SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1901.08706 · 2019

Emergent Linguistic Phenomena in Multi-Agent Communication Games

arXiv · PDF · Open in the Atlas

Code that ran

16 of 20 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 16 have no contract verdict; 0 failures of our instrument; 4 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/1901.08706.json · badge: /api/ran/1901.08706.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
lgraesser/MultimodalGame canonical (classifier label)
classifier basis: census
16 ran · 4 no recorded run
FunctionStatusWhere it lives
add_elem Ran
ran on a synthesized input
lgraesser/MultimodalGame/analyze_messages.py
code served (permissive licence) · get_code("2a4d0f5a6e0b0f11")
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 numpy as np

# Construct a valid m_dict starting structure
m_dict = {
    'total': 0,
    'correct': 0,
    'incorrect': 0,
}

# Define all required arguments
message = "test_message"
shape = "circle"
color = "red"
answer_type = "correct"

ARGS = {
    'm_dict': m_dict,
    'message': message,
    'shape': shape,
    'color': color,
    'answer_type': answer_type,
}
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 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
basic_block Ran
ran on a synthesized input
lgraesser/MultimodalGame/utils/package_data.py
code served (permissive licence) · get_code("0e4b65c138fa4aa5")
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 torch
import torch.nn as nn

torch.manual_seed(0)

class BasicBlockLayer(nn.Module):
    def __init__(self):
        super().__init__()
        self.conv1 = nn.Conv2d(3, 4, kernel_size=3, padding=1)
        self.bn1 = nn.BatchNorm2d(4)
        self.relu = nn.ReLU(inplace=True)
        self.conv2 = nn.Conv2d(4, 4, kernel_size=3, padding=1)
        self.bn2 = nn.BatchNorm2d(4)
        self.downsample = None
    
    def forward(self, x):
        residual = x
        out = self.conv1(x)
        out = self.bn1(out)
        out = self.relu(out)
        out = self.conv2(out)
        out = self.bn2(out)
        if self.downsample is not None:
            residual = self.downsample(x)
        out += residual
        return out

layer = BasicBlockLayer()

ARGS = {
    'layer': layer,
    'relu': False
}
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 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
build_eval_list Ran
ran on a synthesized input
lgraesser/MultimodalGame/community_util.py
code served (permissive licence) · get_code("727f055137f254ec")
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. 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 logging

# Set up logging
logging.basicConfig(level=logging.DEBUG)

# Create a simple logger
debuglogger = logging.getLogger('main_logger')

# Define ARGS dictionary with all required parameters
# pools_num: list of integers representing number of agents in each pool
# community_type: string indicating the type of community ('chain' or other)
# train_vec_prob: 1D array that will be reshaped to (total_agents, total_agents)

pools_num = [3, 2, 4]  # 3 pools with 3, 2, and 4 agents respectively
total_agents = sum(pools_num)  # 3 + 2 + 4 = 9 agents

# Create a training matrix as a flattened vector
# This represents which agent pairs have trained together
# Shape will be reshaped to (9, 9) = 81 elements
train_vec_prob = np.random.rand(total_agents * total_agents)

community_type = 'chain'

ARGS = {
    'pools_num': pools_num,
    'community_type': community_type,
    'train_vec_prob': train_vec_prob
}
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 — a list 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
build_train_matrix Ran
ran on a synthesized input
lgraesser/MultimodalGame/community_util.py
code served (permissive licence) · get_code("a6308dec32de38bc")
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 numpy as np
import torch
import logging

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

# Set up logging to avoid issues
logging.basicConfig(level=logging.INFO)

ARGS = {
    'pools_num': [3, 4, 2],
    'community_type': 'dense',
    'intra_pool_connect_p': [0.5, 0.6, 0.4],
    'inter_pool_connect_p': 0.3,
    'adjust_train_ratio': True,
    'intra_inter_ratio': 1.0
}
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 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_tensor_to_string Ran
ran on a synthesized input
lgraesser/MultimodalGame/analyze_messages.py
code served (permissive licence) · get_code("f41a389cd0a31a31")
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

torch.manual_seed(0)

# Create a 1D binary tensor with values 0 and 1
message = torch.tensor([0, 1, 0, 1, 1, 0])

ARGS = {
    'message': message
}
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
custom_dtype Ran
ran on a synthesized input
lgraesser/MultimodalGame/utils/package_data.py
code served (permissive licence) · get_code("65343d95239ca6e7")
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 VoidDType. 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 torch

# Analyze the function:
# - outp: iterable that gets zipped with request, each element o has .shape attribute
# - request: iterable that gets zipped with outp, each element r is used as a tuple key
# The function iterates through pairs and builds a schema with tuples containing:
#   - r (string key from request)
#   - np.float32 (dtype)
#   - size (tuple constructed from o.shape)

# Create tensors with distinct shapes for outp
torch.manual_seed(0)
outp = [
    torch.randn(2, 4),      # shape (2, 4)
    torch.randn(2, 3, 5),   # shape (2, 3, 5)
]

