Back

A survey on large language model based autonomous agents

Lei WangChengbang MaXueyang FengZeyu ZhangHao-ran YangJingsen ZhangZhi-Yang ChenJiakai TangXu ChenYankai Lin

article2023Frontiers of Computer Science3,843 citations

Establishes a unified framework for constructing large language model-based autonomous agents while systematically reviewing their applications across science and engineering, current evaluation methods, and open research challenges.

Listen

Artificial intelligence research is shifting rapidly from isolated, rule-based systems toward autonomous agents capable of operating in complex, dynamic, real-world environments. Traditional reinforcement learning agents struggle to replicate human-like decision-making because they lack broad background knowledge and struggle in open-ended domains. The rapid rise of large language models provides an opportunity to overcome these hurdles by leveraging extensive world knowledge and natural language comprehension. The article evaluates how to construct, apply, and benchmark autonomous agents powered by large language models, establishing a unified foundation for researchers and decision-makers navigating this emerging technology.

To address this objective, the article conducts a systematic literature review analyzing over one hundred recent studies. It synthesizes these works to create comprehensive taxonomies across three core dimensions: architectural construction and capability acquisition, practical applications across diverse fields, and subjective and objective evaluation strategies.

Key findings show that effective autonomous agents depend on a unified four-part architecture: a profiling module to establish identity, a hybrid memory module combining immediate context and long-term storage, a planning module that breaks down complex tasks, and an action module that executes decisions directly or via external tools. The synthesis also finds that agents improve performance through two main pathways: model fine-tuning on annotated or real-world data, and engineering strategies such as structured prompt design, multi-agent debate, and iterative trial-and-error without model fine-tuning. Across applications, language model agents demonstrate substantial value in social science simulations, experimental planning in natural sciences, and automated software development and robotics. Finally, evaluating these agents requires combining subjective human review with structured, objective benchmarks that measure goal completion rates, behavioral similarity to humans, and operational efficiency.

These findings suggest that autonomous agents can automate complex end-to-end workflows and safely simulate human behaviors, potentially lowering development costs and accelerating research across scientific and engineering disciplines. However, several operational and safety risks remain, including model hallucinations producing incorrect code or guidance, prompt brittleness across different underlying models, and high computational latency caused by iterative reasoning loops.

Organizations evaluating or deploying language model agents should prioritize architectures that incorporate closed-loop feedback and tool integration to catch and correct execution errors automatically. For complex simulations and high-stakes workflows, practitioners should run targeted pilot programs combining automated benchmarks with human oversight to validate accuracy, alignment, and safety before broad deployment.

Because the field remains in its early stages, confidence in agent autonomy should be tempered by existing limitations, including context window constraints, unpredictable role-playing fidelity, and inference latency. Decision-makers should treat these systems as assistive tools requiring ongoing monitoring until more robust prompt standards, alignment controls, and domain-specific benchmarks are established.

Renmin University of Chinaautonomous agentsengineeringlanguage modelAgentAgentsarXiv
arXiv: 2308.11432Paitesanshi/LLM-Agent-Survey
Cover for A survey on large language model based autonomous agents

Abstract

Autonomous agents have long been a prominent research focus in both academic and industry communities. Previous research in this field often focuses on training agents with limited knowledge within isolated environments, which diverges significantly from human learning processes, and thus makes the agents hard to achieve human-like decisions. Recently, through the acquisition of vast amounts of web knowledge, large language models (LLMs) have demonstrated remarkable potential in achieving human-level intelligence. This has sparked an upsurge in studies investigating LLM-based autonomous agents. In this paper, we present a comprehensive survey of these studies, delivering a systematic review of the field of LLM-based autonomous agents from a holistic perspective. More specifically, we first discuss the construction of LLM-based autonomous agents, for which we propose a unified framework that encompasses a majority of the previous work. Then, we present a comprehensive overview of the diverse applications of LLM-based autonomous agents in the fields of social science, natural science, and engineering. Finally, we delve into the evaluation strategies commonly used for LLM-based autonomous agents. Based on the previous studies, we also present several challenges and future directions in this field. To keep track of this field and continuously update our survey, we maintain a repository of relevant references at this https URL.

Topics

autonomous agentsengineeringlanguage modellarge language modelssocial sciencesagent basedautonomous robot

Keywords

AgentAgentsarXivLanguagearXiv Preprint arXivLarge LanguageLLMsMemoryBasedModels

Table of Contents

  • 1 Introduction
  • 2 LLM-based Autonomous Agent Construction
  • 2.1 Agent Architecture Design
  • 2.1.1 Profiling Module
  • 2.1.2 Memory Module
  • 2.1.3 Planning Module
  • 2.1.4 Action Module
  • 2.2 Agent Capability Acquisition
  • 3 LLM-based Autonomous Agent Application
  • 3.1 Social Science
  • 3.2 Natural Science
  • 3.3 Engineering
  • 4 LLM-based Autonomous Agent Evaluation
  • 4.1 Subjective Evaluation
  • 4.2 Objective Evaluation
  • 5 Related Surveys
  • 6 Challenges
  • 6.1 Role-playing Capability
  • 6.2 Generalized Human Alignment
  • 6.3 Prompt Robustness
  • 6.4 Hallucination
  • 6.5 Knowledge Boundary
  • 6.6 Efficiency
  • 7 Conclusion
  • References

Knowls

  1. Knowl 1 — Unified Architectural Framework for LLM-Based Autonomous Agents

    model/method

    A unified architectural framework organizes large language model (LLM) based autonomous agents into four foundational modules:

    1. Profiling Module: Configures and identifies the role, characteristics, and domain identity of the agent.
    2. Memory Module: Records past observations, behaviors, and environmental states, allowing the agent to accumulate knowledge, adapt, and maintain behavioral consistency over time.
    3. Planning Module: Deconstructs complex tasks into structured sub-tasks, sequencing future actions through reasoning with or without external feedback.
    4. Action Module: Translates the planned decisions and recalled memories into concrete external operations or internal state changes.

    In this architectural workflow, the profiling module guides and constrains both the memory and planning modules, and all three modules jointly govern the execution of the action module.

  2. Knowl 2 — Memory Module: Structures, Formats, and Operations

    model/method

    The memory module enables an autonomous agent to store environmental perceptions and utilize historical interactions to guide future reasoning.

    Memory Structures

    • Unified Memory: Simulates short-term memory exclusively, holding immediate context and interaction history directly within the prompt context window via in-context learning.
    • Hybrid Memory: Explicitly decouples memory into a short-term buffer (recent perceptions and conversational turns in the context window) and an external long-term repository (past behaviors, abstract insights, and reference plans stored in external vector stores or databases for similarity-based retrieval).

    Storage Formats

    • Natural Language: Stores raw text descriptions to preserve rich semantic context.
    • Embeddings: Encodes memory segments into dense vectors for high-efficiency semantic similarity search.
    • Databases: Employs relational or symbolic databases to execute structured queries (e.g., via SQL) for deterministic addition, modification, and deletion.
    • Structured Lists / Trees: Organizes sub-goals and hierarchical plans into structured lists or triplets.

    Memory Operations

    • Memory Reading: Retrieves relevant, recent, and important memories given a query or task context.
    • Memory Writing: Ingests new records while managing memory duplication (merging or summarizing duplicate trajectory lists) and memory overflow (evicting oldest items via FIFO or explicit deletion).
    • Memory Reflection: Periodically generates higher-level abstract insights by synthesizing raw, low-level experiential records.

