SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2502.20073 · EMNLP · 2025

Collab-Overcooked: Benchmarking and Evaluating Large Language Models as Collaborative Agents

Hao Xu, Hao Fu, Caixia Yuan, Xiaojie Wang, Haochen Sun, Lei Ren, Shuwen Zhang

arXiv · PDF · Open in the Atlas

Code that ran

6 of 7 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 3 have no contract verdict; 3 failures of our instrument; 1 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2502.20073.json · badge: /api/ran/2502.20073.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
yusaemeow/collab-overcooked canonical (classifier label)
classifier basis not recorded on the link
6 ran · 1 no recorded run
FunctionStatusWhere it lives
check_port_in_use Ran
ran on a synthesized input
yusaemeow/collab-overcooked/src/collab/web_util.py
code served (permissive licence) · get_code("11c72b2a5ed15b25")
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 bool, ndim 0, dtype family float, 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 socket

# This function attempts to connect to a port and checks if it's in use.
# It requires:
# - port: a port number (as string or int)
# - host: a hostname/IP address (defaults to '127.0.0.1')

# Since we can't actually test real network connections in a sandbox,
# we'll provide arguments that won't cause the function to hang or error
# before it tries the socket connection.

# port should be a string representation of a port number
# host should be a valid hostname or IP address

ARGS = {
    'port': '9999',
    'host': '127.0.0.1'
}
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 bool, ndim 0, dtype family float, 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
convert_messages_to_prompt Ran
ran on a synthesized input
yusaemeow/collab-overcooked/src/collab/utils.py
code served (permissive licence) · get_code("d68e86e16e480219")
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

def convert_messages_to_prompt(messages):
    """
    Converts a list of messages(for chat) to a prompt (for completion) for OpenAI's API.

    :param messages:
    :return: prompt
    """
    prompt = ''
    for message in messages:
        prompt += f"{message['content']}\n"
    return prompt


# Setup code to construct valid arguments
torch.manual_seed(0)

messages = [
    {'content': 'Hello, how are you?'},
    {'content': 'I am doing well, thank you.'},
    {'content': 'What is your name?'}
]

ARGS = {
    'messages': messages
}
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
load_dict_from_file Ran
ran; our contract draft was wrong, not the code
yusaemeow/collab-overcooked/lib/overcooked_ai/overcooked_ai_py/utils.py
code served (permissive licence) · get_code("591c3a0919cdc129")
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 a dictionary representation
temp_dir = tempfile.mkdtemp()
filepath = os.path.join(temp_dir, "test_dict.py")

# Write a simple dictionary to the file
test_dict = {
    "key1": "value1",
    "key2": 42,
    "key3": [1, 2, 3],
    "key4": {"nested": "dict"}
}

with open(filepath, 'w') as f:
    f.write(repr(test_dict))

ARGS = {
    "filepath": filepath
}
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
load_dict_from_txt Ran
ran; our contract draft was wrong, not the code
yusaemeow/collab-overcooked/lib/overcooked_ai/overcooked_ai_py/utils.py
code served (permissive licence) · get_code("1db1eb104270f970")
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 tempfile
import os

# Create a temporary directory and file for testing
temp_dir = tempfile.mkdtemp()
temp_filepath = os.path.join(temp_dir, "test_config")

# Create a valid dict and write it to a .txt file
test_dict = {
    "model_name": "test_model",
    "learning_rate": 0.001,
    "batch_size": 32,
    "layers": [64, 128, 256],
    "use_dropout": True
}

# Write the dict to file with .txt extension
with open(temp_filepath + ".txt", "w") as f:
    f.write(str(test_dict))

# Define ARGS with the filename parameter (without .txt extension)
# The function will add .txt via fix_filetype
ARGS = {
    "filename": temp_filepath
}
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
load_pickle Ran
ran; our contract draft was wrong, not the code
yusaemeow/collab-overcooked/lib/overcooked_ai/overcooked_ai_py/utils.py
code served (permissive licence) · get_code("c770dc82622e81c4")
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 pickle
import os

# Create a temporary directory and file for testing
temp_dir = tempfile.mkdtemp()
temp_file = os.path.join(temp_dir, "test_data")

# Create and save a pickle file
test_data = {"key": "value", "numbers": [1, 2, 3, 4, 5]}
with open(temp_file + ".pickle", "wb") as f:
    pickle.dump(test_data, f)