# Create request strings
request = ['feature_a', 'feature_b']

ARGS = {
    'outp': outp,
    'request': request,
}
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 VoidDType — a VoidDType output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a VoidDType 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
downsize Ran
ran on a synthesized input
lgraesser/MultimodalGame/dataset_loader.py
code served (permissive licence) · get_code("f9a910c22ed228de")
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 4, dtype family float, shape (2, 4, 4, 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 numpy as np
from skimage.transform import resize

# Create sample image data
# Shape: (batch_size, width, height, channels)
batch_size = 2
width = 8
height = 6
channels = 3

# Create random image data in valid range [0, 1] for skimage
ims = np.random.rand(batch_size, width, height, channels).astype(np.float32)

# Target size for downsampling
size = 4

ARGS = {
    "ims": ims,
    "size": size,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, skimage — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 4, dtype family float, shape (2, 4, 4, 3) — shape (2, 4, 4, 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
label_mapping Ran
ran on a synthesized input
lgraesser/MultimodalGame/utils/package_data.py
code served (permissive licence) · get_code("a3fa3e5837b92279")
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 tempfile
import os

# Create a temporary file with label mappings
temp_dir = tempfile.mkdtemp()
desc_path = os.path.join(temp_dir, "labels.txt")

# Write label data in the expected format: label_id,label,desc
with open(desc_path, 'w') as f:
    f.write("0,cat,a feline animal\n")
    f.write("1,dog,a canine animal\n")
    f.write("2,bird,a flying animal\n")

ARGS = {
    'desc_path': desc_path
}
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 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
loglikelihood Ran
ran on a synthesized input
lgraesser/MultimodalGame/model_symmetric.py
code served (permissive licence) · get_code("f8890e03987bcbdf")
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, 1) 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 log_prob tensor: (N=2, C=4) - batch size 2, 4 classes
log_prob = torch.randn(2, 4).log_softmax(dim=1)

# Create target tensor: (N=2,) - batch size 2, each element is a class index in [0, 3]
target = torch.tensor([1, 3]).unsqueeze(1)

ARGS = {
    "log_prob": log_prob,
    "target": target,
}
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, 1) — shape (2, 1) 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
read_log_load Ran
ran on a synthesized input
lgraesser/MultimodalGame/misc.py
code served (permissive licence) · get_code("e64293b23b188b52")
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 json
import tempfile
import os

# Create a temporary file with log content containing JSON
temp_file = tempfile.NamedTemporaryFile(mode='w', delete=False, suffix='.log')
temp_filename = temp_file.name

# Write log content with Flag Values section
log_content = """Some log lines
Before the flag values
Flag Values
{
  "param1": "value1",
  "param2": 42
}
More lines after
Flag Values
{
  "param3": "value3",
  "param4": 100
}
End of log
"""

temp_file.write(log_content)
temp_file.close()

ARGS = {
    'filename': temp_filename,
    'last': True
}
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 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
recursively_set_device Ran
ran on a synthesized input
lgraesser/MultimodalGame/misc.py
code served (permissive licence) · get_code("4cf10e714cf8e774")
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. 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 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)

# Create a simple nn.Module with a parameter that has .cpu() and .cuda() methods
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 3)
    
    def forward(self, x):
        return self.linear(x)

# Create test inputs
model = SimpleModel()
tensor = torch.randn(2, 4)
nested_dict = {
    'model': model,
    'tensor': tensor,
    'nested': {
        'another_tensor': torch.randn(2, 3)
    }
}
list_input = [tensor, model, torch.randn(2, 4)]
tuple_input = (tensor, torch.randn(2, 3))

ARGS = {
    'inp': nested_dict,
    'gpu': -1
}
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
sample_agents Ran
ran on a synthesized input
lgraesser/MultimodalGame/community_util.py
code served (permissive licence) · get_code("52e46b251663ef26")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family 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 numpy as np

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

# Create a probability distribution over agent pairs
# This should sum to 1.0
train_probs = np.array([0.2, 0.3, 0.25, 0.25])

# Create a mapping from flattened index to agent pairs
# Each element is a tuple of (agent1_idx, agent2_idx)
agent_idx_list = [
    (0, 1),
    (0, 2),
    (1, 2),
    (1, 3)
]

ARGS = {
    "train_probs": train_probs,
    "agent_idx_list": agent_idx_list
}
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 tuple, ndim 1, dtype family int — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
sparks Ran
ran on a synthesized input
lgraesser/MultimodalGame/old_files/sparks.py
code served (permissive licence) · get_code("10d24b28d4ca56b5")
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
import numpy as np

parts = u' ▁▂▃▄▅▆▇▉'

def sparks(nums):
    fraction = max(nums) / float(len(parts) - 1)
    return ''.join((parts[int(round(x / fraction))] for x in nums))

# Create test data - nums should be a list/array of numbers
nums = [0, 2.5, 5.0, 7.5, 10.0]

ARGS = {
    'nums': nums
}
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 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
torch_load Ran
ran on a synthesized input
lgraesser/MultimodalGame/misc.py
code served (permissive licence) · get_code("ceb3746f90391b0e")
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
import tempfile
import os

torch.manual_seed(0)

