Large Language Model based Multi-Agents: A Survey of Progress and Challenges

Taicheng Guo1 , Xiuying Chen2 , Yaqi Wang3∗ , Ruidi Chang , Shichao Pei4 ,
Nitesh V. Chawla1 , Olaf Wiest1 , Xiangliang Zhang1†
1
University of Notre Dame, 2 King Abdullah University of Science and Technology
3
Southern University of Science and Technology, 4 University of Massachusetts Boston
{tguo2, nchawla, owiest, xzhang33}@nd.edu, [email protected], [email protected],
[email protected], [email protected]
arXiv:2402.01680v2 [cs.CL] 19 Apr 2024

Abstract decision-making in a wide range of contexts [Yao et al.,

2023; Shinn et al., 2023; Li et al., 2023d]. Timely survey
Large Language Models (LLMs) have achieved re- papers systematically summarize the progress of LLM-based
markable success across a wide array of tasks. agents, as seen in works [Xi et al., 2023; Wang et al., 2023b].
Due to the impressive planning and reasoning abil-
Based on the inspiring capabilities of the single LLM-
ities of LLMs, they have been used as autonomous
based agent, LLM-based Multi-Agents have been proposed
agents to do many tasks automatically. Recently,
to leverage the collective intelligence and specialized pro-
based on the development of using one LLM as a
files and skills of multiple agents. Compared to systems us-
single planning or decision-making agent, LLM-
ing a single LLM-powered agent, multi-agent systems offer
based multi-agent systems have achieved consid-
advanced capabilities by 1) specializing LLMs into various
erable progress in complex problem-solving and
distinct agents, each with different capabilities, and 2) en-
world simulation. To provide the community with
abling interactions among these diverse agents to simulate
an overview of this dynamic field, we present this
complex real-world environments effectively. In this context,
survey to offer an in-depth discussion on the essen-
multiple autonomous agents collaboratively engage in plan-
tial aspects of multi-agent systems based on LLMs,
ning, discussions, and decision-making, mirroring the co-
as well as the challenges. Our goal is for readers to
operative nature of human group work in problem-solving
gain substantial insights on the following questions:
tasks. This approach capitalizes on the communicative ca-
What domains and environments do LLM-based
pabilities of LLMs, leveraging their ability to generate text
multi-agents simulate? How are these agents pro-
for communication and respond to textual inputs. Further-
filed and how do they communicate? What mech-
more, it exploits LLMs’ extensive knowledge across vari-
anisms contribute to the growth of agents’ capaci-
ous domains and their latent potential to specialize in spe-
ties? For those interested in delving into this field
cific tasks. Recent research has demonstrated promising re-
of study, we also summarize the commonly used
sults in utilizing LLM-based multi-agents for solving vari-
datasets or benchmarks for them to have convenient
ous tasks, such as software development [Hong et al., 2023;
access. To keep researchers updated on the latest
Qian et al., 2023], multi-robot systems [Mandi et al., 2023;
studies, we maintain an open-source GitHub repos-
Zhang et al., 2023c], society simulation [Park et al., 2023;
itory, dedicated to outlining the research on LLM-
Park et al., 2022], policy simulation [Xiao et al., 2023;
based multi-agent systems.
Hua et al., 2023], and game simulation [Xu et al., 2023c;
Wang et al., 2023c]. Due to the nature of interdisciplinary
1 Introduction study in this field, it has attracted a diverse range of re-
searchers, expanding beyond AI experts to include those from
Large Language Models (LLMs) have recently shown re- social science, psychology, and policy research. The vol-
markable potential in reaching a level of reasoning and plan- ume of research papers is rapidly increasing, as shown in
ning capabilities comparable to humans. This ability ex- Fig. 1 (inspired by the design in [Gao et al., 2023b]), thus
actly aligns with the expectations of humans for autonomous broadening the impact of LLM-based Multi-Agent research.
agents that can perceive the surroundings, make decisions, Nonetheless, earlier efforts were undertaken independently,
and take actions in response [Xi et al., 2023; Wooldridge and resulting in an absence of a systematic review to summarize
Jennings, 1995; Russell and Norvig, 2009; Guo et al., 2023; them, establish comprehensive blueprint of this field, and ex-
Liang et al., 2023]. Hence, LLM-based agent has been stud- amine future research challenges. This underscores the sig-
ied and rapidly developed to understand and generate human- nificance of our work and serves as the motivation behind pre-
like instructions, facilitating sophisticated interactions and senting this survey paper, dedicated to the research on LLM-
∗
This work was done when Yaqi was visiting students at the Uni- based multi-agent systems.
versity of Notre Dame. We expect that our survey can make significant contribu-
†
Corresponding author. tions to both the research and development of LLMs and to
Figure 1: The rising trend in the research field of LLM-based Multi-Agents. For Problem Solving and World Simulation, we categorize
current work into several categories and count the number of papers of different types at 3-month intervals. The number at each leaf node
denotes the count of papers within that category.

a wider range of interdisciplinary studies employing LLMs. in specific ways; 3) agent communication, which examines
Readers will gain a comprehensive overview of LLM-based how agents exchange messages and collaborate; and 4) agent
Multi-Agent (LLM-MA) systems, grasp the fundamental capability acquisition, which explores how agents develop
concepts involved in establishing multi-agent systems based their abilities to effectively solve problems. An additional
on LLMs, and catch the latest research trends and applica- perspective for reviewing studies about LLM-MA is their ap-
tions in this dynamic field. We recognize that this field is in plication. In Section 4, we categorize current applications
its early stages and is rapidly evolving with fresh methodolo- into two primary streams: multi-agents for problem-solving
gies and applications. To provide a sustainable resource com- and multi-agents for world simulation. To guide individuals
plementing our survey paper, we maintain an open-source in identifying appropriate tools and resources, we present
GitHub repository1 . We hope that our survey will inspire fur- open-source implementation frameworks for studying LLM-
ther exploration and innovation in this field, as well as appli- MA, as well as the usable datasets and benchmarks in Sec-
cations across a wide array of research disciplines. tion 5. Based on the previous summary, we open the dis-
To assist individuals from various backgrounds in under- cussion for future research challenges and opportunities in
standing LLM-MA techniques and to complement existing Section 6. The conclusions are summarized in Section 7.
surveys by tackling unresolved questions, we have organized
our survey paper in the following manner. After laying out 2 Background
the background knowledge in Section 2, we address a piv- 2.1 Single-Agent Systems Powered LLMs
otal question: How are LLM-MA systems aligned with the
collaborative task-solving environment? To answer this, we We introduce the background by first outlining the capabili-
present a comprehensive schema for positioning, differenti- ties of a single-agent system based on LLMs, following the
ating, and connecting various aspects of LLM-MA systems discussion presented in [Weng, 2023].
in Section 3. We delve into this question by discussing: 1) Decision-making Thought: This term denotes the capabil-
the agents-environment interface, which details how agents ity of LLM-based agents, guided by prompts, to break down
interact with the task environment; 2) agent profiling, which complex tasks into smaller subgoals [Khot et al., 2023], think
explains how an agent is characterized by an LLM to behave through each part methodically (sometimes exploring mul-
tiple paths) [Yao et al., 2023], and learn from past experi-
1
https://github.com/taichengguo/LLM MultiAgents Survey Papers ences [Shinn et al., 2023] to perform better decision-making
on complex tasks. This capability enhances the autonomy checking roles. Following these actions, agents receive feed-
of a single LLM-based agent and bolsters its effectiveness in back from the environment, informing them of the game’s
problem-solving. current state. This information guides the agents in adjust-
Tool-use: LLM-based agents’ tool-use capability allows ing their strategies over time, responding to the evolving
them to leverage external tools and resources to accom- gameplay and interactions with other agents. The Agents-
plish tasks, enhancing their functional capabilities and oper- Environment Interface refers to the way in which agents in-
ate more effectively in diverse and dynamic environments [Li teract with and perceive the environment. It’s through this
et al., 2023d; Ruan et al., 2023; Gao et al., 2023b]. interface that agents understand their surroundings, make de-
cisions, and learn from the outcomes of their actions. We
Memory: This ability refers to the capability of LLM- categorize the current interfaces in LLM-MA systems into
based agent for conducting in-context learning [Dong et al., three types, Sandbox, Physcial, and None, as detailed in Ta-
2023a] as short memory or external vector database [Lewis et ble 1. The Sandbox refers to a simulated or virtual environ-
al., 2021] as long memory to preserve and retrieve informa- ment built by human where agents can interact more freely
tion over prolonged periods [Wang et al., 2023b]. This ability and experiment with various actions and strategies. This kind
enables a single LLM-based agent to maintain contextual co- of interface is widely used in software development (code
herence and enhance learning from interactions. interpreter as simulated environment) [Hong et al., 2023],
gaming (using game rules as simulated environment) [Mao
2.2 Single-Agent VS. Multi-Agent Systems et al., 2023], etc. The Physical is a real-world environment
Single-Agent systems empowered by LLMs have shown in- where agents interact with physical entities and obey real-
spiring cognitive abilities [Sumers et al., 2023]. The con- world physics and constraints. In physical space, agents nor-
struction of such systems concentrates on formulating their mally need to take actions that can have direct physical out-
internal mechanisms and interactions with the external en- comes. For example, in tasks such as sweeping the floor,
vironment. Conversely, LLM-MA systems emphasize di- making sandwiches, packing groceries, and arranging cab-
verse agent profiles, inter-agent interactions, and collective inets, robotic agents are required to perform actions itera-
decision-making processes. From this perspective, more dy- tively, observe the physical environment, and continuously
namic and complex tasks can be tackled by the collaboration refine their actions [Mandi et al., 2023]. Lastly, None refers
of multiple autonomous agents, each of which is equipped to scenarios where there is no specific external environment,
with unique strategies and behaviors, and engaged in com- and agents do not interact with any environment. For exam-
munication with one another. ple, many applications [Du et al., 2023; Xiong et al., 2023;
Chan et al., 2023] utilize multiple agents to debate a ques-
3 Dissecting LLM-MA Systems: Interface, tion to reach a consensus. These applications primarily focus
on communication among agents and do not depend on the
Profiling, Communication, and Capabilities external environment.