  3. Knowl 3 — Memory Extraction Scoring Function

    equation

    To extract the most meaningful memory item m∗m^* from the entire set of stored memories MM given a current task or context query qq, candidate memories are scored based on recency, relevance, and importance according to the objective function:

    m∗=arg⁡min⁡m∈M[αsrec(q,m)+βsrel(q,m)+γsimp(m)]m^* = \arg\min_{m \in M} \left[ \alpha s^{\text{rec}}(q, m) + \beta s^{\text{rel}}(q, m) + \gamma s^{\text{imp}}(m) \right]

    where:

    • qq is the query vector or context describing the agent's current situation or target task.
    • MM is the set of all stored memory records.
    • srec(q,m)s^{\text{rec}}(q, m) is the recency scoring function measuring how recently memory mm was accessed or created.
    • srel(q,m)s^{\text{rel}}(q, m) is the relevance scoring function measuring semantic similarity between memory mm and query qq (e.g., computed via vector similarity algorithms like FAISS, HNSW, or LSH).
    • simp(m)s^{\text{imp}}(m) is the intrinsic importance scoring function evaluating the inherent significance of memory mm independently of the query qq.
    • α,β,γ≥0\alpha, \beta, \gamma \ge 0 are non-negative balancing hyperparameters that weigh the relative contribution of each scoring criterion.

  4. Knowl 4 — Planning Module Without Feedback

    model/method

    Planning without feedback generates task execution trajectories prior to action execution, without iteratively modifying intermediate steps based on run-time environmental signals. It encompasses three primary paradigms:

    • Single-Path Reasoning: Decomposes a complex problem into a strictly linear cascade of sub-steps. Strategies include few-shot Chain-of-Thought (CoT), Zero-Shot-CoT (trigger prompts like "think step by step"), Re-Prompting (verifying prerequisite constraints before finalizing steps), and decoupled planning systems (e.g., ReWOO and HuggingGPT) that separate high-level plan decomposition from observation execution.
    • Multi-Path Reasoning: Explores candidate intermediate steps organized into tree or graph topologies. Strategies include Self-Consistency CoT (sampling multiple reasoning paths and selecting by majority voting), Tree of Thoughts (evaluating branching reasoning thoughts via BFS or DFS searches), Graph of Thoughts (expanding tree search to arbitrary graph structures), and Monte Carlo Tree Search based world models (e.g., RAP).
    • External Planner Integration: Offloads combinatorial planning to formal external solvers. The LLM translates natural language task descriptions into domain definition representations (such as Planning Domain Definition Language, PDDL), calls a classical search-based external planner to generate valid plans, and converts the output back into natural language.

  5. Knowl 5 — Planning Module With Feedback

    model/method

    In dynamic or long-horizon environments, planning with feedback allows autonomous agents to iteratively inspect, recalibrate, and update their plans based on signals gathered during execution. Feedback sources include:

    • Environmental Feedback: Observations, execution errors, and state transitions generated by real or simulated environments (e.g., ReAct's thought-act-observation triplets, compiler errors, task progress metrics, or 3D scene-graph simulator checks).
    • Human Feedback: Direct natural language instructions, evaluations, and preference corrections from human interactors to align goals and eliminate agent hallucinations.
    • Model Feedback: Internal evaluation signals generated by auxiliary language models or self-reflection routines. Approaches include self-refinement loops (generating an output, critiquing it with an LLM evaluator, and revising), specialized verification models (e.g., checker/sorter agents in InterAct), and verbal reinforcement learning (e.g., Reflexion), where textual summaries of failure trajectories serve as in-context prompts for subsequent attempts.

  6. Knowl 6 — Action Module Dimensions: Goals, Production, Space, and Impact

    model/method

    The action module operationalizes agent decisions into concrete outputs across four core dimensions:

    1. Action Goal: Categorized into:
      • Task Completion: Executing concrete functional goals (e.g., crafting items or generating code).
      • Communication: Exchanging structured or natural language messages with human users or other agents for multi-agent coordination.
      • Environment Exploration: Exploring unfamiliar states to balance exploration and exploitation.
    2. Action Production: Mechanisms by which actions are generated:
      • Memory Recollection: Triggering actions directly by retrieving relevant past experiences from the memory module.
      • Plan Following: Executing actions sequentially according to pre-generated plans from the planning module.
    3. Action Space: The set of accessible actions, divided into:
      • External Tools: Interacting through Web APIs, command-line interfaces, compilers, databases, and domain-specific external models.
      • Internal Knowledge: Relying entirely on the LLM's intrinsic conversational, planning, and commonsense reasoning capabilities.
    4. Action Impact: The downstream effects of actions, including modifying the external environment, altering internal agent states (updating memory streams and belief states), and triggering subsequent actions.

  7. Knowl 7 — Agent Capability Acquisition Strategies

    model/method

    Autonomous agents acquire domain-specific problem-solving capabilities via two primary paradigms:

    Capability Acquisition With Fine-Tuning

    Model parameters are updated using domain-specific supervision:

    • Human-Annotated Datasets: Fine-tuning on curated expert demonstrations, educational dialogues, or human interaction logs.
    • LLM-Generated Datasets: Scaling instruction datasets synthetically by prompting LLMs over API hubs or simulated sandboxes.
    • Real-World Interaction Datasets: Training LLMs on real-world logs from web navigation, database transactions, or software repositories.

    Capability Acquisition Without Fine-Tuning

    Model parameters remain fixed, enhancing performance through operational design:

    • Prompt Engineering: Structuring role definitions, demonstration exemplars, listener mental state estimates, and reflection prompts within the input context.
    • Mechanism Engineering: Implementing structured operational workflows, including:
      • Trial-and-Error: Using pre-defined validation critics to supply failure details that prompt plan revision.
      • Crowd-Sourcing / Debate: Enforcing iterative consensus across multiple debating agents.
      • Experience Accumulation: Storing executable skills or successful action scripts in dynamic libraries for subsequent retrieval.
      • Self-Driven Evolution: Enabling agents to set intrinsic sub-goals and improve policies through autonomous exploration and peer teaching.

  8. Knowl 8 — Profiling Module Generation Methods

    model/method

    The profiling module defines an agent's demographic attributes, psychological traits, professional roles, and social relationships. Profiles are established via three core methods:

    • Handcrafting Method: Developers manually write explicit role definitions and behavioral specifications into the system prompt (e.g., assigning specific software engineering roles like architect or tester, or defining Big Five personality dimensions). It offers granular control but becomes labor-intensive at scale.
    • LLM-Generation Method: An LLM generates diverse agent profiles automatically based on high-level population rules and optional seed examples. This enables scalable population construction but can reduce control over exact trait distributions.
    • Dataset Alignment Method: Demographic and behavioral profiles are extracted directly from real-world human survey datasets (e.g., election study participants) and formatted into system prompts, ensuring the simulated population reflects empirical distributions.

  9. Knowl 9 — Taxonomy of LLM-Based Autonomous Agent Applications

    model/method

    LLM-based autonomous agents are deployed across three broad disciplinary categories:

    • Social Science:
      • Psychology: Replicating human subject cognitive experiments and providing mental health conversational support.
      • Political Science & Economics: Simulating voting behavior, political discourse analysis, and agent-based economic market dynamics.
      • Social Simulation: Studying information propagation, community regulation, and opinion dynamics in multi-agent sandboxes.
      • Jurisprudence: Assisting legal judgment simulation via multi-judge voting and legal question answering.
      • Research Assistance: Automating literature summarization, keyword extraction, and hypothesis generation.
    • Natural Science:
      • Documentation & Data Management: Mining scientific literature and validating database entries for chemical and materials discovery.
      • Experiment Assistance: Automating experimental protocol design, chemical reaction planning, and safety analysis.
      • Science Education: Interactive tutoring systems for university-level mathematics and programming.
    • Engineering:
      • Civil Engineering: Interactive 3D structural design and block construction.
      • Computer Science & Software Engineering: End-to-end multi-agent software development (coding, debugging, code review), database root-cause analysis, and automated penetration testing.
      • Industrial Automation: Digital-twin-driven flexible production planning and control.
      • Robotics & Embodied AI: High-level task planning mapped to low-level robotic manipulation and navigation skills.

  10. Knowl 10 — Subjective and Objective Evaluation Methodologies for LLM-Based Agents

    model/method

    Autonomous agent evaluation is categorized into subjective and objective frameworks:

    Subjective Evaluation

    Assesses agent behavior via qualitative human or LLM proxy judgments:

    • Human Annotation: Human judges score or rank agent behaviors across dimensions such as helpfulness, honesty, harmlessness, and social intelligence.
    • Turing Test: Evaluators attempt to distinguish agent-generated responses from human-generated responses in open-ended or role-specific interactions.
    • LLM-as-a-Judge / Multi-Agent Debate: Utilizing advanced LLMs or structured multi-agent debates to conduct standardized qualitative evaluations at reduced cost.

