": Args(test_file)
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
top_filtering Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
dladustn95/enLanguageModel/interact.py
pointer only (licence: NONE) · get_code("a07f6a302fe3b9eb")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, dladustn95/enLanguageModel 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.

top_filtering Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
dladustn95/ENMODEL/bertTran_interact.py
pointer only (licence: NONE) · get_code("7996c90fe508c5c8")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, dladustn95/ENMODEL 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.

top_filtering Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
thu-coai/CDial-GPT/infer.py
code served (permissive licence) · get_code("695eaa24be6f2a1d")
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) and returned Tensor, ndim 1, dtype family float, shape (10,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn.functional as F

torch.manual_seed(0)

# Create a small logits tensor (vocabulary size of 10)
logits = torch.randn(10)

# Define arguments for the top_filtering function
ARGS = {
    'logits': logits,
    'top_k': 5,
    'top_p': 0.9,
    'threshold': -2.0,
    'filter_value': -float('Inf')
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 1, dtype family float, shape (10,) — shape (10,) 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
top_k_top_p_filtering Ran
ran on a synthesized input
cerebroai/AskIt/decoder.py
code served (permissive licence) · get_code("31631ed21a9dd3b3")
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.functional as F

torch.manual_seed(0)

# Create a small batch of logits with shape (batch_size=2, vocab_size=4)
logits = torch.randn(2, 4)

ARGS = {
    "logits": logits,
    "top_k": 2,
    "top_p": 0.9,
    "filter_value": -float('Inf')
}
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
translate Ran
ran on a synthesized input
liliaellouz/wabi-sabi/integrate_summary.py
code served (permissive licence) · get_code("b4f520ecd9ceaa5e")
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 torch

ARGS = {
    'word': 'she'
}
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 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
cosine_similarity Not yet run
unverified (harvested, no recorded run)
cerebroai/AskIt/run_pipeline.py
code served (permissive licence) · get_code("c423433d8bc52711")
How it ran

no invocation recorded.

download_model_folder Not yet run
unverified (harvested, no recorded run)
cerebroai/AskIt/model.py
code served (permissive licence) · get_code("41b82c54e1b15d6d")
How it ran

no invocation recorded.

evaluation Not yet run
unverified (harvested, no recorded run)
this paper's copy was not recorded; identical code first harvested from samsonleegh/convai_smile
pointer only; the stored copy get_code serves, from samsonleegh/convai_smile, is pointer only (licence: NONE) · get_code("e46d6b86df789052")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. 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.

generate Not yet run
unverified (harvested, no recorded run)
noriyukipy/gptchat/gptchat/chatlm/lib.py
code served (permissive licence) · get_code("1d561f2cbca372e3")
How it ran

no invocation recorded.

init_model Not yet run
unverified (harvested, no recorded run)
noriyukipy/gptchat/gptchat/gpt2.py
code served (permissive licence) · get_code("d0fd73ea11340a75")
How it ran

no invocation recorded.

load_model Not yet run
unverified (harvested, no recorded run)
cerebroai/AskIt/model.py
code served (permissive licence) · get_code("2f849c7661306814")
How it ran

no invocation recorded.

load_or_init_model Not yet run
unverified (harvested, no recorded run)
noriyukipy/gptchat/gptchat/gpt2.py
code served (permissive licence) · get_code("88a39d354890aa74")
How it ran

no invocation recorded.

top_filtering Not yet run
unverified (harvested, no recorded run)
KhueNguyen312/Persona-Chatbot/Backend/model/interact.py
pointer only (licence: NONE) · get_code("cc247ba2437e3e84")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, KhueNguyen312/Persona-Chatbot 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.

Also present in other papers' code

Of 26 samples lifted from this paper's linked repositories, 26 carry a function identity and 0 do not. Of the 25 distinct functions with an identity, 5 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 4 later) and 20 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
top_filtering 7 other papers
7 later
1904.09751 (2019-04)
THE CURIOUS CASE OF NEURAL TEXT DeGENERATION
1911.00536 (2019-11)
DialoGPT: Large-Scale Generative Pre-training for Conversational Response Generation
2002.00163 (2020-02)
Bridging Text and Video: A Universal Multimodal Transformer for Video-Audio Scene-Aware Dialog
and 4 more not named here
average_distributed_scalar 3 other papers
3 later
2008.09075 (2020-08)
Controlling Dialogue Generation with Semantic Exemplars
2009.12005 (2020-09)
MinTL: Minimalist Transfer Learning for Task-Oriented Dialogue Systems
2310.15896 (2023-10)
BianQue: Balancing the Questioning and Suggestion Ability of Health LLMs with Multi-turn Health Conversations Polished by ChatGPT
top_filtering 3 other papers
3 later
2004.12752 (2020-04)
The Gutenberg Dialogue Dataset
2008.09075 (2020-08)
Controlling Dialogue Generation with Semantic Exemplars
2308.09033 (2023-08)
Uni-NLX: Unifying Textual Explanations for Vision and Vision-Language Tasks
evaluation 1 other paper
1 dated earlier
1801.07243 (2018-01)
Personalizing Dialogue Agents: I have a dog, do you have pets too?
make_logdir 1 other paper
1 later
2004.12752 (2020-04)
The Gutenberg Dialogue Dataset

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 introduce a new approach to generative data-driven dialogue systems (e.g. chatbots) called TransferTransfo which is a combination of a Transfer learning based training scheme and a high-capacity Transformer model. Fine-tuning is performed by using a multi-task objective which combines several unsupervised prediction tasks. The resulting fine-tuned model shows strong improvements over the current state-of-the-art end-to-end conversational models like memory augmented seq2seq and information-retrieval models. On the privately held PERSONA-CHAT dataset of the Conversational Intelligence Challenge 2, this approach obtains a new state-of-the-art, with respective perplexity, Hits@1 and F1 metrics of 16.28 (45 % absolute improvement), 80.7 (46 % absolute improvement) and 19.5 (20 % absolute improvement).

For agents

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

get_harvested_code_for_paper("1901.08149")
get_code_for_paper("1901.08149")
have("1901.08149")

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

curl https://syntology.ai/api/ran/1901.08149.json

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

[![Syntology run record](https://syntology.ai/api/ran/1901.08149.svg)](https://syntology.ai/paper/1901.08149)

Connect an agent — have() is free.