In this section, we delve into the intricacies of LLM-MA sys-
tems, where multiple autonomous agents engage in collabo- 3.2 Agents Profiling
rative activities akin to human group dynamics in problem-
solving scenarios. A critical inquiry we address is how In LLM-MA systems, agents are defined by their traits, ac-
these LLM-MA systems are aligned to their operational envi- tions, and skills, which are tailored to meet specific goals.
ronments and the collective objectives they are designed to Across various systems, agents assume distinct roles, each
achieve. To shed light on this, we present the general ar- with comprehensive descriptions encompassing characteris-
chitecture of these systems in Fig. 2. Our analysis dissects tics, capabilities, behaviors, and constraints. For instance,
the operational framework of these systems, focusing on four in gaming environments, agents might be profiled as players
key aspects: the agents-environment interface, agent profil- with varying roles and skills, each contributing differently to
ing, agent communication, and agent capability acquisition. the game’s objectives. In software development, agents could
take on the roles of product managers and engineers, each
3.1 Agents-Environment Interface with responsibilities and expertise that guide the development
The operational environments defines the specific contexts or process. Similarly, in a debating platform, agents might be
settings in which the LLM-MA systems are deployed and designated as proponents, opponents, or judges, each with
interact. For example, these environments can be like soft- unique functions and strategies to fulfill their roles effectively.
ware development [Hong et al., 2023], gaming [Mao et al., These profiles are crucial for defining the agents’ interactions
2023], and various other domains such as financial markets and effectiveness within their respective environments. Table
[Li et al., 2023g] or even social behavior modeling [Park et 1 lists the agent Profiles in recent LLM-MA works.
al., 2023]. The LLM-based agents perceive and act within Regarding the Agent Profiling Methods, we categorized
the environment, which in turn influences their behavior and them into three types: Pre-defined, Model-Generated, and
decision making. For example, in the Werewolf Game simu- Data-Derived. In the Pre-defined cases, agent profiles are
lation, the sandbox environment sets the game’s framework, explicitly defined by the system designers. The Model-
including transitions from day to night, discussion periods, Generated method creates agent profiles by models, e.g.,
voting mechanics, and reward rules. Agents, such as were- large language models. The Data-Derived method involves
wolves and the Seer, perform specific actions like killing or constructing agent profiles based on pre-existing datasets.
 Figure 2: The Architecture of LLM-MA Systems.

3.3 Agents Communication

The communication between agents in LLM-MA systems is
the critical infrastructure supporting collective intelligence.
We dissect agent communication from three perspectives: 1)
Communication Paradigms: the styles and methods of in-
teraction between agents; 2) Communication Structure: the
organization and architecture of communication networks
within the multi-agent system; and 3) Communication Con-
tent exchanged between agents.
Communication Paradigms: Current LLM-MA systems
mainly take three paradigms for communication: Coopera-
tive, Debate, and Competitive. Cooperative agents work to-
gether towards a shared goal or objectives, typically exchang-
ing information to enhance a collective solution. The Debate
paradigm is employed when agents engage in argumentative
interactions, presenting and defending their own viewpoints
or solutions, and critiquing those of others. This paradigm
is ideal for reaching a consensus or a more refined solution.
Competitive agents work towards their own goals that might
be in conflict with the goals of other agents. Figure 3: The Agent Communication Structure.
Communication Structure: Fig. 3 shows four typical
communication structures in LLM-MA systems. Layered
communication is structured hierarchically, with agents at work, where agents directly communicate with each other,
each level having distinct roles and primarily interacting a structure commonly employed in world simulation appli-
within their layer or with adjacent layers. [Liu et al., 2023] cations. Centralized communication involves a central agent
introduces a framework called Dynamic LLM-Agent Net- or a group of central agents coordinating the system’s com-
work (DyLAN), which organizes agents in a multi-layered munication, with other agents primarily interacting through
feed-forward network. This setup facilitates dynamic inter- this central node. Shared Message Pool is proposed by
actions, incorporating features like inference-time agent se- MetaGPT [Hong et al., 2023] to improve the communication
lection and an early-stopping mechanism, which collectively efficiency. This communication structure maintains a shared
enhance the efficiency of cooperation among agents. De- message pool where agents publish messages and subscribe
centralized communication operates on a peer-to-peer net- to relevant messages based on their profiles, thereby boosting
communication efficiency. to decide subsequent actions as seen in Memory-based solu-
tions, agents can dynamically self-evolve by modifying them-
Communication Content: In LLM-MA systems, the selves such as altering their initial goals and planning strate-
Communication Content typically takes the form of text. The gies, and training themselves based on feedback or commu-
specific content varies widely and depends on the particular nication logs. [Nascimento et al., 2023] proposes a self-
application. For example, in software development, agents control loop process to allow each agent in the multi-agents
may communicate with each other about code segments. In systems to be self-managed and self-adaptive to dynamic en-
simulations of games like Werewolf, agents might discuss vironments, thereby improving the cooperation efficiency of
their analyses, suspicions, or strategies. multiple agents. [Zhang et al., 2023b] introduces ProA-
gent which anticipates teammates’ decisions and dynami-
3.4 Agents Capabilities Acquisition cally adjusts each agent’s strategies based on the communi-
The Agents Capabilities Acquisition is a crucial process in cation logs between agents, facilitating mutual understand-
LLM-MA, enabling agents to learn and evolve dynamically. ing and improving collaborative planning capability. [Wang
In this context, there are two fundamental concepts: the types et al., 2023a] discusses a Learning through Communication
of feedback from which agents should learn to enhance their (LTC) paradigm, using the communication logs of multi-
capabilities, and the strategies for agents to adjust themselves agents to generate datasets to train or fine-tune LLMs. LTC
to effectively solve complex problems. enables continuous adaptation and improvement of agents
through interaction with their environments and other agents,
Feedback: Feedback involves the critical information that breaking the limits of in-context learning or supervised fine-
agents receive about the outcome of their actions, helping the tuning, which don’t fully utilize the feedback received dur-
agents learn the potential impact of their actions and adapt ing interactions with the environment and external tools
to complex and dynamic problems. In most studies, the for- for continuous training. Self-Evolution enables agents’ au-
mat of feedback provided to agents is textual. Based on the tonomous adjustment in their profiles or goals, rather than
sources from which agents receive this feedback, it can be just learning from historical interactions. 3) Dynamic Gen-
categorized into four types. 1) Feedback from Environ- eration. In some scenarios, the system can generate new
ment, e.g., from either real world environments or virtual agents on-the-fly during its operation [Chen et al., 2023a;
environments [Wang et al., 2023b]. It is prevalent in most Chen et al., 2023c]. This capability enables the system to
LLM-MA for problem-solving scenarios, including Software scale and adapt effectively, as it can introduce agents that are
Development (agents obtain feedback from Code Interpreter), specifically designed to address current needs and challenges.
and Embodied multi-agents systems (robots obtain feedback With the scaling up LLM-MA with a larger number of
from real-world or Simulated environments). 2) Feedback agents, the escalating complexity of managing various kinds
from Agents Interactions means that the feedback comes of agents has been a critical problem. Agents Orchestration
from the judgement of other agents or from agents communi- emerged as a pivotal challenge and began to gain attention
cations. It is common in problem-solving scenarios like sci- in [Moura, 2023; Dibia, 2023]. We will further discuss this
ence debates, where agents learn to critically evaluate and re- topic in Section 6.4.
fine the conclusions through communications. In world sim-
ulation scenarios such as Game Simulation, agents learn to
refine strategies based on previous interactions between other
4 Applications
agents. 3) Human Feedback comes directly from humans LLM-MA systems have been used in a wide range of appli-
and is crucial for aligning the multi-agent system with hu- cations. We summarize two kinds of applications in Table 1:
man values and preferences. This kind of feedback is widely Problem Solving and World Simulation. We elaborate on
used in most “Human-in-the-loop” applications [Wang et al., these applications below. Note that this is a fast growing re-
2021]. Last 4) None. In some cases, there is no feedback pro- search field and new applications appear almost everyday. We
vided to the agents. This often happens for world simulation maintain an open source repository to report the latest work.
works focused on analyzing simulated results rather than the
planning capabilities of agents. In such scenarios, like prop- 4.1 LLM-MA for Problem Solving
agation simulation, the emphasis is on result analysis, and The main motivation of using LLM-MA for problem solving
hence, feedback is not a component of the system. is to harness the collective capabilities of agents with spe-
cialized expertise. These agents, each acting as individuals,
Agents Adjustment to Complex Problems: To enhance collaborate to address complex problems effectively, such as
their capabilities, agents in LLM-MA systems can adapt software development, embodied agents, science experiments
through three main solutions. 1) Memory. Most LLM- and science debate. These application examples are intro-
MA systems leverage a memory module for agents to ad- duced next.
