SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2006.02419 · 2020

Emergent Multi-Agent Communication in the Deep Learning Era

arXiv · PDF · Open in the Atlas

Code that ran

7 of 11 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; 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/2006.02419.json · badge: /api/ran/2006.02419.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
brendon-boldt/xferbench community (classifier label)
classifier basis not recorded on the link
7 ran · 4 no recorded run
FunctionStatusWhere it lives
get_analysis_path Ran
ran on a synthesized input
brendon-boldt/xferbench/xferbench/analysis.py
code served (permissive licence) · get_code("ecb43e6cee6af315")
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 PosixPath. 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 pathlib import Path
import tempfile

# Create a temporary directory for testing
temp_dir = tempfile.mkdtemp()
temp_path = Path(temp_dir)

ARGS = {
    'path': temp_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 PosixPath — a PosixPath output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a PosixPath 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
get_bpe_class Ran
ran on a synthesized input
brendon-boldt/xferbench/xferbench/model/config.py
code served (permissive licence) · get_code("c245ddbbb5623db1")
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 type. 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 tokenizers
from typing import Literal

BpeMethod = Literal['char', 'byte']

def get_bpe_class(method: BpeMethod) -> type[tokenizers.implementations.base_tokenizer.BaseTokenizer]:
    match method:
        case 'byte':
            return tokenizers.ByteLevelBPETokenizer
        case 'char':
            return tokenizers.CharBPETokenizer
        case _:
            from typing import assert_never
            assert_never(method)

# Test both valid method values
ARGS = {
    'method': 'byte'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
tokenizers — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type type — a type output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a type 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
get_checkpoint Ran
ran on a synthesized input
brendon-boldt/xferbench/xferbench/model/common.py
code served (permissive licence) · get_code("ba728bc56edbd416")
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 tempfile
import os
from pathlib import Path

# Create a temporary directory structure with checkpoint directories
temp_dir = tempfile.mkdtemp()

# Create some checkpoint directories
os.makedirs(os.path.join(temp_dir, "checkpoint-100"), exist_ok=True)
os.makedirs(os.path.join(temp_dir, "checkpoint-500"), exist_ok=True)
os.makedirs(os.path.join(temp_dir, "checkpoint-250"), exist_ok=True)

# The function expects model_source_base to be a path containing checkpoint-* directories
ARGS = {
    "model_source_base": temp_dir
}
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
get_clm_df Ran
ran on a synthesized input
brendon-boldt/xferbench/xferbench/analysis.py
code served (permissive licence) · get_code("6b329d2cbdb7977c")
How it ran

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

from pathlib import Path
import pandas as pd
import tempfile
import os

# Create a temporary directory structure with mock result files
temp_dir = tempfile.mkdtemp()
temp_path = Path(temp_dir)

# Create nested directory structure: source/target-variant/result.txt
source_dir = temp_path / "source1" / "target1-v1"
source_dir.mkdir(parents=True, exist_ok=True)
result_file = source_dir / "result.txt"
result_file.write_text("0.95")

source_dir2 = temp_path / "source2" / "target2-v2"
source_dir2.mkdir(parents=True, exist_ok=True)
result_file2 = source_dir2 / "result.txt"
result_file2.write_text("0.87")

ARGS = {
    'path': temp_path,
    'overwrite': False,
    'fail_no_cache': False,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_data_collator Ran
ran on a synthesized input
brendon-boldt/xferbench/xferbench/model/clm.py
code served (permissive licence) · get_code("6d31d0bd5eaad58d")
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 DataCollatorForLanguageModeling. 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 transformers
from transformers import PreTrainedTokenizer

torch.manual_seed(0)

class SimpleTokenizer:
    """Minimal tokenizer stand-in for DataCollatorForLanguageModeling"""
    def __init__(self):
        self.pad_token_id = 0
        self.eos_token_id = 2
        self.bos_token_id = 1
        self.unk_token_id = 3
        self.vocab_size = 100
        self.model_max_length = 512
    
    def __call__(self, text, **kwargs):
        # Return a dict with input_ids as expected by collator
        if isinstance(text, str):
            tokens = [1, 10, 11, 12, 2]  # Simple mock tokenization
        else:
            tokens = [1, 10, 11, 12, 2]
        return {"input_ids": tokens, "attention_mask": [1] * len(tokens)}
    
    def pad(self, encoded_inputs, **kwargs):
        # Mock padding
        return encoded_inputs