    Objective Evaluation

    Quantifies performance using reproducible metrics and standardized benchmarks:

    • Metrics:
      • Task Success Metrics: Success rate, reward score, goal coverage, and output accuracy.
      • Human Similarity Metrics: Trajectory accuracy, dialogue similarity, and mimicry of human response distributions.
      • Efficiency Metrics: Number of planning steps, inference latency, development cost, and clarification dialogue turns.
    • Evaluation Protocols: Real-world simulation (interactive game environments), social evaluation (multi-agent collaboration and social IQ tasks), multi-task evaluation (cross-domain capability testing), and software testing (bug reproduction and test coverage benchmarks).

  11. Knowl 11 — Open Challenges and Research Frontiers in LLM-Based Agents

    limitation

    The development and deployment of LLM-based autonomous agents face six key challenges:

    1. Role-Playing Capability: LLMs struggle to accurately simulate niche, highly specialized, or newly emerging roles not well represented in web training corpora, often lacking authentic cognitive psychological self-awareness.
    2. Generalized Human Alignment: Standard LLM alignment enforces uniform, prosocial values (e.g., safety filters), which impedes realistic social simulations that require depicting non-cooperative, biased, or adversarial human behaviors to study societal problems.
    3. Prompt Robustness: Agent architectures rely on interconnected multi-module prompt frameworks where minor changes to one module's prompt can destabilize downstream planning and action behavior across different LLM backbones.
    4. Hallucination: Unmitigated generation of confident but false information can produce incorrect code, compromised security, or unsafe actions in embodied and scientific domains.
    5. Knowledge Boundary Management: When simulating real-world human behavior, LLMs may inadvertently exploit their comprehensive web knowledge base (e.g., future movie information), violating epistemic constraints that real humans would operate under.
    6. Inference Efficiency: Autoregressive generation latency compounds significantly when agents issue multiple LLM queries per action cycle for memory extraction, multi-path reasoning, and tool parsing.

Coverage note — None was omitted; all major architectural frameworks, memory/planning/action taxonomies, capability acquisition methods, application categories, evaluation strategies, and identified research challenges contributed by the survey have been systematically captured.