just their behavior. Agents store information from previ-
ous interactions and feedback in their memory. When per- 4.1.1 Software Development
forming actions, they can retrieve relevant, valuable memo- Given that software development is a complex endeavor re-
ries, particularly those containing successful actions for simi- quiring the collaboration of various roles like product man-
lar past goals, as highlighted in [Wang et al., 2023b]. This agers, programmers, and testers, LLM-MA systems are typ-
process aids in enhancing their current actions. 2) Self- ically set to emulate these distinct roles and collaborate to
Evolution. Instead of only relying on the historical records address the intricate challenge. Following the waterfall or
 Agents
Agents-Env. Agents Profiling Agents Capabilities Acquisition
Communication
Interface
Motivation Research Domain & Goals Work Profiling Profiles Agents
Paradigms Structure Feedback from
methods (examples) Adjustment
Environment,
Pre-defined, CTO, Memory,
[Qian et al., 2023] Sandbox Cooperative Layered Agent interaction,
Model-Generated programmer Self-Evolution
Human
Environment,
Product Manager, Layered, Memory,
Software development [Hong et al., 2023] Sandbox Pre-defined Cooperative Agent interaction,
Engineer Shared Message Pool Self-Evolution
Human
Pre-defined, Analyst, Environment, Memory,
[Dong et al., 2023b] Sandbox Cooperative Layered
Model-Generated coder Agent interaction Self-Evolution
Multi-robot Sandbox, Centralized, Environment,
[Chen et al., 2023d] Pre-defined Robots Cooperative Memory
planning Physical Decentralized Agent interaction
Embodied Multi-robot Sandbox, Environment,
[Mandi et al., 2023] Pre-defined Robots Cooperative Decentralized Memory
Agents collaboration Physical Agent interaction
Multi-Agents Environment,
Problem [Zhang et al., 2023c] Sandbox Pre-defined Robots Cooperative Decentralized Memory
cooperation Agent interaction
Solving
Strategy planers,
Science Optimization Environment,
[Zheng et al., 2023] Physical Pre-defined literature Cooperative Centralized Memory
Experiments of MOF Human
collector, coder
Improving
[Du et al., 2023] None Pre-defined Agents Debate Decentralized Agent interaction Memory
Factuality
Proponent,
Science Examining, Centralized,
[Xiong et al., 2023] None Pre-defined Opponent, Debate Agent interaction Memory
Debate Inter-Consistency Decentralized
Judge
Evaluators Centralized,
[Chan et al., 2023] None Pre-defined Agents Debate Agent interaction Memory
for debates Decentralized
Multi-Agents Cardiology, Debate, Centralized,
[Tang et al., 2023] None Pre-defined Agent interaction Memory
for Medication Surgery Cooperative Decentralized
Modest Community Pharmacy, Environment,
[Park et al., 2023] Sandbox Model-generated - - Memory
(25 persons) shopkeeper Agent interaction
Online community Pre-defined, Camping, Dynamic
[Park et al., 2022] None - - Agent interaction
(1000 persons) Model-generated fishing Generation
Pre-defined, Real-world
Society Emotion propagation [Gao et al., 2023a] None - - Agent interaction Memory
Model-generated user
Real-time Real-world Environment,
[Kaiya et al., 2023] Sandbox Pre-defined - - Memory
social interactions user Agent interaction
NIN, NINL,
Opinion dynamics [Li et al., 2023a] None Pre-defined - - Agent interaction Memory
NIL
Seer, Cooperative,
[Xu et al., 2023b] Environment,
WereWolf Sandbox Pre-defined werewolf, Debate, Decentralized Memory
World [Xu et al., 2023c] Agent interaction
villager Competitive
Simulation
Servant, Cooperative,
[Light et al., 2023a] Environment,
Gaming Avalon Sandbox Pre-defined Merlin, Debate, Decentralized Memory
[Wang et al., 2023c] Agent interaction
Assassin Competitive
Cooperative, Environment,
Welfare Diplomacy [Mukobi et al., 2023] Sandbox Pre-defined Countries Decentralized Memory
Competitive Agent interaction
Human behavior
[Aher et al., 2023] Sandbox Pre-defined Humans - - Agent interaction Memory
Simulation
Psychology Collaboration Cooperative,
[Zhang et al., 2023d] None Pre-defined Agents Decentralized Agent interaction Memory
Exploring Debate
Macroeconomic Pre-defined,
[Li et al., 2023e] None Labor Cooperative Decentralized Agent interaction Memory
simulation Model-generated
Information Pre-defined, Cooperative, Environment,
Economy [Anonymous, 2023] Sandbox Buyer Decentralized Memory
Marketplaces Data-Derived Competitive Agent interaction
Improving Environment,
[Li et al., 2023g] Physical Pre-defined Trader Debate Decentralized Memory
financial trading Agent interaction
Pre-defined, Restaurant, Environment, Memory,
Economic theories [Zhao et al., 2023] Sandbox Competitive Decentralized
Model-Generated Customer Agent interaction Self-Evolution
Simulating Users from
Recommender [Zhang et al., 2023a] Sandbox Data-Derived - - Environment Memory
user behaviors MovieLens-1M
Systems
Simulating user-item Pre-defined, User Agents Environment,
[Zhang et al., 2023e] Sandbox Cooperative Decentralized Memory
interactions Data-Derived Item Agents Agent interaction
Public
Policy [Xiao et al., 2023] None Pre-defined Residents Cooperative Decentralized Agent interaction Memory
Administration
Making
War Simulation [Hua et al., 2023] None Pre-defined Countries Competitive Decentralized Agent interaction Memory
Human Behaviors [Ghaffarzadegan Pre-defined, Conformity Environment,
Sandbox Cooperative Decentralized Memory
Disease to epidemics et al., 2023] Model-Generated traits Agent interaction
Memory,
[Williams Pre-defined, Adults aged Environment,
Public health Sandbox Cooperative Decentralized Dynamic
et al., 2023] Model-Generated 18 to 64 Agent interaction
Generation

Table 1: Summary of the LLM-MA studies. We categorize current work according to their motivation, research domains and goals, and detail
each work from different aspects regarding Agents-Environment Interface, Agents Profiling, Agents Communication and Agents Capability
Acquisition. “-” denotes that a particular element is not specifically mentioned in this work.

Standardized Operating Procedures (SOPs) workflow of the in software development, autonomously collaborating to gen-
software development, the communication structure among erate code. Moreover, [Qian et al., 2023] presents an end-to-
agents is usually layered. Agents generally interact with the end framework for software development, utilizing multiple
code interpreter, other agents or human to iteratively refine agents for software development without incorporating ad-
the generated code. [Li et al., 2023b] first proposes a sim- vanced human teamwork experience. [Hong et al., 2023] first
ple role-play agent framework, which utilizes the interplay incorporates human workflow insights for more controlled
of two roles to realize autonomous programming based on and validated performance. It encodes SOPs into prompts to
one-sentence user instruction. It provides insights into the enhance structured coordination. [Huang et al., 2023a] delves
“cognitive” processes of communicative agents. [Dong et al., deeper into multi-agent based programming by solving the
2023b] makes LLMs work as distinct “experts” for sub-tasks problem of balancing code snippet generation with effective
test case generation, execution, and optimization. by a joint debating process. Through multiple rounds of de-
bate, the agents converge on a single, consensus answer. [Du
4.1.2 Embodied Agents et al., 2023] leverages the multi-agents debate process on a
Most embodied agents applications inherently utilize multi- set of six different reasoning and factual accuracy tasks and
ple robots working together to perform complex real-world demonstrates that LLM-MA debating can improve factual-
planning and manipulation tasks such as warehouse manage- ity. [Xiong et al., 2023] focuses on the commonsense rea-
ment with heterogeneous robot capabilities. Hence, LLM- soning tasks and formulates a three-stage debate to align with
MA can be used to model robots with different capabilities real-world scenarios including fair debate, mismatched de-
and cooperate with each other to solve real-world physical bate, and roundtable debate. The paper also analyzes the
tasks. [Dasgupta et al., 2023] first explores the potential to inter-consistency between different LLMs and claims that de-
use LLM as an action planner for embedded agents. [Mandi bating can improve the inter-consistency. [Tang et al., 2023]
et al., 2023] introduces RoCo, a novel approach for multi- also utilizes multiple LLM-based agents as distinct domain
robot collaboration that uses LLMs for high-level commu- experts to do the collaborative discussion on a medical report
nication and low-level path planning. Each robotic arm is to reach a consensus for medical diagnosis.
equipped with an LLM, cooperating with inverse kinemat-
ics and collision checking. Experimental results demonstrate 4.2 LLM-MA for World Simulation
the adaptability and success of RoCo in collaborative tasks.