# Create a simple model
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 3)
    
    def forward(self, x):
        return self.linear(x)

# Create instances
model1 = SimpleModel()
model2 = SimpleModel()

# Create optimizers
optimizer1 = torch.optim.SGD(model1.parameters(), lr=0.01)
optimizer2 = torch.optim.Adam(model2.parameters(), lr=0.001)

# Create a checkpoint file
checkpoint_data = {
    'models': {
        'model1': model1.state_dict(),
        'model2': model2.state_dict(),
    },
    'optimizers': {
        'optimizer1': optimizer1.state_dict(),
        'optimizer2': optimizer2.state_dict(),
    },
    'data': {'epoch': 5, 'loss': 0.123}
}

# Save checkpoint to a temporary file
with tempfile.NamedTemporaryFile(mode='w', delete=False, suffix='.pt') as f:
    temp_filename = f.name

torch.save(checkpoint_data, temp_filename)

# Create new model and optimizer instances for loading
loaded_model1 = SimpleModel()
loaded_model2 = SimpleModel()
loaded_optimizer1 = torch.optim.SGD(loaded_model1.parameters(), lr=0.01)
loaded_optimizer2 = torch.optim.Adam(loaded_model2.parameters(), lr=0.001)

ARGS = {
    'filename': temp_filename,
    'models_dict': {
        'model1': loaded_model1,
        'model2': loaded_model2,
    },
    'optimizers_dict': {
        'optimizer1': loaded_optimizer1,
        'optimizer2': loaded_optimizer2,
    }
}
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
try_mkdir Ran
ran on a synthesized input
lgraesser/MultimodalGame/utils/download_data.py
code served (permissive licence) · get_code("5aab172cceed1b91")
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 int, ndim 0, dtype family int, 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 os
import tempfile

# Create a temporary directory for testing
temp_dir = tempfile.gettempdir()
test_path = os.path.join(temp_dir, "test_mkdir_" + str(os.getpid()))

# Clean up if it exists from a previous run
if os.path.exists(test_path):
    os.rmdir(test_path)

ARGS = {
    "path": test_path
}
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 int, ndim 0, dtype family int, 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
upscale Ran
ran on a synthesized input
lgraesser/MultimodalGame/dataset_loader.py
code served (permissive licence) · get_code("445f43025741f25e")
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 4, dtype family float, shape (2, 227, 227, 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 numpy as np

# Create a small batch of dummy images
# Shape: (batch_size, width, height, channels)
# Using small but distinct dimensions to probe shape handling
batch_size = 2
width = 64
height = 48
channels = 3

# Create dummy image data with random values in [0, 1] range
ims = np.random.rand(batch_size, width, height, channels).astype(np.float32)

ARGS = {
    "ims": ims
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, skimage — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 4, dtype family float, shape (2, 227, 227, 3) — shape (2, 227, 227, 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
calc_message_mean_and_std Not yet run
unverified (harvested, no recorded run)
lgraesser/MultimodalGame/model_symmetric.py
code served (permissive licence) · get_code("f653105ff7ba6ade")
How it ran

no invocation recorded.

calc_ratios Not yet run
unverified (harvested, no recorded run)
lgraesser/MultimodalGame/analyze_messages.py
code served (permissive licence) · get_code("1399d03e6a433f97")
How it ran

no invocation recorded.

clean_and_tokenize Not yet run
unverified (harvested, no recorded run)
lgraesser/MultimodalGame/dataset_loader.py
code served (permissive licence) · get_code("0b5e5ad6c0494100")
How it ran

no invocation recorded.

flickr_name Not yet run
unverified (harvested, no recorded run)
lgraesser/MultimodalGame/utils/download_data.py
code served (permissive licence) · get_code("400e6ee276cd2cc3")
How it ran

no invocation recorded.

Also present in other papers' code

Of 20 samples lifted from this paper's linked repositories, 20 carry a function identity and 0 do not. Of the 20 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 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.

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

In this work, we propose a computational framework in which agents equipped with communication capabilities simultaneously play a series of referential games, where agents are trained using deep reinforcement learning. We demonstrate that the framework mirrors linguistic phenomena observed in natural language: i) the outcome of contact between communities is a function of inter- and intra-group connectivity; ii) linguistic contact either converges to the majority protocol, or in balanced cases leads to novel creole languages of lower complexity; and iii) a linguistic continuum emerges where neighboring languages are more mutually intelligible than farther removed languages. We conclude that intricate properties of language evolution need not depend on complex evolved linguistic capabilities, but can emerge from simple social exchanges between perceptually-enabled agents playing communication games.

For agents

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

get_harvested_code_for_paper("1901.08706")
get_code_for_paper("1901.08706")
have("1901.08706")

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

curl https://syntology.ai/api/ran/1901.08706.json

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

[![Syntology run record](https://syntology.ai/api/ran/1901.08706.svg)](https://syntology.ai/paper/1901.08706)

Connect an agent — have() is free.