References

  1. 1.Mnih V, Kavukcuoglu K, Silver D, Rusu A A, Veness J, Bellemare M G, Graves A, Riedmiller M, Fidjeland A K, Ostrovski G, Petersen S, Beattie C, Sadik A, Antonoglou I, King H, Kumaran D, Wierstra D, Legg S, Hassabis D. Human-level control through deep reinforcement learning. Nature, 2015, 518(7540): 529–533
  2. 2.Lillicrap T P, Hunt J J, Pritzel A, Heess N, Erez T, Tassa Y, Silver D, Wierstra D. Continuous control with deep reinforcement learning. 2019, arXiv preprint arXiv: 1509.02971
  3. 3.Schulman J, Wolski F, Dhariwal P, Radford A, Klimov O. Proximal policy optimization algorithms. 2017, arXiv preprint arXiv: 1707.06347
  4. 4.Haarnoja T, Zhou A, Abbeel P, Levine S. Soft actor-critic: off-policy maximum entropy deep reinforcement learning with a stochastic actor. In: Proceedings of the 35th International Conference on Machine Learning. 2018, 1861−1870
  5. 5.Brown T B, Mann B, Ryder N, Subbiah M, Kaplan J D, Dhariwal P, Neelakantan A, Shyam P, Sastry G, Askell A, Agarwal S, Herbert-Voss A, Krueger G, Henighan T, Child R, Ramesh A, Ziegler D M, Wu J, Winter C, Hesse C, Chen M, Sigler E, Litwin M, Gray S, Chess B, Clark J, Berner C, McCandlish S, Radford A, Sutskever I, Amodei D. Language models are few-shot learners. In: Proceedings of the 34th Conference on Neural Information Processing Systems. 2020, 1877−1901
  6. 6.Radford A, Wu J, Child R, Luan D, Amodei D, Sutskever I. Language models are unsupervised multitask learners. OpenAI Blog, 2019, 1(8): 9
  7. 7.OpenAI. GPT-4 technical report. 2024, arXiv preprint arXiv: 2303.08774
  8. 8.Anthropic. Model card and evaluations for Claude models. See Files.anthropic.com/production/images/Model-Card-Claude-2, 2023
  9. 9.Touvron H, Lavril T, Izacard G, Martinet X, Lachaux M A, Lacroix T, Rozière B, Goyal N, Hambro E, Azhar F, Rodriguez A, Joulin A, Grave E, Lample G. LLaMA: open and efficient foundation language models. 2023, arXiv preprint arXiv: 2302.13971
  10. 10.Touvron H, Martin L, Stone K, Albert P, Almahairi A, et al. Llama 2: open foundation and fine-tuned chat models. 2023, arXiv preprint arXiv: 2307.09288
  11. 11.Chen X, Li S, Li H, Jiang S, Qi Y, Song L. Generative adversarial user model for reinforcement learning based recommendation system. In: Proceedings of the 36th International Conference on Machine Learning. 2019, 1052−1061
  12. 12.Shinn N, Cassano F, Gopinath A, rasimhan K, Yao S. Reflexion: language agents with verbal reinforcement learning. NaIn: Proceedings of the 37th Conference on Neural Information Processing Systems. 2023, 36
  13. 13.Shen Y, Song K, Tan X, Li D, Lu W, Zhuang Y. HuggingGPT: solving AI tasks with chatGPT and its friends in hugging face. In: Proceedings of the 37th Conference on Neural Information Processing Systems. 2023, 36
  14. 14.Qin Y, Liang S, Ye Y, Zhu K, Yan L, Lu Y, Lin Y, Cong X, Tang X, Qian B, Zhao S, Hong L, Tian R, Xie R, Zhou J, Gerstein M, Li D, Liu Z, Sun M. ToolLLM: facilitating large language models to master 16000+ real-world APIs. 2023, arXiv preprint arXiv: 2307.16789
  15. 15.Schick T, Dwivedi-Yu J, Dessì R, Raileanu R, Lomeli M, Hambro E, Zettlemoyer L, Cancedda N, Scialom T. Toolformer: language models can teach themselves to use tools. In: Proceedings of the 37th Conference on Neural Information Processing Systems. 2023, 36
  16. 16.Zhu X, Chen Y, Tian H, Tao C, Su W, Yang C, Huang G, Li B, Lu L, Wang X, Qiao Y, Zhang Z, Dai J. Ghost in the minecraft: generally capable agents for open-world environments via large language models with text-based knowledge and memory. 2023, arXiv preprint arXiv: 2305.17144
  17. 17.Sclar M, Kumar S, West P, Suhr A, Choi Y, Tsvetkov Y. Minding language models’ (lack of) theory of mind: a plug-and-play multi-character belief tracker. In: Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics. 2023, 13960–13980
  18. 18.Qian C, Cong X, Liu W, Yang C, Chen W, Su Y, Dang Y, Li J, Xu J, Li S, Liu Z, Sun M. Communicative agents for software development. 2023, arXiv preprint arXiv: 2307.07924
  19. 19.Chen W, Su Y, Zuo J, Yang C, Yuan C, Chan C, Yu H, Lu Y, Hung Y, Qian C, Qin Y, Cong X, Xie R, Liu Z, Sun M, Zhou, J. Agentverse: Facilitating multi-agent collaboration and exploring emergent behaviors in agents. arXiv preprint arXiv:2308.10848 .
  20. 20.Park J S, O’Brien J, Cai C J, Morris M R, Liang P, Bernstein M S. Generative agents: interactive simulacra of human behavior. In: Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology. 2023, 2
  21. 21.Zhang H, Du W, Shan J, Zhou Q, Du Y, Tenenbaum J B, Shu T, Gan C. Building cooperative embodied agents modularly with large language models. 2024, arXiv preprint arXiv: 2307.02485
  22. 22.Hong S, Zhuge M, Chen J, Zheng X, Cheng Y, Zhang C, Wang J, Wang Z, Yau S K S, Lin Z, Zhou L, Ran C, Xiao L, Wu C, Schmidhuber J. MetaGPT: meta programming for a multi-agent collaborative framework. 2023, arXiv preprint arXiv: 2308.00352
  23. 23.Dong Y, Jiang X, Jin Z, Li G. Self-collaboration code generation via chatGPT. 2023, arXiv preprint arXiv: 2304.07590
  24. 24.Serapio-García G, Safdari M, Crepy C, Sun L, Fitz S, Romero P, Abdulhai M, Faust A, Matarić M. Personality traits in large language models. 2023, arXiv preprint arXiv: 2307.00184
  25. 25.Johnson J A. Measuring thirty facets of the five factor model with a 120-item public domain inventory: development of the IPIP-NEO-120. Journal of Research in Personality, 2014, 51: 78–89
  26. 26.John O P, Donahue E M, Kentle R L. Big five inventory. Journal of personality and social psychology, 1991.
  27. 27.Deshpande A, Murahari V, Rajpurohit T, Kalyan A, Narasimhan K. Toxicity in chatGPT: analyzing persona-assigned language models. In: Proceedings of Findings of the Association for Computational Linguistics. 2023, 1236–1270
  28. 28.Wang L, Zhang J, Yang H, Chen Z, Tang J, Zhang Z, Chen X, Lin Y, Song R, Zhao W X, Xu J, Dou Z, Wang J, Wen J R. User behavior simulation with large language model based agents. 2024, arXiv preprint arXiv: 2306.02552
  29. 29.Argyle L P, Busby E C, Fulda N, Gubler J R, Rytting C, Wingate D. Out of one, many: using language models to simulate human samples. Political Analysis, 2023, 31(3): 337–351
  30. 30.Fischer K A. Reflective linguistic programming (RLP): a stepping stone in socially-aware AGI (socialAGI). 2023, arXiv preprint arXiv: 2305.12647
  31. 31.Rana K, Haviland J, Garg S, Abou-Chakra J, Reid I, Suenderhauf N. SayPlan: grounding large language models using 3D scene graphs for scalable robot task planning. In: Proceedings of the 7th Conference on Robot Learning. 2023, 23−72
  32. 32.Zhu A, Martin L, Head A, Callison-Burch C. CALYPSO: LLMs as dungeon master’s assistants. In: Proceedings of the 19th AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment. 2023, 380−390
  33. 33.Wang Z, Cai S, Chen G, Liu A, Ma X, Liang Y. Describe, explain, plan and select: interactive planning with large language models enables open-world multi-task agents. 2023, arXiv preprint arXiv: 2302.01560
  34. 34.Lin J, Zhao H, Zhang A, Wu Y, Ping H, Chen Q. AgentSims: an open-source sandbox for large language model evaluation. 2023, arXiv preprint arXiv: 2308.04026
  35. 35.Wang B, Liang X, Yang J, Huang H, Wu S, Wu P, Lu L, Ma Z, Li Z. Enhancing large language model with self-controlled memory framework. 2024, arXiv preprint arXiv: 2304.13343
  36. 36.Ng Y, Miyashita D, Hoshi Y, Morioka Y, Torii O, Kodama T, Deguchi J. SimplyRetrieve: a private and lightweight retrieval-centric generative AI tool. 2023, arXiv preprint arXiv: 2308.03983
  37. 37.Huang Z, Gutierrez S, Kamana H, Macneil S. Memory sandbox: transparent and interactive memory management for conversational agents. In: Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology. 2023, 97
  38. 38.Wang G, Xie Y, Jiang Y, Mandlekar A, Xiao C, Zhu Y, Fan L, Anandkumar A. Voyager: an open-ended embodied agent with large language models. 2023, arXiv preprint arXiv: 2305.16291
  39. 39.Zhong W, Guo L, Gao Q, Ye H, Wang Y. MemoryBank: enhancing large language models with long-term memory. 2023, arXiv preprint arXiv: 2305.10250
  40. 40.Hu C, Fu J, Du C, Luo S, Zhao J, Zhao H. ChatDB: augmenting LLMs with databases as their symbolic memory. 2023, arXiv preprint arXiv: 2306.03901
  41. 41.Zhou X, Li G, Liu Z. LLM as DBA. 2023, arXiv preprint arXiv: 2308.05481
  42. 42.Modarressi A, Imani A, Fayyaz M, Schütze H. RET-LLM: towards a general read-write memory for large language models. 2023, arXiv preprint arXiv: 2305.14322
  43. 43.Schuurmans D. Memory augmented large language models are computationally universal. 2023, arXiv preprint arXiv: 2301.04589
  44. 44.Zhao A, Huang D, Xu Q, Lin M, Liu Y J, Huang G. Expel: LLM agents are experiential learners. 2023, arXiv preprint arXiv: 2308.10144
  45. 45.Wei J, Wang X, Schuurmans D, Bosma M, Ichter B, Xia F, Chi E H, Le Q V, Zhou D. Chain-of-thought prompting elicits reasoning in large language models. In: Proceedings of the 36th Conference on Neural Information Processing Systems. 2022, 24824−24837
  46. 46.Kojima T, Gu S S, Reid M, Matsuo Y, Iwasawa Y. Large language models are zero-shot reasoners. In: Proceedings of the 36th Conference on Neural Information Processing Systems. 2022, 22199−22213