[Zhang et al., 2023c] presents CoELA, a Cooperative Em- Another mainstream application scenario of LLM-MA is the
bodied Language Agent, managing discussions and task plan- world simulation. Research in this area is rapidly growing
ning in an LLM-MA setting. This challenging setting is and spans a diverse range of fields including social sciences,
featured with decentralized control, complex partial observa- gaming, psychology, economics, policy-making, etc. The key
tion, costly communication, and multi-objective long-horizon reason for employing LLM-MA in world simulations lies in
tasks. [Chen et al., 2023d] investigates communication chal- their exceptional role-playing abilities, which are crucial for
lenges in scenarios involving a large number of robots, as realistically depicting various roles and viewpoints in a sim-
assigning each robot an LLM will be costly and unpracti- ulated world. The environment of world simulation projects
cal due to the long context. The study compares four com- is usually crafted to reflect the specific scenario being simu-
munication frameworks, centralized, decentralized, and two lated, with agents designed in various profiles to match this
hybrid models, to evaluate their effectiveness in coordinating context. Unlike the problem solving systems that focus on
complex multi-agent tasks. [Yu et al., 2023] proposes Co- agent cooperation, world simulation systems involve diverse
NavGPT for multi-robot cooperative visual target navigation, methods of agent management and communication, reflecting
integrating LLM as a global planner to assign frontier goals the complexity and variety of real-world interactions. Next,
to each robot. [Chen et al., 2023b] proposes an LLM-based we explore simulations conducted in diverse fields.
consensus-seeking framework, which can be applied as a co- 4.2.1 Societal Simulation
operative planner to a multi-robot aggregation task. In societal simulation, LLM-MA models are used to simu-
4.1.3 Science Experiments late social behaviors, aiming to explore the potential social
Like multiple agents play as different specialists and cooper- dynamics and propagation, test social science theories, and
ate to solve the Software Development and Embodied Agents populate virtual spaces and communities with realistic social
problem, multiple agents can also be used to form a science phenomena [Park et al., 2023]. Leveraging LLM’s capabili-
team to conduct science experiments. One important differ- ties, agents with unique profiles engage in extensive commu-
ence from previous applications lies in the crucial role of hu- nication, generating rich behavioral data for in-depth social
man oversight, due to the high expenses of the science ex- science analysis.
periments and the hallucination of the LLM agents. Human The scale of societal simulation has expanded over time,
experts are at the center of these agents to process the infor- beginning with smaller, more intimate settings and progress-
mation of agents and give feedback to the agents. [Zheng et ing to larger, more intricate ones. Initial work by [Park et al.,
al., 2023] utilizes multiple LLM-based agents, each focusing 2023] introduces generative agents within an interactive sand-
on specific tasks for the science experiments including strat- box environment reminiscent of the sims, allowing end users
egy planning, literature search, coding, robotic operations, to engage with a modest community of 25 agents through nat-
and labware design. All these agents interact with humans ural language. At the same time, [Park et al., 2022] develops
to work collaboratively to optimize the synthesis process of Social Simulacra, which constructs a simulated community
complex materials. of 1,000 personas. This system takes a designer’s vision for
a community—its goals, rules, and member personas—and
4.1.4 Science Debate simulates it, generating behaviors like posting, replying, and
LLM-MA can be set for science debating scenarios, where even anti-social actions. Building on this, [Gao et al., 2023a]
agents debate with each other to enhance the collective rea- takes the concept further by constructing vast networks com-
soning capabilities in tasks such as Massive Multitask Lan- prising 8,563 and 17,945 agents, respectively, designed to
guage Understanding (MMLU) [Hendrycks et al., 2020], simulate social networks focused on the topics of Gender Dis-
Math problems [Cobbe et al., 2021], and StrategyQA [Geva crimination and Nuclear Energy. This evolution showcases
et al., 2021]. The main idea is that each agent initially of- the increasing complexity and size of simulated environments
fers its own analysis of a problem, which is then followed in recent research. Recent studies such as [Chen et al., 2023b;
Kaiya et al., 2023; Li et al., 2023a; Li et al., 2023f; Ziems et This method focuses on observing and analyzing their varied
al., 2023] highlight the evolving complexity in multi-agent behaviors through statistical methods. Here, each agent op-
systems, LLM impacts on social networks, and their integra- erates independently, without interacting with others, essen-
tion into social science research. tially representing different individuals. Another approach
aligns more closely with societal simulations, where multiple
4.2.2 Gaming
agents interact and communicate with each other. In this sce-
LLM-MA is well-suited for creating simulated gaming en- nario, psychological theories are applied to understand and
vironments, allowing agents to assume various roles within analyze the emergent behavioral patterns. This method fa-
games. This technology enables the development of con- cilitates the study of interpersonal dynamics and group be-
trolled, scalable, and dynamic settings that closely mimic haviors, providing insights into how individual psychological
human interactions, making it ideal for testing a range of traits influence collective actions. [Ma et al., 2023] explores
game theory hypotheses [Mao et al., 2023; Xu et al., 2023b; the psychological implications and outcomes of employing
Gong et al., 2023]. Most games simulated by LLM-MA rely LLM-based conversational agents for mental well-being sup-
heavily on natural language communication, offering a sand- port. It emphasizes the need for carefully evaluating the use
box environment within different game settings for exploring of LLM-based agents in mental health applications from a
or testing game theory hypotheses including reasoning, coop- psychological perspective. [Kovač et al., 2023] introduces
eration, persuasion, deception, leadership, etc. a tool named SocialAI school for creating interactive envi-
[Akata et al., 2023] leverages behavioral game theory to
ronments simulating social interactions. It draws from devel-
examine LLMs’ behavior in interactive social settings, partic- opmental psychology to understand how agents can acquire,
ularly their performance in games like the iterated Prisoner’s demonstrate, and evolve social skills such as joint attention,
Dilemma and Battle of the Sexes. Furthermore, [Xu et al., communication, and cultural learning. [Zhang et al., 2023d]
2023b] proposes a framework using ChatArena library [Wu explores how LLM agents, with distinct traits and thinking
et al., 2023b] for engaging LLMs in communication games patterns, emulate human-like social behaviors such as confor-
like Werewolf, using retrieval and reflection on past commu- mity and majority rule. This integration of psychology into
nications for improvement, as well as the Chain-of-Thought the understanding of agent collaboration offers a novel lens
mechanism [Wei et al., 2022]. [Light et al., 2023b] explores for examining and enhancing the mechanisms behind LLM-
the potential of LLM agents in playing Resistance Avalon, in- based multi-agents systems. [Aher et al., 2023] introduces
troducing AVALONBENCH, a comprehensive game environ- Turing Experiments to evaluate the extent to which large lan-
ment and benchmark for further developing advanced LLMs guage models can simulate different aspects of human behav-
and multi-agent frameworks. [Wang et al., 2023c] also fo- iors. The Turing Experiments replicate classical experiments
cuses on the capabilities of LLM Agents in dealing with mis- and phenomena in psychology, economics, and sociology us-
information in the Avalon game, proposing the Recursive ing a question-answering format to mimic experimental con-
Contemplation (ReCon) framework to enhance LLMs’ ability ditions. They also design a prompt that is used to simulate
to discern and counteract deceptive information. [Xu et al., the responses of multiple different individuals by varying the
2023c] introduces a framework combining LLMs with rein- name. By simulating various kinds of individuals via LLM,
forcement learning (RL) to develop strategic language agents they show that larger models replicate human behavior more
for the Werewolf game. It introduces a new approach to use faithfully, but they also reveal a hyper-accuracy distortion, es-
RL policy in the case that the action and state sets are not pre- pecially in knowledge-based tasks.
defined but in the natural language setting. [Mukobi et al.,
2023] designs the “Welfare Diplomacy”, a general-sum vari- 4.2.4 Economy
ant of the zero-sum board game Diplomacy, where players
must balance military conquest and domestic welfare. It also LLM-MA is used to simulate economic and financial trading
offers an open-source benchmark, aiming to help improve the environments mainly because it can serve as implicit com-
cooperation ability of multi-agent AI systems. On top of that, putational models of humans. In these simulations, agents
there is a work [Li et al., 2023c] in a multi-agent cooperative are provided with endowments, and information, and set with
text game testing the agents’ Theory of Mind (ToM), the abil- pre-defined preferences, allowing for an exploration of their
ity to reason about the concealed mental states of others and actions in economic and financial contexts. This is similar to
is fundamental to human social interactions, collaborations, the way economists model ’homo economicus’, the character-
and communications. [Fan et al., 2023] comprehensively as- ization of man in some economic theories as a rational person
sesses the capability of LLMs as rational players, and iden- who pursues wealth for his own self-interest [Horton, 2023].
tifies the weaknesses of LLM-based Agents that even in the There are several studies demonstrate the diverse applications
explicit game process, agents may still overlook or modify of LLM-MA in simulating economic scenarios, encompass-
refined beliefs when taking actions. ing macroeconomic activities, information marketplaces, fi-
nancial trading, and virtual town simulations. Agents in-
4.2.3 Psychology teract in cooperative or debate, decentralized environments.
In psychological simulation studies, like in the societal simu- [Li et al., 2023e] employs LLMs for macroeconomic simu-
lation, multiple agents are utilized to simulate humans with lation, featuring prompt-engineering-driven agents that emu-
various traits and thought processes. However, unlike so- late human-like decision-making, thereby enhancing the real-
cietal simulations, one approach in psychology involves di- ism of economic simulations compared to rule-based or other
rectly applying psychological experiments to these agents. AI agents. [Anonymous, 2023] explores the buyer’s inspec-
 Motivation Domain Datasets and Benchmarks Used by Data Link
HumanEval [Hong et al., 2023] Link
Software Development MBPP [Hong et al., 2023] Link
SoftwareDev [Hong et al., 2023] Link
RoCoBench [Mandi et al., 2023] Link
Communicative Watch-And-Help (C-WAH) [Zhang et al., 2023c] Link
Embodied AI [Zhang et al., 2023c]
ThreeDWorld Multi-Agent Transport (TDW-MAT) Link
Problem Solving HM3D v0.2 [Yu et al., 2023] Link
MMLU [Tang et al., 2023] Link
MedQA [Tang et al., 2023] Link
PubMedQA [Tang et al., 2023] Link
Science Debate [Du et al., 2023]
GSM8K Link
StrategyQA [Xiong et al., 2023] Link
Chess Move Validity [Du et al., 2023] Link
SOTOPIA [Zhou et al., 2023b] /
Society Gender Discrimination [Gao et al., 2023a] /
Nuclear Energy [Gao et al., 2023a] /
Werewolf [Xu et al., 2023b] /
Avalon [Light et al., 2023b] /
Welfare Diplomacy [Mukobi et al., 2023] /
Gaming [Agashe et al., 2023]
Layout in the Overcooked-AI environment /
World Simulation Chameleon [Xu et al., 2023a] Link
Undercover [Xu et al., 2023a] Link
Ultimatum Game TE [Aher et al., 2023] Link
Psychology Garden Path TE [Aher et al., 2023] Link
Wisdom of Crowds TE [Aher et al., 2023] Link
MovieLens-1M [Zhang et al., 2023a] Link
Recommender System [Zhang et al., 2023e]
Amazon review dataset /
Policy Making Board Connectivity Evaluation [Hua et al., 2023] Link

Table 2: Datasets and Benchmarks commonly used in LLM-MA studies. “ / ” denotes the unavailability of data link.