ARGS = {
    "filename": temp_file
}
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
retry_with_exponential_backoff Ran
ran on a synthesized input
yusaemeow/collab-overcooked/src/collab/utils.py
code served (permissive licence) · get_code("e164796d18f375ac")
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 random
import time
import openai

# Create a mock openai.RateLimitError if it doesn't exist
if not hasattr(openai, 'RateLimitError'):
    class RateLimitError(Exception):
        pass
    openai.RateLimitError = RateLimitError

# Define a simple test function that we'll wrap
call_count = 0

def test_function(x, y=10):
    global call_count
    call_count += 1
    # Succeed on the third attempt
    if call_count < 3:
        raise openai.RateLimitError("Rate limited")
    return x + y

# Define ARGS with all parameters
ARGS = {
    'func': test_function,
    'initial_delay': 0.01,
    'exponential_base': 2.0,
    'jitter': True,
    'max_retries': 10,
    'errors': (openai.RateLimitError,)
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
openai — 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
if_two_sentence_similar_meaning Not yet run
unverified (harvested, no recorded run)
yusaemeow/collab-overcooked/src/collab/modules.py
code served (permissive licence) · get_code("d7f810efb3411a9a")
How it ran

no invocation recorded.

Also present in other papers' code

Of 7 samples lifted from this paper's linked repositories, 7 carry a function identity and 0 do not. Of the 7 distinct functions with an identity, 4 are also present in other papers whose code we hold (4 with at least one paper dated earlier, 0 the same month, 0 later) and 3 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
load_dict_from_file 3 other papers
3 dated earlier
2106.10853 (2021-06)
On the Importance of Environments in Human-Robot Coordination
2302.04831 (2023-02)
Cooperative Open-ended Learning Framework for Zero-shot Coordination
2310.05208 (2023-10)
ZSC-Eval: An Evaluation Toolkit and Benchmark for Multi-agent Zero-shot Coordination
load_dict_from_txt 3 other papers
3 dated earlier
2106.10853 (2021-06)
On the Importance of Environments in Human-Robot Coordination
2302.04831 (2023-02)
Cooperative Open-ended Learning Framework for Zero-shot Coordination
2310.05208 (2023-10)
ZSC-Eval: An Evaluation Toolkit and Benchmark for Multi-agent Zero-shot Coordination
load_pickle 3 other papers
3 dated earlier
2106.10853 (2021-06)
On the Importance of Environments in Human-Robot Coordination
2302.04831 (2023-02)
Cooperative Open-ended Learning Framework for Zero-shot Coordination
2310.05208 (2023-10)
ZSC-Eval: An Evaluation Toolkit and Benchmark for Multi-agent Zero-shot Coordination
convert_messages_to_prompt 1 other paper
1 dated earlier
2302.04831 (2023-02)
Cooperative Open-ended Learning Framework for Zero-shot Coordination

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

Large Language Models (LLMs) based agent systems have made great strides in realworld applications beyond traditional NLP tasks. This paper proposes a new LLMbased Multi-Agent System (LLM-MAS) benchmark, Collab-Overcooked, built on the popular Overcooked-AI game with more applicable and challenging tasks in interactive environments. Collab-Overcooked extends existing benchmarks in two novel ways. First, it provides a multi-agent framework supporting diverse tasks and objectives and encourages collaboration through natural language communication. Second, it introduces a spectrum of process-oriented evaluation metrics to assess the fine-grained collaboration capabilities of different LLM agents, a dimension often overlooked in prior work. We conduct extensive experiments with 13 popular LLMs and show that, while the LLMs exhibit a strong ability in goal interpretation, there are significant shortcomings in active collaboration and continuous adaptation, which are critical for efficiently fulfilling complex tasks. Notably, we highlight the strengths and weaknesses of LLM-MAS and provide insights for improving and evaluating LLM-MAS on a unified and open-source benchmark. The environments, 30 open-ended tasks, and the evaluation package are publicly available at https://github. com/YusaeMeow/Collab-Overcooked.

For agents

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

get_harvested_code_for_paper("2502.20073")
get_code_for_paper("2502.20073")
have("2502.20073")

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

curl https://syntology.ai/api/ran/2502.20073.json

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

[![Syntology run record](https://syntology.ai/api/ran/2502.20073.svg)](https://syntology.ai/paper/2502.20073)

Connect an agent — have() is free.