  47. 47.Raman S S, Cohen V, Rosen E, Idrees I, Paulius D, Tellex S. Planning with large language models via corrective re-prompting. In: Proceedings of Foundation Models for Decision Making Workshop at Neural Information Processing Systems. 2022
  48. 48.Xu B, Peng Z, Lei B, Mukherjee S, Liu Y, Xu D. ReWOO: decoupling reasoning from observations for efficient augmented language models. 2023, arXiv preprint arXiv: 2305.18323
  49. 49.Wang X, Wei J, Schuurmans D, Le Q V, Chi E H, Narang S, Chowdhery A, Zhou D. Self-consistency improves chain of thought reasoning in language models. In: Proceedings of the 11th International Conference on Learning Representations. 2023
  50. 50.Yao S, Yu D, Zhao J, Shafran I, Griffiths T L, Cao Y, Narasimhan K. Tree of thoughts: deliberate problem solving with large language models. In: Proceedings of the 37th Conference on Neural Information Processing Systems. 2023, 36
  51. 51.Wang Y, Jiang Z, Chen Z, Yang F, Zhou Y, Cho E, Fan X, Huang X, Lu Y, Yang Y. RecMind: Large language model powered agent for recommendation. 2023, arXiv preprint arXiv: 2308.14296
  52. 52.Besta M, Blach N, Kubicek A, Gerstenberger R, Podstawski M, Gianinazzi L, Gajda J, Lehmann T, Niewiadomski H, Nyczyk P, Hoefler T. Graph of thoughts: solving elaborate problems with large language models. 2024, arXiv preprint arXiv: 2308.09687
  53. 53.Sel B, Al-Tawaha A, Khattar V, Jia R, Jin M. Algorithm of thoughts: enhancing exploration of ideas in large language models. 2023, arXiv preprint arXiv: 2308.10379
  54. 54.Huang W, Abbeel P, Pathak D, Mordatch I. Language models as zero-shot planners: extracting actionable knowledge for embodied agents. In: Proceedings of the 39th International Conference on Machine Learning. 2022, 9118−9147
  55. 55.Gramopadhye M, Szafir D. Generating executable action plans with environmentally-aware language models. In: Proceedings of 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems. 2023, 3568−3575
  56. 56.Hao S, Gu Y, Ma H, Hong J, Wang Z, Wang D, Hu Z. Reasoning with language model is planning with world model. In: Proceedings of 2023 Conference on Empirical Methods in Natural Language Processing. 2023, 8154–8173
  57. 57.Liu B, Jiang Y, Zhang X, Liu Q, Zhang S, Biswas J, Stone P. LLM+P: empowering large language models with optimal planning proficiency. 2023, arXiv preprint arXiv: 2304.11477
  58. 58.Dagan G, Keller F, Lascarides A. Dynamic planning with a LLM. 2023, arXiv preprint arXiv: 2308.06391
  59. 59.Yao S, Zhao J, Yu D, Du N, Shafran I, Narasimhan K R, Cao Y. ReAct: synergizing reasoning and acting in language models. In: Proceedings of the 11th International Conference on Learning Representations. 2023
  60. 60.Song C H, Sadler B M, Wu J, Chao W L, Washington C, Su Y. LLM-planner: few-shot grounded planning for embodied agents with large language models. In: Proceedings of 2023 IEEE/CVF International Conference on Computer Vision. 2023, 2986−2997
  61. 61.Huang W, Xia F, Xiao T, Chan H, Liang J, Florence P, Zeng A, Tompson J, Mordatch I, Chebotar Y, Sermanet P, Jackson T, Brown N, Luu L, Levine S, Hausman K, Ichter B. Inner monologue: embodied reasoning through planning with language models. In: Proceedings of the 6th Conference on Robot Learning, 2023, 1769−1782
  62. 62.Madaan A, Tandon N, Gupta P, Hallinan S, Gao L, Wiegreffe S, Alon U, Dziri N, Prabhumoye S, Yang Y, Gupta S, Majumder B P, Hermann K, Welleck S, Yazdanbakhsh A, Clark P. Self-refine: iterative refinement with self-feedback. Advances in Neural Information Processing Systems, 2024, 36.
  63. 63.Miao N, Teh Y W, Rainforth T. SelfCheck: using LLMs to zero-shot check their own step-by-step reasoning. 2023, arXiv preprint arXiv: 2308.00436
  64. 64.Chen P L, Chang C S. InterAct: exploring the potentials of chatGPT as a cooperative agent. 2023, arXiv preprint arXiv: 2308.01552
  65. 65.Chen Z, Zhou K, Zhang B, Gong Z, Zhao X, Wen J R. ChatCoT: tool-augmented chain-of-thought reasoning on chat-based large language models. In: Proceedings of Findings of the Association for Computational Linguistics. 2023, 14777–14790
  66. 66.Nakano R, Hilton J, Balaji S, Wu J, Ouyang L, Kim C, Hesse C, Jain S, Kosaraju V, Saunders W, Jiang X, Cobbe K, Eloundou T, Krueger G, Button K, Knight M, Chess B, Schulman J. WebGPT: browser-assisted question-answering with human feedback. 2022, arXiv preprint arXiv: 2112.09332
  67. 67.Ruan J, Chen Y, Zhang B, Xu Z, Bao T, Du G, Shi S, Mao H, Li Z, Zeng X, Zhao R. TPTU: large language model-based AI agents for task planning and tool usage. 2023, arXiv preprint arXiv: 2308.03427
  68. 68.Patil S G, Zhang T, Wang X, Gonzalez J E. Gorilla: large language model connected with massive APIs. 2023, arXiv preprint arXiv: 2305.15334
  69. 69.Li M, Zhao Y, Yu B, Song F, Li H, Yu H, Li Z, Huang F, Li Y. API-bank: a comprehensive benchmark for tool-augmented LLMs. In: Proceedings of 2023 Conference on Empirical Methods in Natural Language Processing. 2023, 3102–3116
  70. 70.Song Y, Xiong W, Zhu D, Wu W, Qian H, Song M, Huang H, Li C, Wang K, Yao R, Tian Y, Li S. RestGPT: connecting large language models with real-world RESTful APIs. 2023, arXiv preprint arXiv: 2306.06624
  71. 71.Liang Y, Wu C, Song T, Wu W, Xia Y, Liu Y, Ou Y, Lu S, Ji L, Mao S, Wang Y, Shou L, Gong M, Duan N. TaskMatrix.AI: Completing tasks by connecting foundation models with millions of APIs. 2023, arXiv preprint arXiv: 2303.16434
  72. 72.Karpas E, Abend O, Belinkov Y, Lenz B, Lieber O, Ratner N, Shoham Y, Bata H, Levine Y, Leyton-Brown K, Muhlgay D, Rozen N, Schwartz E, Shachaf G, Shalev-Shwartz S, Shashua A, Tenenholtz M. MRKL systems: a modular, neuro-symbolic architecture that combines large language models, external knowledge sources and discrete reasoning. 2022, arXiv preprint arXiv: 2205.00445
  73. 73.Ge Y, Hua W, Mei K, Tan J, Xu S, Li Z, Zhang Y. OpenAGI: When LLM meets domain experts. In: Proceedings of the 37th Conference on Neural Information Processing Systems, 2023, 36
  74. 74.Surís D, Menon S, Vondrick C. ViperGPT: visual inference via python execution for reasoning. 2023, arXiv preprint arXiv: 2303.08128
  75. 75.Bran A M, Cox S, Schilter O, Baldassari C, White A D, Schwaller P. ChemCrow: augmenting large-language models with chemistry tools. 2023, arXiv preprint arXiv: 2304.05376
  76. 76.Yang Z, Li L, Wang J, Lin K, Azarnasab E, Ahmed F, Liu Z, Liu C, Zeng M, Wang L. MM-REACT: Prompting chatGPT for multimodal reasoning and action. 2023, arXiv preprint arXiv: 2303.11381
  77. 77.Gao C, Lan X, Lu Z, Mao J, Piao J, Wang H, Jin D, Li Y. S3: social-network simulation system with large language model-empowered agents. 2023, arXiv preprint arXiv: 2307.14984
  78. 78.Ichter B, Brohan A, Chebotar Y, Finn C, Hausman K, et al. Do as I can, not as I say: grounding language in robotic affordances. In: Proceedings of the 6th Conference on Robot Learning. 2023, 287−318
  79. 79.Liu H, Sferrazza C, Abbeel P. Chain of hindsight aligns language models with feedback. arXiv preprint arXiv: 2302.02676
  80. 80.Yao S, Chen H, Yang J, Narasimhan K. WebShop: towards scalable real-world Web interaction with grounded language agents. In: Proceedings of the 36th Conference on Neural Information Processing Systems. 2022, 20744−20757
  81. 81.Dan Y, Lei Z, Gu Y, Li Y, Yin J, Lin J, Ye L, Tie Z, Zhou Y, Wang Y, Zhou A, Zhou Z, Chen Q, Zhou J, He L, Qiu X. EduChat: a large-scale language model-based chatbot system for intelligent education. 2023, arXiv preprint arXiv: 2308.02773
  82. 82.Lin B Y, Fu Y, Yang K, Brahman F, Huang S, Bhagavatula C, Ammanabrolu P, Choi Y, Ren X. SwiftSage: a generative agent with fast and slow thinking for complex interactive tasks. In: Proceedings of the 37th Conference on Neural Information Processing Systems. 2023, 36
  83. 83.Evans J S B T, Stanovich K E. Dual-process theories of higher cognition: advancing the debate. Perspectives on Psychological Science, 2013, 8(3): 223–241
  84. 84.Liu R, Yang R, Jia C, Zhang G, Zhou D, Dai A M, Yang D, Vosoughi S. Training socially aligned language models on simulated social interactions. 2023, arXiv preprint arXiv: 2305.16960
  85. 85.Weng X, Gu Y, Zheng B, Chen S, Stevens S, Wang B, Sun H, Su Y. Mind2Web: towards a generalist agent for the Web. In: Proceedings of the 37th Conference on Neural Information Processing Systems. 2023, 36
  86. 86.Sun R, Arik S O, Nakhost H, Dai H, Sinha R, Yin P, Pfister T. SQL-PaLm: improved large language model adaptation for text-to-SQL. 2023, arXiv preprint arXiv: 2306.00739
  87. 87.Yao W, Heinecke S, Niebles J C, Liu Z, Feng Y, Xue L, Murthy R, Chen Z, Zhang J, Arpit D, Xu R, Mui P, Wang H, Xiong C, Savarese S. Retroformer: retrospective large language agents with policy gradient optimization, 2023, arXiv preprint arXiv: 2308.02151
  88. 88.Shu Y, Zhang H, Gu H, Zhang P, Lu T, Li D, Gu N. RAH! RecSys-assistant-human: a human-centered recommendation framework with LLM agents. 2023, arXiv preprint arXiv: 2308.09904
  89. 89.Mandi Z, Jain S, Song S. RoCo: dialectic multi-robot collaboration with large language models. 2023, arXiv preprint arXiv: 2307.04738