tion paradox in an information marketplace, reveals improved causal relationships in recommendation tasks. In Agent4Rec
decision-making and answer quality when agents temporar- work, agents are used to simulate users and they do not com-
ily access information before purchase. [Li et al., 2023g] municate with each other. Different from Agent4Rec work,
presents an LLM-MA framework for financial trading, em- [Zhang et al., 2023e] treats both users and items as agents,
phasizing a layered memory system, debate mechanisms, and optimizing them collectively to reflect and adjust to real-
individualized trading characters, thereby fortifying decision- world interaction disparities. This work emphasizes simulat-
making robustness. [Zhao et al., 2023] utilizes LLM-based ing user-item interactions and propagates preferences among
Agents to simulate a virtual town with restaurant and cus- agents, capturing the collaborative filtering essence.
tomer agents, yielding insights aligned with sociological and
economic theories. These studies collectively illuminate the
4.2.6 Policy Making
broad spectrum of applications and advancements in employ-
ing LLMs for diverse economic simulation scenarios. Similar to simulations in gaming and economic scenarios,
Policy Making requires strong decision-making capabilities
4.2.5 Recommender Systems to realistic and dynamic complex problems. LLM-MA can
The use of the LLM-MA in recommender systems is similar be used to simulate the policy making via simulating a virtual
to that in psychology since studies in both fields involve the government or simulating the impact of various policies on
consideration of extrinsic and intrinsic human factors such as different communities. These simulations provide valuable
cognitive processes and personality [Lex and Schedl, 2022]. insights into how policies are formulated and their potential
One way to use LLM-MA in recommender systems is to di- effects, aiding policymakers in understanding and anticipat-
rectly introduce items to multiple LLM-based agents within ing the consequences of their decisions [Farmer and Axtell,
diverse traits and conduct statistics of the preferences of dif- 2022]. The research outlined in [Xiao et al., 2023] is cen-
ferent agents. Another way is to treat both users and items tered on simulating a township water pollution crisis. It sim-
as agents and the user-item communication as interactions, ulated a town located on an island including a demographic
simulating the preference propagation. To bridge the gap be- structure of different agents and township head and advisor.
tween offline metrics and real-world performance in recom- Within the water pollution crisis simulation, this work pro-
mendation systems, Agent4Rec [Zhang et al., 2023a] intro- vides an in-depth analysis of how a virtual government entity
duces a simulation platform based on LLM-MA. 1000 gener- might respond to such a public administration challenge and
ative agents are initialized with the MovieLens-1M dataset to how information transfer in the social network in this crisis.
simulate complex user interactions in a recommendation en- [Hua et al., 2023] introduces WarAgent to simulate key his-
vironment. Agent4Rec shows that LLM-MA can effectively torical conflicts and provides insights for conflict resolution
mimic real user preferences and behaviors, provide insights and understanding, with potential applications in preventing
into phenomena like the filter bubble effect, and help uncover future international conflicts.
4.2.7 Disease Propagation Simulation search applications use different datasets and benchmarks.
Leveraging the societal simulation capabilities of LLM-MA In the Problem solving scenarios, most datasets and bench-
can also be used to simulate disease propagation. The most marks are used to evaluate the planning and reasoning capa-
recent study in [Williams et al., 2023] delves into the use of bilities by Multiple agents cooperation or debate. In World
LLM-MA in simulating disease spread. The research show- Simulation scenarios, datasets and benchmarks are used to
cases through various simulations how these LLM-based evaluate the alignment between the simulated world and real-
agents can accurately emulate human responses to disease world or analyze the behaviors of different agents. However,
outbreaks, including behaviors like self-quarantine and iso- in certain research applications like Science Team operations
lation during heightened case numbers. The collective be- for experiments and economic modeling, there is still a need
havior of these agents mirrors the complex patterns of multi- for comprehensive benchmarks. The development of such
ple waves typically seen in pandemics, eventually stabilizing benchmarks would greatly enhance the ability to gauge the
into an endemic state. Impressively, their actions contribute success and applicability of LLM-MA in these complex and
to the attenuation of the epidemic curve. [Ghaffarzadegan et dynamic fields.
al., 2023] also discusses the epidemic propagation simulation
and decomposes the simulation into two parts: the Mechanis- 6 Challenges and Opportunities
tic Model which represents the information or propagation of
the virus and the Decision-Making Model which represents Studies of LLM-MA frameworks and applications are ad-
the agents’ decision-making process when facing the virus. vancing rapidly, giving rise to numerous challenges and op-
portunities. We identified several critical challenges and po-
tential areas for future study.
5 Implementation Tools and Resources
5.1 Multi-Agents Framework 6.1 Advancing into Multi-Modal Environment
We provide a detailed introduction to three open-source Most previous work on LLM-MA has been focused on text-
multi-agent frameworks: MetaGPT [Hong et al., 2023], based environments, excelling in processing and generating
CAMEL [Li et al., 2023b], and Autogen [Wu et al., 2023a]. text. However, there is a notable lack in multi-modal set-
They are all frameworks that utilize language models for tings, where agents would interact with and interpret data
complex task-solving with a focus on multi-agent collabora- from multiple sensory inputs and generate multiple outputs
tion, but they differ in their approaches and applications. such as images, audio, video, and physical actions. Inte-
MetaGPT is designed to embed human workflow processes grating LLMs into multi-modal environments presents addi-
into the operation of language model agents, thereby reducing tional challenges, such as processing diverse data types and
the hallucination problem that often arises in complex tasks. enabling agents to understand each other and respond to more
It does this by encoding Standard Operating Procedures into than just textual information.
the system and using an assembly line approach to assign spe-
cific roles to different agents. 6.2 Addressing Hallucination
CAMEL, or Communicative Agent Framework, is oriented The hallucination problem is a significant challenge in LLMs
towards facilitating autonomous cooperation among agents. and single LLM-based Agent systems. It refers to the phe-
It uses a novel technique called inception prompting to guide nomenon where the model generates text that is factually in-
conversational agents towards fulfilling tasks that are consis- correct [Huang et al., 2023b]. However, this problem takes
tent with human objectives. This framework also serves as a on an added layer of complexity in a multi-agent setting. In
tool for generating and studying conversational data, help- such scenarios, one agent’s hallucination can have a cascad-
ing researchers understand how communicative agents be- ing effect. This is due to the interconnected nature of multi-
have and interact. agent systems, where misinformation from one agent can be
AutoGen is a versatile framework that allows for the cre- accepted and further propagated by others in the network.
ation of applications using language models. It is distinctive Therefore, detecting and mitigating hallucinations in LLM-
for its high level of customization, enabling developers to pro- MA is not just a crucial task but also presents a unique set
gram agents using both natural language and code to define of challenges. It involves not only correcting inaccuracies at
how these agents interact. This versatility enables its use in the level of individual agents but also managing the flow of
diverse fields, from technical areas such as coding and math- information between agents to prevent the spread of these in-
ematics to consumer-focused sectors like entertainment. accuracies throughout the system.
More recently, [Chen et al., 2023c; Chen et al., 2023a]
introduce frameworks for dynamic multi-agent collabora- 6.3 Acquiring Collective Intelligence
tion, while [Zhou et al., 2023a; Li et al., 2023h; Xie et In traditional multi-agent systems, agents often use reinforce-
al., 2023] present platforms and libraries for building au- ment learning to learn from offline training datasets. How-
tonomous agents, emphasizing their adaptability in task- ever, LLM-MA systems mainly learn from instant feedback,
solving and social simulations. such as interactions with the environment or humans, as we
discussed in Section 3. This learning style requires a reli-
5.2 Datasets and Benchmarks able interactive environment and it would be tricky to design
We summarize commonly used datasets or benchmarks for such an interactive environment for many tasks, limiting the
LLM-MA study in Table 2. We observe that different re- scalability of LLM-MA systems. Moreover, the prevailing
approaches in current research involve employing Memory 6.6 Applications and Beyond
and Self-Evolution techniques to adjust agents based on feed- The potential of LLM-MA systems extends far beyond their
back. While effective for individual agents, these methods do current applications, holding great promise for advanced
not fully capitalize on the potential collective intelligence of computational problem-solving in fields such as finance, edu-
the agent network. They adjust agents in isolation, overlook- cation, healthcare, environmental science, urban planning and
ing the synergistic effects that can emerge from coordinated so on. As we have discussed, LLM-MA systems possess the
multi-agent interactions. Hence, jointly adjusting multiple capability to tackle complex problems and simulate various
agents and achieving optimal collective intelligence is still a aspects of the real world. While the current role-playing ca-
critical challenge for LLM-MA. pabilities of LLMs may have limitations, ongoing advance-
ments in LLM technology suggest a bright future. It is an-
6.4 Scaling Up LLM-MA Systems ticipated to have more sophisticated methodologies, applica-
LLM-MA systems are composed of a number of individual tions, datasets, and benchmarks tailored for diverse research
LLM-based agents, posing a significant challenge of scala- fields. Furthermore, there are opportunities to explore LLM-
bility regarding the number of agents. From the computa- MA systems from various theoretical perspectives, such as
tional complexity perspective, each LLM-based agent, typ- Cognitive Science [Sumers et al., 2023], Symbolic Artificial
ically built on large language models like GPT-4, demands Intelligence, Cybernetics, Complex Systems, and Collective
substantial computational power and memory. Scaling up the Intelligence. Such a multi-faceted approach could contribute
number of these agents in an LLM-MA system significantly to a more comprehensive understanding and innovative appli-
increases resource requirements. In scenarios with limited cations in this rapidly evolving field.