# Create a minimal tokenizer instance
tokenizer = SimpleTokenizer()

ARGS = {
    "tokenizer": tokenizer
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, transformers — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataCollatorForLanguageModeling — a DataCollatorForLanguageModeling output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataCollatorForLanguageModeling 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
get_encoder Ran
ran on a synthesized input
brendon-boldt/xferbench/xferbench/model/common.py
code served (permissive licence) · get_code("e286dbe12a8ba106")
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 numpy as np

torch.manual_seed(0)

class MockTokenizer:
    def __call__(self, text=None, text_target=None, return_special_tokens_mask=False, 
                 return_length=False, truncation=False, max_length=None, padding=False):
        if isinstance(text, dict):
            # Handle dict input with 'text' key
            text_list = text['text'] if isinstance(text.get('text'), list) else [text.get('text', '')]
        elif isinstance(text, list):
            text_list = text
        else:
            text_list = [text] if text else ['sample text']
        
        batch_size = len(text_list)
        
        # Return structure matching what the function expects
        result = {
            'input_ids': [
                [101, 2054, 2003, 102] * (i + 1) for i in range(batch_size)
            ]
        }
        
        if return_special_tokens_mask:
            result['special_tokens_mask'] = [
                [0, 0, 0, 1] * (i + 1) for i in range(batch_size)
            ]
        
        if return_length:
            result['length'] = [len(ids) for ids in result['input_ids']]
        
        return result

tokenizer = MockTokenizer()
block_size = 512

ARGS = {
    'tokenizer': tokenizer,
    'block_size': block_size
}
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 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
get_target_languages Ran
ran on a synthesized input
brendon-boldt/xferbench/xferbench/model/config.py
code served (permissive licence) · get_code("5c46ca498a81cb2c")
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 pydantic

class PydanticModel(pydantic.BaseModel):
    model_config = pydantic.ConfigDict(protected_namespaces=())

class RunConfig(PydanticModel):
    """Interface for executing an action.

    This is inteded to be useable from the command line via argparse,
    programmatically via Python, or from JSON config files.
    """
    command: str = 'default'
    source: str | None = None
    target: str | None = None
    overwrite: bool = False
    unit_test: bool = False
    save_prefix: str | None = None
    config: str | None = None
    cpu_ok: bool = False
    danish_only: bool = False
    extra_name: str | None = None
    result_path: str | None = None
    base_path: str | None = None

_target_languages = ['da', 'eu', 'fa', 'fi', 'he', 'id', 'ja', 'kk', 'ro', 'ur']

def get_target_languages(rc: RunConfig) -> list[str]:
    if rc.unit_test:
        return ['eval1', 'eval2']
    else:
        return _target_languages

# Setup code to construct valid arguments
rc_unit_test = RunConfig(
    command='test',
    source='src',
    target='tgt',
    overwrite=False,
    unit_test=True,
    save_prefix='test_prefix',
    config='config.json',
    cpu_ok=True,
    danish_only=False,
    extra_name='extra',
    result_path='/results',
    base_path='/base'
)

rc_normal = RunConfig(
    command='run',
    source='source_file',
    target='target_file',
    overwrite=True,
    unit_test=False,
    save_prefix='save_prefix',
    config=None,
    cpu_ok=False,
    danish_only=True,
    extra_name=None,
    result_path=None,
    base_path='/data'
)

ARGS = {
    'rc': rc_unit_test
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
pydantic — 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
check_complete Not yet run
unverified (harvested, no recorded run)
brendon-boldt/xferbench/xferbench/model/common.py
code served (permissive licence) · get_code("fc06548a004b1d3e")
How it ran

no invocation recorded.

get_tune_data Not yet run
unverified (harvested, no recorded run)
brendon-boldt/xferbench/xferbench/model/mlm.py
code served (permissive licence) · get_code("d065a6889de1c9fc")
How it ran

no invocation recorded.

get_tune_data Not yet run
unverified (harvested, no recorded run)
brendon-boldt/xferbench/xferbench/model/mt.py
code served (permissive licence) · get_code("b041e6488bd4a5ad")
How it ran

no invocation recorded.

load_tokenizer Not yet run
unverified (harvested, no recorded run)
brendon-boldt/xferbench/xferbench/model/mt.py
code served (permissive licence) · get_code("10a49786ffff4dd4")
How it ran

no invocation recorded.

Also present in other papers' code

Of 11 samples lifted from this paper's linked repositories, 11 carry a function identity and 0 do not. Of the 11 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 11 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

The ability to cooperate through language is a defining feature of humans. As the perceptual, motory and planning capabilities of deep artificial networks increase, researchers are studying whether they also can develop a shared language to interact. From a scientific perspective, understanding the conditions under which language evolves in communities of deep agents and its emergent features can shed light on human language evolution. From an applied perspective, endowing deep networks with the ability to solve problems interactively by communicating with each other and with us should make them more flexible and useful in everyday life. This article surveys representative recent language emergence studies from both of these two angles.

For agents

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

get_harvested_code_for_paper("2006.02419")
get_code_for_paper("2006.02419")
have("2006.02419")

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

curl https://syntology.ai/api/ran/2006.02419.json

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

[![Syntology run record](https://syntology.ai/api/ran/2006.02419.svg)](https://syntology.ai/paper/2006.02419)

Connect an agent — have() is free.