  90. 90.Zhang C, Liu L, Wang J, Wang C, Sun X, Wang H, Cai M. PREFER: prompt ensemble learning via feedback-reflect-refine. 2023, arXiv preprint arXiv: 2308.12033
  91. 91.Du Y, Li S, Torralba A, Tenenbaum J B, Mordatch I. Improving factuality and reasoning in language models through multiagent debate. 2023, arXiv preprint arXiv: 2305.14325
  92. 92.Zhang C, Yang Z, Liu J, Han Y, Chen X, Huang Z, Fu B, Yu G. AppAgent: multimodal agents as smartphone users. 2023, arXiv preprint arXiv: 2312.13771
  93. 93.Madaan A, Tandon N, Clark P, Yang Y. Memory-assisted prompt editing to improve GPT-3 after deployment. In: Proceedings of 2022 Conference on Empirical Methods in Natural Language Processing. 2022, 2833–2861
  94. 94.Colas C, Teodorescu L, Oudeyer P Y, Yuan X, Côté M A. Augmenting autotelic agents with large language models. In: Proceedings of the 2nd Conference on Lifelong Learning Agents. 2023, 205–226
  95. 95.Nascimento N, Alencar P, Cowan D. Self-adaptive large language model (LLM)-based multiagent systems. In: Proceedings of 2023 IEEE International Conference on Autonomic Computing and Self-Organizing Systems Companion. 2023, 104−109
  96. 96.Saha S, Hase P, Bansal M. Can language models teach weaker agents? Teacher explanations improve students via personalization. 2023, arXiv preprint arXiv: 2306.09299
  97. 97.Zhuge M, Liu H, Faccio F, Ashley D R, Csordás R, Gopalakrishnan A, Hamdi A, Hammoud H A A K, Herrmann V, Irie K, Kirsch L, Li B, Li G, Liu S, Mai J, Piękos P, Ramesh A, Schlag I, Shi W, Stanić A, Wang W, Wang Y, Xu M, Fan D P, Ghanem B, Schmidhuber J. Mindstorms in natural language-based societies of mind. 2023, arXiv preprint arXiv: 2305.17066
  98. 98.Park J S, Popowski L, Cai C, Morris M R, Liang P, Bernstein M S. Social simulacra: creating populated prototypes for social computing systems. In: Proceedings of the 35th Annual ACM Symposium on User Interface Software and Technology. 2022, 74
  99. 99.Li G, Hammoud H A A K, Itani H, Khizbullin D, Ghanem B. CAMEL: communicative agents for "mind" exploration of large language model society. 2023, arXiv preprint arXiv: 2303.17760
  100. 100.AutoGPT. See Github.com/Significant-Gravitas/Auto, 2023
  101. 101.Chen L, Wang L, Dong H, Du Y, Yan J, Yang F, Li S, Zhao P, Qin S, Rajmohan S, Lin Q, Zhang D. Introspective tips: large language model for in-context decision making. 2023, arXiv preprint arXiv: 2305.11598
  102. 102.Aher G V, Arriaga R I, Kalai A T. Using large language models to simulate multiple humans and replicate human subject studies. In: Proceedings of the 40th International Conference on Machine Learning. 2023, 337−371
  103. 103.Akata E, Schulz L, Coda-Forno J, Oh S J, Bethge M, Schulz E. Playing repeated games with large language models. 2023, arXiv preprint arXiv: 2305.16867
  104. 104.Ma Z, Mei Y, Su Z. Understanding the benefits and challenges of using large language model-based conversational agents for mental well-being support. In: Proceedings of AMIA Symposium. 2023, 1105−1114
  105. 105.Ziems C, Held W, Shaikh O, Chen J, Zhang Z, Yang D. Can large language models transform computational social science? 2024, arXiv preprint arXiv: 2305.03514
  106. 106.Horton J J. Large language models as simulated economic agents: what can we learn from homo silicus? 2023, arXiv preprint arXiv: 2301.07543
  107. 107.Li S, Yang J, Zhao K. Are you in a masquerade? Exploring the behavior and impact of large language model driven social bots in online social networks. 2023, arXiv preprint arXiv: 2307.10337
  108. 108.Li C, Su X, Han H, Xue C, Zheng C, Fan C. Quantifying the impact of large language models on collective opinion dynamics. 2023, arXiv preprint arXiv: 2308.03313
  109. 109.Kovač G, Portelas R, Dominey P F, Oudeyer P Y. The SocialAI school: insights from developmental psychology towards artificial socio-cultural agents. 2023, arXiv preprint arXiv: 2307.07871
  110. 110.Williams R, Hosseinichimeh N, Majumdar A, Ghaffarzadegan N. Epidemic modeling with generative agents. 2023, arXiv preprint arXiv: 2307.04986
  111. 111.Shi J, Zhao J, Wang Y, Wu X, Li J, He L. CGMI: configurable general multi-agent interaction framework. 2023, arXiv preprint arXiv: 2308.12503
  112. 112.Cui J, Li Z, Yan Y, Chen B, Yuan L. ChatLaw: open-source legal large language model with integrated external knowledge bases. 2023, arXiv preprint arXiv: 2306.16092
  113. 113.Hamilton S. Blind judgement: agent-based supreme court modelling with GPT. 2023, arXiv preprint arXiv: 2301.05327
  114. 114.Bail C A. Can generative AI improve social science? 2023
  115. 115.Boiko D A, MacKnight R, Gomes G. Emergent autonomous scientific research capabilities of large language models. 2023, arXiv preprint arXiv: 2304.05332
  116. 116.Kang Y, Kim J. ChatMOF: an autonomous AI system for predicting and generating metal-organic frameworks. 2023, arXiv preprint arXiv: 2308.01423
  117. 117.Swan M, Kido T, Roland E, Santos R P D. Math agents: computational infrastructure, mathematical embedding, and genomics. 2023, arXiv preprint arXiv: 2307.02502
  118. 118.Drori I, Zhang S, Shuttleworth R, Tang L, Lu A, Ke E, Liu K, Chen L, Tran S, Cheng N, Wang R, Singh N, Patti T L, Lynch J, Shporer A, Verma N, Wu E, Strang G. A neural network solves, explains, and generates university math problems by program synthesis and few-shot learning at human level. Proceedings of the National Academy of Sciences of the United States of America, 2022, 119(32): e2123433119
  119. 119.Chen M, Tworek J, Jun H, Yuan Q, de Oliveira Pinto H P, et al. Evaluating large language models trained on code. 2021, arXiv preprint arXiv: 2107.03374
  120. 120.Liffiton M, Sheese B E, Savelka J, Denny P. CodeHelp: using large language models with guardrails for scalable support in programming classes. In: Proceedings of the 23rd Koli Calling International Conference on Computing Education Research. 2023, 8
  121. 121.Matelsky J K, Parodi F, Liu T, Lange R D, Kording K P. A large language model-assisted education tool to provide feedback on open-ended responses. 2023, arXiv preprint arXiv: 2308.02439
  122. 122.Mehta N, Teruel M, Sanz P F, Deng X, Awadallah A H, Kiseleva J. Improving grounded language understanding in a collaborative environment by interacting with agents through help feedback. 2024, arXiv preprint arXiv: 2304.10750
  123. 123.SmolModels. See Githubcom/smol-ai/developer website, 2023
  124. 124.DemoGPT. See Github.com/melih-unsal/Demo website, 2023
  125. 125.GPT-engineer. See Github.com/AntonOsika/gpt website, 2023
  126. 126.Li H, Hao Y, Zhai Y, Qian Z. The hitchhiker’s guide to program analysis: a journey with large language models. 2023, arXiv preprint arXiv: 2308.00245
  127. 127.He Z, Wu H, Zhang X, Yao X, Zheng S, Zheng H, Yu B. ChatEDA: a large language model powered autonomous agent for EDA. In: Proceedings of the 5th ACM/IEEE Workshop on Machine Learning for CAD. 2023, 1−6
  128. 128.Deng G, Liu Y, Mayoral-Vilches V, Liu P, Li Y, Xu Y, Zhang T, Liu Y, Pinzger M, Rass S. PentestGPT: an LLM-empowered automatic penetration testing tool. 2023, arXiv preprint arXiv: 2308.06782
  129. 129.Xia Y, Shenoy M, Jazdi N, Weyrich M. Towards autonomous system: flexible modular production system enhanced with large language model agents. In: Proceedings of the 2023 IEEE 28th International Conference on Emerging Technologies and Factory Automation. 2023, 1−8
  130. 130.Ogundare O, Madasu S, Wiggins N. Industrial engineering with large language models: a case study of chatGPT’s performance on oil & gas problems. In: Proceedings of the 2023 11th International Conference on Control, Mechatronics and Automation. 2023, 458−461
  131. 131.Hu B, Zhao C, Zhang P, Zhou Z, Yang Y, Xu Z, Liu B. Enabling intelligent interactions between an agent and an LLM: a reinforcement learning approach. 2024, arXiv preprint arXiv: 2306.03604
  132. 132.Wu Y, Min S Y, Bisk Y, Salakhutdinov R, Azaria A, Li Y, Mitchell T, Prabhumoye S. Plan, eliminate, and track−language models are good teachers for embodied agents. 2023, arXiv preprint arXiv: 2305.02412
  133. 133.Zhang D, Chen L, Zhang S, Xu H, Zhao Z, Yu K. Large language models are semi-parametric reinforcement learning agents. In: Proceedings of the 37th Conference on Neural Information Processing Systems. 2023, 36
  134. 134.Di P N, Byravan A, Hasenclever L, Wulfmeier M, Heess N, Riedmiller M. Towards a unified agent with foundation models. 2023, arXiv preprint arXiv: 2307.09668
  135. 135.Dasgupta I, Kaeser-Chen C, Marino K, Ahuja A, Babayan S, Hill F, Fergus R. Collaborating with language models for embodied reasoning. 2023, arXiv preprint arXiv: 2302.00763
  136. 136.Zhou W, Peng X, Riedl M O. Dialogue shaping: empowering agents through NPC interaction. 2023, arXiv preprint arXiv: 2307.15833
  137. 137.Nottingham K, Ammanabrolu P, Suhr A, Choi Y, Hajishirzi H, Singh S, Fox R. Do embodied agents dream of pixelated sheep: embodied decision making using language guided world modelling. In: Proceedings of the 40th International Conference on Machine Learning. 2023, 26311–26325