computational resource, it would be challenging to develop
these LLM-MA systems. 7 Conclusion
Additionally, as the number of agents in an LLM-MA sys- LLM-based Multi-Agents have shown inspiring collective in-
tem increases, additional complexities and research opportu- telligence and rapidly garnered increasing interest among re-
nities emerge, particularly in areas like efficient agent coor- searchers. In this survey, we first systematically review the
dination, communication, and understanding the scaling laws development of LLM-MA systems by positioning, differen-
of multi-agents. For instance, with more LLM-based agents, tiating, and connecting them from various aspects, regard-
the intricacy of ensuring effective coordination and commu- ing the agents-environment interface, the characterization of
nication rises significantly. As highlighted in [Dibia, 2023], agents by LLMs, the strategies for managing agent communi-
designing advanced Agents Orchestration methodologies is cation and the paradigms for capability acquisition. We also
increasingly important. These methodologies aim to opti- summarized LLM-MA applications for problem-solving and
mize agents workflows, task assignments tailored to differ- world simulation. By also highlighting the commonly used
ent agents, and communication patterns across agents such as datasets and benchmarks and discussing challenges and fu-
communication constraints between agents. Effective Agents ture opportunities, we hope that this survey can serve as a use-
Orchestration facilitates harmonious operation among agents, ful resource for researchers across various research fields, in-
minimizing conflicts and redundancies. Additionally, explor- spiring future research to explore the potential of LLM-based
ing and defining the scaling laws that govern the behavior and Multi-Agents.
efficiency of multi-agent systems as they grow larger remains
an important area of research. These aspects highlight the References
need for innovative solutions to optimize LLM-MA systems, [Agashe et al., 2023] Saaket Agashe, Yue Fan, and Xin Eric
making them both effective and resource-efficient. Wang. Evaluating multi-agent coordination abilities in
large language models, 2023.
6.5 Evaluation and Benchmarks
[Aher et al., 2023] Gati Aher, Rosa I. Arriaga, and
We have summarized the datasets and benchmarks currently Adam Tauman Kalai. Using large language models
available for LLM-MA in Table 2. This is a starting point, and to simulate multiple humans and replicate human subject
far from being comprehensive. We identify two significant studies, 2023.
challenges in evaluating LLM-MA systems and benchmark- [Akata et al., 2023] Elif Akata, Lion Schulz, Julian Coda-
ing their performance against each other. Firstly, as discussed Forno, Seong Joon Oh, Matthias Bethge, and Eric Schulz.
in [Xu et al., 2023a], much of the existing research focuses Playing repeated games with large language models. arXiv
on evaluating individual agents’ understanding and reason- preprint arXiv:2305.16867, 2023.
ing within narrowly defined scenarios. This focus tends to
overlook the broader and more complex emergent behaviors [Anonymous, 2023] Anonymous. Rethinking the buyer’s in-
that are integral to multi-agent systems. Secondly, there is a spection paradox in information markets with language
notable shortfall in the development of comprehensive bench- agents. In Submitted to The Twelfth International Con-
marks across several research domains, such as Science Team ference on Learning Representations, 2023. under review.
for Experiment Operations, Economic analysis, and Disease [Chan et al., 2023] Chi-Min Chan, Weize Chen, Yusheng
propagation simulation. This gap presents an obstacle to ac- Su, Jianxuan Yu, Wei Xue, Shanghang Zhang, Jie Fu, and
curately assessing and benchmarking the full capabilities of Zhiyuan Liu. Chateval: Towards better llm-based evalua-
LLM-MA systems in these varied and crucial fields. tors through multi-agent debate, 2023.
[Chen et al., 2023a] Guangyao Chen, Siwei Dong, Yu Shu, large language model-empowered agents. arXiv preprint
Ge Zhang, Jaward Sesay, Börje F Karlsson, Jie Fu, and arXiv:2307.14984, 2023.
Yemin Shi. Autoagents: A framework for automatic agent [Gao et al., 2023b] Yunfan Gao, Yun Xiong, Xinyu Gao,
generation. arXiv preprint arXiv:2309.17288, 2023.
Kangxiang Jia, Jinliu Pan, Yuxi Bi, Yi Dai, Jiawei
[Chen et al., 2023b] Huaben Chen, Wenkang Ji, Lufeng Xu, Sun, and Haofen Wang. Retrieval-augmented generation
and Shiyu Zhao. Multi-agent consensus seeking via large for large language models: A survey. arXiv preprint
language models. arXiv preprint arXiv:2310.20151, 2023. arXiv:2312.10997, 2023.
[Chen et al., 2023c] Weize Chen, Yusheng Su, Jingwei Zuo, [Geva et al., 2021] Mor Geva, Daniel Khashabi, Elad Segal,
Cheng Yang, Chenfei Yuan, Chen Qian, Chi-Min Chan, Tushar Khot, Dan Roth, and Jonathan Berant. Did aris-
Yujia Qin, Yaxi Lu, Ruobing Xie, et al. Agentverse: Facil- totle use a laptop? a question answering benchmark with
itating multi-agent collaboration and exploring emergent implicit reasoning strategies, 2021.
behaviors in agents. arXiv preprint arXiv:2308.10848,
2023. [Ghaffarzadegan et al., 2023] Navid Ghaffarzadegan, Aritra
Majumdar, Ross Williams, and Niyousha Hosseinichimeh.
[Chen et al., 2023d] Yongchao Chen, Jacob Arkin, Yang Generative agent-based modeling: Unveiling social sys-
Zhang, Nicholas Roy, and Chuchu Fan. Scalable multi- tem dynamics through coupling mechanistic models
robot collaboration with large language models: Cen- with generative artificial intelligence. arXiv preprint
tralized or decentralized systems? arXiv preprint arXiv:2309.11456, 2023.
arXiv:2309.15943, 2023.
[Gong et al., 2023] Ran Gong, Qiuyuan Huang, Xiaojian
[Cobbe et al., 2021] Karl Cobbe, Vineet Kosaraju, Moham-
Ma, Hoi Vo, Zane Durante, Yusuke Noda, Zilong Zheng,
mad Bavarian, Mark Chen, Heewoo Jun, Lukasz Kaiser,
Song-Chun Zhu, Demetri Terzopoulos, Li Fei-Fei, et al.
Matthias Plappert, Jerry Tworek, Jacob Hilton, Reiichiro
Mindagent: Emergent gaming interaction. arXiv preprint
Nakano, et al. Training verifiers to solve math word prob-
arXiv:2309.09971, 2023.
lems. arXiv preprint arXiv:2110.14168, 2021.
[Dasgupta et al., 2023] Ishita Dasgupta, Christine Kaeser- [Guo et al., 2023] Taicheng Guo, Kehan Guo, Zhengwen
Chen, Kenneth Marino, Arun Ahuja, Sheila Babayan, Liang, Zhichun Guo, Nitesh V Chawla, Olaf Wiest, Xi-
Felix Hill, and Rob Fergus. Collaborating with lan- angliang Zhang, et al. What indeed can gpt models do
guage models for embodied reasoning. arXiv preprint in chemistry? a comprehensive benchmark on eight tasks.
arXiv:2302.00763, 2023. arXiv preprint arXiv:2305.18365, 2023.
[Dibia, 2023] Victor Dibia. Multi-agent llm applica- [Hendrycks et al., 2020] Dan Hendrycks, Collin Burns,
tions — a review of current research, tools, and Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song,
challenges. https://newsletter.victordibia.com/p/ and Jacob Steinhardt. Measuring massive multitask lan-
multi-agent-llm-applications-a-review, 2023. guage understanding. arXiv preprint arXiv:2009.03300,
2020.
[Dong et al., 2023a] Qingxiu Dong, Lei Li, Damai Dai,
Ce Zheng, Zhiyong Wu, Baobao Chang, Xu Sun, Jingjing [Hong et al., 2023] Sirui Hong, Xiawu Zheng, Jonathan
Xu, Lei Li, and Zhifang Sui. A survey on in-context learn- Chen, Yuheng Cheng, Ceyao Zhang, Zili Wang, Steven
ing, 2023. Ka Shing Yau, Zijuan Lin, Liyang Zhou, Chenyu Ran,
[Dong et al., 2023b] Yihong Dong, Xue Jiang, Zhi Jin, and et al. Metagpt: Meta programming for multi-agent col-
laborative framework. arXiv preprint arXiv:2308.00352,
Ge Li. Self-collaboration code generation via chatgpt,
2023.
2023.
[Du et al., 2023] Yilun Du, Shuang Li, Antonio Torralba, [Horton, 2023] John J Horton. Large language models as
Joshua B. Tenenbaum, and Igor Mordatch. Improving fac- simulated economic agents: What can we learn from homo
tuality and reasoning in language models through multia- silicus? Technical report, National Bureau of Economic
gent debate, 2023. Research, 2023.
[Fan et al., 2023] Caoyun Fan, Jindou Chen, Yaohui Jin, and [Hua et al., 2023] Wenyue Hua, Lizhou Fan, Lingyao Li,
Hao He. Can large language models serve as rational play- Kai Mei, Jianchao Ji, Yingqiang Ge, Libby Hemphill, and
ers in game theory? a systematic analysis. arXiv preprint Yongfeng Zhang. War and peace (waragent): Large lan-
arXiv:2312.05488, 2023. guage model-based multi-agent simulation of world wars,
2023.
[Farmer and Axtell, 2022] J. Doyne Farmer and Robert L.
Axtell. Agent-Based Modeling in Economics and Finance: [Huang et al., 2023a] Dong Huang, Qingwen Bu, Jie M.
Past, Present, and Future. INET Oxford Working Papers Zhang, Michael Luck, and Heming Cui. Agentcoder:
2022-10, Institute for New Economic Thinking at the Ox- Multi-agent-based code generation with iterative testing
ford Martin School, University of Oxford, June 2022. and optimisation, 2023.
[Gao et al., 2023a] Chen Gao, Xiaochong Lan, Zhihong Lu, [Huang et al., 2023b] Lei Huang, Weijiang Yu, Weitao Ma,
Jinzhu Mao, Jinghua Piao, Huandong Wang, Depeng Jin, Weihong Zhong, Zhangyin Feng, Haotian Wang, Qiang-
and Yong Li. S3 : Social-network simulation system with long Chen, Weihua Peng, Xiaocheng Feng, Bing Qin, et al.
 A survey on hallucination in large language models: Prin- [Li et al., 2023g] Yang Li, Yangyang Yu, Haohang Li, Zhi
ciples, taxonomy, challenges, and open questions. arXiv Chen, and Khaldoun Khashanah. Tradinggpt: Multi-agent
preprint arXiv:2311.05232, 2023. system with layered memory and distinct characters for
[Kaiya et al., 2023] Zhao Kaiya, Michelangelo Naim, Jo- enhanced financial trading performance, 2023.
vana Kondic, Manuel Cortes, Jiaxin Ge, Shuying Luo, [Li et al., 2023h] Yuan Li, Yixuan Zhang, and Lichao Sun.
Guangyu Robert Yang, and Andrew Ahn. Lyfe agents: Metaagents: Simulating interactions of human behaviors
Generative agents for low-cost real-time social interac- for llm-based task-oriented coordination via collaborative
tions. arXiv preprint arXiv:2310.02172, 2023. generative agents. arXiv preprint arXiv:2310.06500, 2023.
[Khot et al., 2023] Tushar Khot, Harsh Trivedi, Matthew [Liang et al., 2023] Zhenwen Liang, Wenhao Yu, Tanmay
Finlayson, Yao Fu, Kyle Richardson, Peter Clark, and Rajpurohit, Peter Clark, Xiangliang Zhang, and Ashwin
Ashish Sabharwal. Decomposed prompting: A modular Kaylan. Let gpt be a math tutor: Teaching math word prob-
approach for solving complex tasks, 2023. lem solvers with customized exercise generation. arXiv
preprint arXiv:2305.14386, 2023.
[Kovač et al., 2023] Grgur Kovač, Rémy Portelas, Peter Ford
Dominey, and Pierre-Yves Oudeyer. The socialai school: [Light et al., 2023a] Jonathan Light, Min Cai, Sheng Shen,
Insights from developmental psychology towards artificial and Ziniu Hu. Avalonbench: Evaluating llms playing the
socio-cultural agents. arXiv preprint arXiv:2307.07871, game of avalon, 2023.
2023. [Light et al., 2023b] Jonathan Light, Min Cai, Sheng Shen,
[Lewis et al., 2021] Patrick Lewis, Ethan Perez, Aleksan- and Ziniu Hu. From text to tactic: Evaluating llms play-
ing the game of avalon. arXiv preprint arXiv:2310.05036,
dra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman
2023.
Goyal, Heinrich Küttler, Mike Lewis, Wen tau Yih,
Tim Rocktäschel, Sebastian Riedel, and Douwe Kiela. [Liu et al., 2023] Zijun Liu, Yanzhe Zhang, Peng Li, Yang
Retrieval-augmented generation for knowledge-intensive Liu, and Diyi Yang. Dynamic llm-agent network: An llm-
nlp tasks, 2021. agent collaboration framework with agent team optimiza-
tion. arXiv preprint arXiv:2310.02170, 2023.
[Lex and Schedl, 2022] Elisabeth Lex and Markus Schedl.
Psychology-informed recommender systems: A human- [Ma et al., 2023] Zilin Ma, Yiyang Mei, and Zhaoyuan Su.
centric perspective on recommender systems. In Proceed- Understanding the benefits and challenges of using large
ings of the 2022 Conference on Human Information In- language model-based conversational agents for mental
teraction and Retrieval, CHIIR ’22, page 367–368, New well-being support. arXiv preprint arXiv:2307.15810,
York, NY, USA, 2022. Association for Computing Ma- 2023.
chinery. [Mandi et al., 2023] Zhao Mandi, Shreeya Jain, and Shuran
[Li et al., 2023a] Chao Li, Xing Su, Chao Fan, Haoying Han, Song. Roco: Dialectic multi-robot collaboration with large
Cong Xue, and Chunmo Zheng. Quantifying the impact language models. arXiv preprint arXiv:2307.04738, 2023.
of large language models on collective opinion dynamics. [Mao et al., 2023] Shaoguang Mao, Yuzhe Cai, Yan Xia,
arXiv preprint arXiv:2308.03313, 2023. Wenshan Wu, Xun Wang, Fengyi Wang, Tao Ge, and Furu
[Li et al., 2023b] Guohao Li, Hasan Abed Al Kader Ham- Wei. Alympics: Language agents meet game theory. arXiv
preprint arXiv:2311.03220, 2023.
moud, Hani Itani, Dmitrii Khizbullin, and Bernard
Ghanem. Camel: Communicative agents for” mind” ex- [Moura, 2023] João Moura. Crewai. https://github.com/
ploration of large scale language model society. arXiv joaomdmoura/crewAI, 2023.
preprint arXiv:2303.17760, 2023. [Mukobi et al., 2023] Gabriel Mukobi, Hannah Erlebach,
[Li et al., 2023c] Huao Li, Yu Quan Chong, Simon Stepput- Niklas Lauffer, Lewis Hammond, Alan Chan, and Jesse
tis, Joseph Campbell, Dana Hughes, Michael Lewis, and Clifton. Welfare diplomacy: Benchmarking language
Katia Sycara. Theory of mind for multi-agent collabora- model cooperation. arXiv preprint arXiv:2310.08901,
tion via large language models, 2023. 2023.
[Li et al., 2023d] Minghao Li, Yingxiu Zhao, Bowen Yu, [Nascimento et al., 2023] Nathalia Nascimento, Paulo Alen-
Feifan Song, Hangyu Li, Haiyang Yu, Zhoujun Li, Fei car, and Donald Cowan. Self-adaptive large language
Huang, and Yongbin Li. Api-bank: A comprehensive model (llm)-based multiagent systems. In 2023 IEEE
benchmark for tool-augmented llms, 2023. International Conference on Autonomic Computing and
Self-Organizing Systems Companion (ACSOS-C), pages
[Li et al., 2023e] Nian Li, Chen Gao, Yong Li, and Qingmin 104–109. IEEE, 2023.
Liao. Large language model-empowered agents for simu- [Park et al., 2022] Joon Sung Park, Lindsay Popowski, Car-
lating macroeconomic activities, 2023.
rie Cai, Meredith Ringel Morris, Percy Liang, and
[Li et al., 2023f] Siyu Li, Jin Yang, and Kui Zhao. Are you Michael S Bernstein. Social simulacra: Creating popu-
in a masquerade? exploring the behavior and impact of lated prototypes for social computing systems. In Pro-
large language model driven social bots in online social ceedings of the 35th Annual ACM Symposium on User In-
networks. arXiv preprint arXiv:2307.10337, 2023. terface Software and Technology, pages 1–18, 2022.
[Park et al., 2023] Joon Sung Park, Joseph C O’Brien, Car- [Williams et al., 2023] Ross Williams, Niyousha Hos-
rie J Cai, Meredith Ringel Morris, Percy Liang, and seinichimeh, Aritra Majumdar, and Navid Ghaffarzade-
Michael S Bernstein. Generative agents: Interac- gan. Epidemic modeling with generative agents. arXiv
tive simulacra of human behavior. arXiv preprint preprint arXiv:2307.04986, 2023.
arXiv:2304.03442, 2023. [Wooldridge and Jennings, 1995] Michael Wooldridge and
[Qian et al., 2023] Chen Qian, Xin Cong, Wei Liu, Cheng Nicholas R. Jennings. Intelligent agents: theory and prac-
Yang, Weize Chen, Yusheng Su, Yufan Dang, Jiahao Li, tice. The Knowledge Engineering Review, 10:115 – 152,
Juyuan Xu, Dahai Li, Zhiyuan Liu, and Maosong Sun. 1995.