  138. 138.Wu Z, Wang Z, Xu X, Lu J, Yan H. Embodied task planning with large language models. 2023, arXiv preprint arXiv: 2307.01848
  139. 139.Wu J, Antonova R, Kan A, Lepert M, Zeng A, Song S, Bohg J, Rusinkiewicz S, Funkhouser T. TidyBot: personalized robot assistance with large language models. In: Proceedings of 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems. 2023, 3546−3553
  140. 140.AgentGPT. See Github.com/reworkd/Agent website, 2023
  141. 141.Ai-legion. See Github.com/eumemic/ai website, 2023
  142. 142.AGiXT. See Githubcom/Josh-XT/AGiXT website, 2023
  143. 143.Xlang. See Githubcom/xlang-ai/xlang website, 2023
  144. 144.Babyagi. See Githubcom/yoheinakajima website, 2023
  145. 145.LangChain. See Docs.langchaincom/docs/ website, 2023
  146. 146.WorkGPT. See Githubcom/team-openpm/workgpt website, 2023
  147. 147.LoopGPT. See Githubcom/farizrahman4u/loopgpt website, 2023
  148. 148.GPT-researcher. See Github.com/assafelovic/gpt website, 2023
  149. 149.Qin Y, Hu S, Lin Y, Chen W, Ding N, Cui G, Zeng Z, Huang Y, Xiao C, Han C, Fung Y R, Su Y, Wang H, Qian C, Tian R, Zhu K, Liang S, Shen X, Xu B, Zhang Z, Ye Y, Li B, Tang Z, Yi J, Zhu Y, Dai Z, Yan L, Cong X, Lu Y, Zhao W, Huang Y, Yan J, Han X, Sun X, Li D, Phang J, Yang X, Wu T, Ji H, Liu Z, Sun M. Tool learning with foundation models. 2023, arXiv preprint arXiv: 2304.08354
  150. 150.Transformers agent. See Huggingface.co/docs/transformers/transformers website, 2023
  151. 151.Mini-agi. See Github.com/muellerberndt/mini website, 2023
  152. 152.SuperAGI. See Github.com/TransformerOptimus/Super website, 2023
  153. 153.Wu Q, Bansal G, Zhang J, Wu Y, Li B, Zhu E, Jiang L, Zhang X, Zhang S, Liu J, Awadallah A H, White R W, Burger D, Wang C. AutoGen: enabling next-gen LLM applications via multi-agent conversation. 2023, arXiv preprint arXiv: 2308.08155
  154. 154.Grossmann I, Feinberg M, Parker D C, Christakis N A, Tetlock P E, Cunningham W A. AI and the transformation of social science research: careful bias management and data fidelity are key. Science, 2023, 380(6650): 1108–1109
  155. 155.Huang X, Lian J, Lei Y, Yao J, Lian D, Xie X. Recommender AI agent: integrating large language models for interactive recommendations. 2023, arXiv preprint arXiv: 2308.16505
  156. 156.Zhang C, Yang K, Hu S, Wang Z, Li G, Sun Y, Zhang C, Zhang Z, Liu A, Zhu S C, Chang X, Zhang J, Yin F, Liang Y, Yang Y. ProAgent: building proactive cooperative agents with large language models. 2024, arXiv preprint arXiv: 2308.11339
  157. 157.Xiang J, Tao T, Gu Y, Shu T, Wang Z, Yang Z, Hu Z. Language models meet world models: embodied experiences enhance language models. In: Proceedings of the 37th Conference on Neural Information Processing Systems. 2023, 36
  158. 158.Lee M, Srivastava M, Hardy A, Thickstun J, Durmus E, Paranjape A, Gerard-Ursin I, Li X L, Ladhak F, Rong F, Wang R E, Kwon M, Park J S, Cao H, Lee T, Bommasani R, Bernstein M, Liang P. Evaluating human-language model interaction. 2024, arXiv preprint arXiv: 2212.09746
  159. 159.Krishna R, Lee D, Fei-Fei L, Bernstein M S. Socially situated artificial intelligence enables learning from human interaction. Proceedings of the National Academy of Sciences of the United States of America, 2022, 119(39): e2115730119
  160. 160.Huang J T, Lam M H, Li E J, Ren S, Wang W, Jiao W, Tu Z, Lyu M R. Emotionally numb or empathetic? Evaluating how LLMs feel using emotionbench. 2024, arXiv preprint arXiv: 2308.03656
  161. 161.Chan C M, Chen W, Su Y, Yu J, Xue W, Zhang S, Fu J, Liu Z. ChatEval: towards better LLM-based evaluators through multi-agent debate. 2023, arXiv preprint arXiv: 2308.07201
  162. 162.Chen A, Phang J, Parrish A, Padmakumar V, Zhao C, Bowman S R, Cho K. Two failures of self-consistency in the multi-step reasoning of LLMs. 2024, arXiv preprint arXiv: 2305.14279
  163. 163.Zhang D, Xu H, Zhao Z, Chen L, Cao R, Yu K. Mobile-env: an evaluation platform and benchmark for LLM-GUI interaction. 2024, arXiv preprint arXiv: 2305.08144
  164. 164.Liang Y, Zhu L, Yang Y. Tachikuma: understading complex interactions with multi-character and novel objects by large language models. 2023, arXiv preprint arXiv: 2307.12573
  165. 165.Choi M, Pei J, Kumar S, Shu C, Jurgens D. Do LLMs understand social knowledge? Evaluating the sociability of large language models with socKET benchmark. In: Proceedings of 2023 Conference on Empirical Methods in Natural Language Processing. 2023, 11370–11403
  166. 166.Liu Z, Yao W, Zhang J, Xue L, Heinecke S, Murthy R, Feng Y, Chen Z, Niebles J C, Arpit D, Xu R, Mui P, Wang H, Xiong C, Savarese S. BOLAA: benchmarking and orchestrating LLM-augmented autonomous agents. 2023, arXiv preprint arXiv: 2308.05960
  167. 167.Liu X, Yu H, Zhang H, Xu Y, Lei X, Lai H, Gu Y, Ding H, Men K, Yang K, Zhang S, Deng X, Zeng A, Du Z, Zhang C, Shen S, Zhang T, Su Y, Sun H, Huang M, Dong Y, Tang J. AgentBench: evaluating LLMs as agents. 2023, arXiv preprint arXiv: 2308.03688
  168. 168.Kang S, Yoon J, Yoo S. Large language models are few-shot testers: exploring LLM-based general bug reproduction. In: Proceedings of the 45th IEEE/ACM International Conference on Software Engineering. 2023, 2312−2323
  169. 169.Jalil S, Rafi S, LaToza T D, Moran K, Lam W. ChatGPT and software testing education: Promises & perils. In: Proceedings of 2023 IEEE International Conference on Software Testing, Verification and Validation Workshops. 2023, 4130−4137
  170. 170.Feldt R, Kang S, Yoon J, Yoo S. Towards autonomous testing agents via conversational large language models. In: Proceedings of the 38th IEEE/ACM International Conference on Automated Software Engineering. 2023, 1688−1693
  171. 171.Zhou S, Xu F F, Zhu H, Zhou X, Lo R, Sridhar A, Cheng X, Ou T, Bisk Y, Fried D, Alon U, Neubig G. WebArena: a realistic Web environment for building autonomous agents. 2023, arXiv preprint arXiv: 2307.13854
  172. 172.Xu B, Liu X, Shen H, Han Z, Li Y, Yue M, Peng Z, Liu Y, Yao Z, Xu D. Gentopia.AI: a collaborative platform for tool-augmented LLMs. In: Proceedings of 2023 Conference on Empirical Methods in Natural Language Processing: System Demonstrations. 2023, 237−245
  173. 173.Chalamalasetti K, Götze J, Hakimov S, Madureira B, Sadler P, Schlangen D. clembench: Using game play to evaluate chat-optimized language models as conversational agents. In: Proceedings of 2023 Conference on Empirical Methods in Natural Language Processing. 2023, 11174–11219
  174. 174.Banerjee D, Singh P, Avadhanam A, Srivastava S. Benchmarking LLM powered chatbots: methods and metrics. 2023, arXiv preprint arXiv: 2308.04624
  175. 175.Lin J, Tomlin N, Andreas J, Eisner J. Decision-oriented dialogue for human-AI collaboration. 2023, arXiv preprint arXiv: 2305.20076
  176. 176.Zhao W X, Zhou K, Li J, Tang T, Wang X, Hou Y, Min Y, Zhang B, Zhang J, Dong Z, Du Y, Yang C, Chen Y, Chen Z, Jiang J, Ren R, Li Y, Tang X, Liu Z, Liu P, Nie J Y, Wen J R. A survey of large language models. 2023, arXiv preprint arXiv: 2303.18223
  177. 177.Yang J, Jin H, Tang R, Han X, Feng Q, Jiang H, Zhong S, Yin B, Hu X. Harnessing the power of LLMs in practice: a survey on chatGPT and beyond. ACM Transactions on Knowledge Discovery from Data, 2024, doi: 10.1145/3649506
  178. 178.Wang Y, Zhong W, Li L, Mi F, Zeng X, Huang W, Shang L, Jiang X, Liu Q. Aligning large language models with human: a survey. 2023, arXiv preprint arXiv: 2307.12966
  179. 179.Huang J, Chang K C C. Towards reasoning in large language models: a survey. In: Proceedings of Findings of the Association for Computational Linguistics: ACL 2023. 2023, 1049–1065
  180. 180.Mialon G, Dessì R, Lomeli M, Nalmpantis C, Pasunuru R, Raileanu R, Rozière B, Schick T, Dwivedi-Yu J, Celikyilmaz A, Grave E, LeCun Y, Scialom T. Augmented language models: a survey. 2023, arXiv preprint arXiv: 2302.07842
  181. 181.Chang Y, Wang X, Wang J, Wu Y, Yang L, Zhu K, Chen H, Yi X, Wang C, Wang Y, Ye W, Zhang Y, Chang Y, Yu P S. A survey on evaluation of large language models. ACM Transactions on Intelligent Systems and Technology, 2023, doi: 10.1145/3641289
  182. 182.Chang T A, Bergen B K. Language model behavior: a comprehensive survey. Computational Linguistics, 2024, doi: 10.1162/coli_a_00492
  183. 183.Li C, Wang J, Zhu K, Zhang Y, Hou W, Lian J, Xie X. Emotionprompt: Leveraging psychology for large language models enhancement via emotional stimulus. 2023, arXiv preprint arXiv: 2307.11760
  184. 184.Zhuo T Y, Li Z, Huang Y, Shiri F, Wang W, Haffari G, Li Y F. On robustness of prompt-based semantic parsing with large pre-trained language model: an empirical study on codex. In: Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics. 2023, 1090–1102
  185. 185.Gekhman Z, Oved N, Keller O, Szpektor I, Reichart R. On the robustness of dialogue history representation in conversational question answering: a comprehensive study and a new prompt-based method. Transactions of the Association for Computational Linguistics, 2023, 11(11): 351–366
  186. 186.Ji Z, Lee N, Frieske R, Yu T, Su D, Xu Y, Ishii E, Bang Y J, Madotto A, Fung P. Survey of hallucination in natural language generation. ACM Computing Surveys, 2023, 55(12): 248