Communicative agents for software development, 2023. [Wu et al., 2023a] Qingyun Wu, Gagan Bansal, Jieyu Zhang,
[Ruan et al., 2023] Jingqing Ruan, Yihong Chen, Bin Zhang, Yiran Wu, Shaokun Zhang, Erkang Zhu, Beibin Li,
Zhiwei Xu, Tianpeng Bao, Guoqing Du, Shiwei Shi, Li Jiang, Xiaoyun Zhang, and Chi Wang. Autogen: En-
Hangyu Mao, Ziyue Li, Xingyu Zeng, and Rui Zhao. Tptu: abling next-gen llm applications via multi-agent conversa-
Large language model-based ai agents for task planning tion framework. arXiv preprint arXiv:2308.08155, 2023.
and tool usage, 2023. [Wu et al., 2023b] Yuxiang Wu, Zhengyao Jiang, Akbir
[Russell and Norvig, 2009] Stuart Russell and Peter Norvig. Khan, Yao Fu, Laura Ruis, Edward Grefenstette, and Tim
Artificial Intelligence: A Modern Approach. Prentice Hall Rocktäschel. Chatarena: Multi-agent language game en-
Press, USA, 3rd edition, 2009. vironments for large language models. GitHub repository,
2023.
[Shinn et al., 2023] Noah Shinn, Federico Cassano, Edward
[Xi et al., 2023] Zhiheng Xi, Wenxiang Chen, Xin Guo,
Berman, Ashwin Gopinath, Karthik Narasimhan, and
Shunyu Yao. Reflexion: Language agents with verbal re- Wei He, Yiwen Ding, Boyang Hong, Ming Zhang, Jun-
inforcement learning, 2023. zhe Wang, Senjie Jin, Enyu Zhou, Rui Zheng, Xiaoran
Fan, Xiao Wang, Limao Xiong, Yuhao Zhou, Weiran
[Sumers et al., 2023] Theodore R Sumers, Shunyu Yao, Wang, Changhao Jiang, Yicheng Zou, Xiangyang Liu,
Karthik Narasimhan, and Thomas L Griffiths. Cogni- Zhangyue Yin, Shihan Dou, Rongxiang Weng, Wensen
tive architectures for language agents. arXiv preprint Cheng, Qi Zhang, Wenjuan Qin, Yongyan Zheng, Xipeng
arXiv:2309.02427, 2023. Qiu, Xuanjing Huang, and Tao Gui. The rise and potential
[Tang et al., 2023] Xiangru Tang, Anni Zou, Zhuosheng of large language model based agents: A survey, 2023.
Zhang, Yilun Zhao, Xingyao Zhang, Arman Cohan, and [Xiao et al., 2023] Bushi Xiao, Ziyuan Yin, and Zixuan
Mark Gerstein. Medagents: Large language models as col- Shan. Simulating public administration crisis: A novel
laborators for zero-shot medical reasoning, 2023. generative agent-based simulation system to lower tech-
[Wang et al., 2021] Zijie J. Wang, Dongjin Choi, Shenyu nology barriers in social science research. arXiv preprint
arXiv:2311.06957, 2023.
Xu, and Diyi Yang. Putting humans in the natural lan-
guage processing loop: A survey, 2021. [Xie et al., 2023] Tianbao Xie, Fan Zhou, Zhoujun Cheng,
Peng Shi, Luoxuan Weng, Yitao Liu, Toh Jing Hua, Jun-
[Wang et al., 2023a] Kuan Wang, Yadong Lu, Michael San-
ning Zhao, Qian Liu, Che Liu, et al. Openagents: An open
tacroce, Yeyun Gong, Chao Zhang, and Yelong Shen. platform for language agents in the wild. arXiv preprint
Adapting llm agents through communication, 2023. arXiv:2310.10634, 2023.
[Wang et al., 2023b] Lei Wang, Chen Ma, Xueyang Feng, [Xiong et al., 2023] Kai Xiong, Xiao Ding, Yixin Cao, Ting
Zeyu Zhang, Hao Yang, Jingsen Zhang, Zhiyuan Chen, Ji- Liu, and Bing Qin. Examining inter-consistency of large
akai Tang, Xu Chen, Yankai Lin, Wayne Xin Zhao, Zhewei language models collaboration: An in-depth analysis via
Wei, and Ji-Rong Wen. A survey on large language model debate, 2023.
based autonomous agents, 2023.
[Xu et al., 2023a] Lin Xu, Zhiyuan Hu, Daquan Zhou,
[Wang et al., 2023c] Shenzhi Wang, Chang Liu, Zilong Hongyu Ren, Zhen Dong, Kurt Keutzer, See Kiong Ng,
Zheng, Siyuan Qi, Shuo Chen, Qisen Yang, Andrew Zhao, and Jiashi Feng. Magic: Investigation of large language
Chaofei Wang, Shiji Song, and Gao Huang. Avalon’s game model powered multi-agent in cognition, adaptability, ra-
of thoughts: Battle against deception through recursive tionality and collaboration, 2023.
contemplation. arXiv preprint arXiv:2310.01320, 2023.
[Xu et al., 2023b] Yuzhuang Xu, Shuo Wang, Peng Li,
[Wei et al., 2022] Jason Wei, Xuezhi Wang, Dale Schuur- Fuwen Luo, Xiaolong Wang, Weidong Liu, and Yang
mans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Liu. Exploring large language models for communication
Denny Zhou, et al. Chain-of-thought prompting elicits games: An empirical study on werewolf. arXiv preprint
reasoning in large language models. Advances in Neural arXiv:2309.04658, 2023.
Information Processing Systems, 35:24824–24837, 2022. [Xu et al., 2023c] Zelai Xu, Chao Yu, Fei Fang, Yu Wang,
[Weng, 2023] Lilian Weng. Llm powered au- and Yi Wu. Language agents with reinforcement learning
tonomous agents. https://lilianweng.github.io/posts/ for strategic play in the werewolf game. arXiv preprint
2023-06-23-agent/, 2023. arXiv:2310.18940, 2023.
[Yao et al., 2023] Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak
Shafran, Thomas L. Griffiths, Yuan Cao, and Karthik
Narasimhan. Tree of thoughts: Deliberate problem solving
with large language models, 2023.
[Yu et al., 2023] Bangguo Yu, Hamidreza Kasaei, and Ming
Cao. Co-navgpt: Multi-robot cooperative visual semantic
navigation using large language models, 2023.
[Zhang et al., 2023a] An Zhang, Leheng Sheng, Yuxin
Chen, Hao Li, Yang Deng, Xiang Wang, and Tat-Seng
Chua. On generative agents in recommendation, 2023.
[Zhang et al., 2023b] Ceyao Zhang, Kaijie Yang, Siyi Hu,
Zihao Wang, Guanghe Li, Yihang Sun, Cheng Zhang,
Zhaowei Zhang, Anji Liu, Song-Chun Zhu, et al. Proa-
gent: Building proactive cooperative ai with large lan-
guage models. arXiv preprint arXiv:2308.11339, 2023.
[Zhang et al., 2023c] Hongxin Zhang, Weihua Du, Jiaming
Shan, Qinhong Zhou, Yilun Du, Joshua B Tenenbaum,
Tianmin Shu, and Chuang Gan. Building cooperative
embodied agents modularly with large language models.
arXiv preprint arXiv:2307.02485, 2023.
[Zhang et al., 2023d] Jintian Zhang, Xin Xu, and Shumin
Deng. Exploring collaboration mechanisms for llm agents:
A social psychology view, 2023.
[Zhang et al., 2023e] Junjie Zhang, Yupeng Hou, Ruobing
Xie, Wenqi Sun, Julian McAuley, Wayne Xin Zhao, Leyu
Lin, and Ji-Rong Wen. Agentcf: Collaborative learning
with autonomous language agents for recommender sys-
tems, 2023.
[Zhao et al., 2023] Qinlin Zhao, Jindong Wang, Yixuan
Zhang, Yiqiao Jin, Kaijie Zhu, Hao Chen, and Xing Xie.
Competeai: Understanding the competition behaviors in
large language model-based agents, 2023.
[Zheng et al., 2023] Zhiling Zheng, Oufan Zhang, Ha L.
Nguyen, Nakul Rampal, Ali H. Alawadhi, Zichao Rong,
Teresa Head-Gordon, Christian Borgs, Jennifer T. Chayes,
and Omar M. Yaghi. Chatgpt research group for optimiz-
ing the crystallinity of mofs and cofs. ACS Central Sci-
ence, 9(11):2161–2170, 2023.
[Zhou et al., 2023a] Wangchunshu Zhou, Yuchen Eleanor
Jiang, Long Li, Jialong Wu, Tiannan Wang, Shi Qiu, Jin-
tian Zhang, Jing Chen, Ruipu Wu, Shuai Wang, et al.
Agents: An open-source framework for autonomous lan-
guage agents. arXiv preprint arXiv:2309.07870, 2023.
[Zhou et al., 2023b] Xuhui Zhou, Hao Zhu, Leena Mathur,
Ruohong Zhang, Haofei Yu, Zhengyang Qi, Louis-
Philippe Morency, Yonatan Bisk, Daniel Fried, Graham
Neubig, and Maarten Sap. Sotopia: Interactive evaluation
for social intelligence in language agents, 2023.
[Ziems et al., 2023] Caleb Ziems, Omar Shaikh, Zhehao
Zhang, William Held, Jiaao Chen, and Diyi Yang. Can
large language models transform computational social sci-
ence? Computational Linguistics, pages 1–53, 2023.