Citation

MLA
Wang, L., et al. “A Survey on Large Language Model Based Autonomous Agents”. Frontiers of Computer Science, vol. 18, no. 6, 2024, https://doi.org/10.1007/s11704-024-40231-1.
APA
Wang, L., Ma, C., Feng, X., Zhang, Z., Yang, H., Zhang, J., Chen, Z., Tang, J., Chen, X., Lin, Y., Zhao, W. X., Wei, Z., & Wen, J. (2024). A survey on large language model based autonomous agents. Frontiers of Computer Science, 18(6). https://doi.org/10.1007/s11704-024-40231-1
Chicago
Wang, L., C. Ma, X. Feng, et al. 2024. “A Survey on Large Language Model Based Autonomous Agents”. Frontiers of Computer Science 18 (6). https://doi.org/10.1007/s11704-024-40231-1.
Harvard
Wang, L. et al. (2024) “A survey on large language model based autonomous agents”, Frontiers of Computer Science, 18(6). Available at: https://doi.org/10.1007/s11704-024-40231-1.
Vancouver
1. Wang L, Ma C, Feng X, et al (2024) A survey on large language model based autonomous agents. Frontiers of Computer Science. https://doi.org/10.1007/s11704-024-40231-1

BibTeX

@article{Wang_2024, title={A survey on large language model based autonomous agents}, volume={18}, ISSN={2095-2236}, url={http://dx.doi.org/10.1007/s11704-024-40231-1}, DOI={10.1007/s11704-024-40231-1}, number={6}, journal={Frontiers of Computer Science}, publisher={Springer Science and Business Media LLC}, author={Wang, Lei and Ma, Chen and Feng, Xueyang and Zhang, Zeyu and Yang, Hao and Zhang, Jingsen and Chen, Zhiyuan and Tang, Jiakai and Chen, Xu and Lin, Yankai and Zhao, Wayne Xin and Wei, Zhewei and Wen, Jirong}, year={2024}, month=Mar }
Metadata:Crossref

Source Code

This paper has an official code repository available. Click below to access the source code.

View Repository

Access the Paper

This paper is available from its original source. Click below to access the PDF.

Open PDF

License: https://creativecommons.org/licenses/by/4.0/