Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

A Call for Collaborative Intelligence: Why Human-Agent Systems Should Precede AI Autonomy

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Fully autonomous LLM agents are premature, and human-agent collaboration should be the priority.

desk verdict A sincere, well-structured position paper that usefully frames the human-agent debate but overclaims 'we show' without a comparative experiment; its utility model is definitional. read the letter →

arxiv 2506.09420 v1 pith:DORHIH7P submitted 2025-06-11 cs.AI cs.CLcs.HCcs.LGcs.MA

Henry Peng Zou , Wei-Chieh Huang , Yaozu Wu , Chunyu Miao , Dongyuan Li , Aiwei Liu , Yue Zhou , Yankai Chen
show 5 more authors
Weizhi Zhang Yangning Li Liancheng Fang Renhe Jiang Philip S. Yu
This is my paper
classification cs.AIcs.CLcs.HCcs.LGcs.MA
keywords human-agentsystemsLLMagentsautonomyhuman-in-the-loopcollaborativeintelligenceAIsafetyhumanfeedbackutilitymodel
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This position paper claims that the rush to fully autonomous LLM agents is premature and that the field should instead build LLM-based Human-Agent Systems, where people remain in the loop to clarify, correct, and control. It argues that autonomy-first thinking misidentifies the goal: progress should be measured by how well systems collaborate with humans, not by how much independence they achieve. A sympathetic reader would care because this choice determines where research, deployment, and regulation effort goes in high-stakes domains like healthcare, finance, and software. The paper grounds the argument in a utility trade-off between human cost and error cost, concluding that the current optimum still requires substantial human participation.

What carries the argument

The central object is the utility model $U = V \cdot S - C_h - C_e$, where $V$ is the value created per successful task, $S$ is the success rate, $C_h$ is human cost, and $C_e$ is error cost, together with the optimal human involvement $h^* = \arg\max_h U(h)$. This reframes the debate as an optimization rather than an ideology: human involvement is not a moral choice but a variable whose optimum shifts with AI capability, and the paper's recommendation follows from the claim that the current $h^*$ is significantly above zero.

What would settle it

Run the same high-stakes task, such as financial portfolio selection or clinical triage, twice, once fully autonomous and once with human oversight, measuring success rate, error cost, and human time; the central claim is falsified if the autonomous version yields equal or higher utility $U = V \cdot S - C_h - C_e$ in a realistic setting.

Watch

Extended reading notes

Core claim

The paper's central claim is normative: deploying fully autonomous LLM-based agents in complex real-world settings today carries reliability, ambiguity-handling, and accountability risks that outweigh their speed and scale benefits, so the field should prioritize LLM-based Human-Agent Systems (LLM-HAS), in which humans clarify goals, correct outputs, and assume control in sensitive situations. It formalizes this trade-off as $U = V \cdot S - C_h - C_e$, with optimal human involvement $h^* = \arg\max_h U(h)$, and argues that current technology puts $h^*$ well above zero even though $h^*$ will fall as AI improves.

Load-bearing premise

The load-bearing premise is that human involvement materially raises success rates and lowers error costs enough to outweigh the extra time, effort, and delay it adds; the paper offers examples and a formula but no measured comparison.

Editorial extensions

If this is right

  • If the central claim holds, evaluation of AI systems should include human workload, cognitive load, and interaction quality, not only agent accuracy on static benchmarks.
  • Research priorities should shift toward feedback integration, lifelong learning, and dynamic optimization in human-agent teams rather than fully autonomous agent pipelines.
  • High-stakes deployments in medicine, finance, and driving should retain explicit human oversight and defined responsibility, at least until AI reliability matures.
  • As agent capability improves, the optimal human involvement $h^*$ falls, making human-agent systems a staged path toward autonomy, with humans withdrawn where measurement justifies it.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The utility framing implies a direct comparative prediction: on tasks with high error cost and ambiguous goals, human-agent systems should outperform fully autonomous systems in total utility, which is testable in a controlled benchmark.
  • A hidden risk the paper leaves open is that if human feedback is slow, noisy, or unavailable, $h^*$ may be lower than the paper assumes, so the recommendation should be treated as domain-conditional rather than global.
  • The framework could be extended to define an 'autonomy readiness' threshold, the point where measured $h^*$ approaches zero, letting regulators gate deployment on observed success rates, human costs, and error costs.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. This position paper argues that the AI community should slow its push toward fully autonomous LLM-based agents and instead prioritize LLM-based Human-Agent Systems (LLM-HAS), in which humans and agents collaborate through guidance, feedback, and oversight. The paper identifies reliability, ambiguity handling, and accountability as key weaknesses of autonomous agents; presents advantages of the HAS paradigm; surveys applications in embodied AI, software development, finance, healthcare, and other domains; lists technical and evaluation challenges; and gives implementation guidelines covering setup, data, model engineering, deployment, and evaluation. Section 7 introduces a utility model U = V·S − C_h − C_e and defines optimal human involvement h* as the maximizing degree of involvement, using this model to frame responses to alternative views. The paper's central normative claim is that progress should be measured not by agent independence but by the quality of human-agent collaboration, and that fully autonomous deployment at the current stage is premature.

Significance. If the paper's central claim were backed by quantitative evidence, it would be a significant contribution to the current debate on LLM agent deployment, offering a concrete framework for reasoning about the trade-off between human involvement and error risk. The paper is valuable as a position statement: it clearly articulates a paradigm, organizes a large body of related work, provides a useful taxonomy of implementation challenges (Section 6), and canvasses counterarguments (Section 7). It also makes an honest admission of the field's evaluation gaps, which is a strength. However, the paper's significance as a research contribution is currently limited by the absence of empirical calibration or controlled comparison: the utility model in Section 7 is definitional, and the key premise that human input improves success rate and reduces error cost enough to offset human cost is asserted rather than demonstrated. The paper does not ship machine-checked proofs or new experimental data; its positive content is a structured argument plus a research agenda, not a tested result.

major comments (4)
  1. [Section 7 (Definitions 7.1 and 7.2) and Section 7.2] The utility framework is definitional rather than evidential: U = V·S − C_h − C_e is introduced with no measurement procedures or constraints, and h* is by construction the argmax of this utility. Consequently, the statement in Section 7.2 that 'current technological limitations mean that the optimal balance still requires significant human participation' is an assertion, not a consequence of the framework. To make the framework load-bearing, the authors should either derive conditions on S, C_h, and C_e under which h* > 0, or calibrate these quantities from existing empirical studies of human-LLM collaboration; otherwise the framework merely restates the position it is meant to support.
  2. [Section 5 (Evaluation) and Sections 7.3–7.4] The paper's own concessions undercut the central premise that human involvement raises success rate S and lowers error cost C_e enough to offset human cost C_h. Section 5 states that 'standard metrics for human workload and efficiency are lacking' and that LLM-simulated human proxies 'often fail to reflect the variability of real human input,' while Section 7.4 concedes that human-in-the-loop can be a bottleneck for time-critical tasks. These admissions mean the paper provides no evidence that U(HAS) > U(autonomous) in any specific domain; the authors should either supply quantitative evidence from human-subject studies or explicitly restrict the thesis to high-stakes, non-time-critical domains where the balance is most plausibly favorable.
  3. [Section 4 (Applications) and Abstract] The claim that human-AI teamwork 'can handle complex tasks better than AI working alone' is supported only by citations to domain examples such as FinArena, healthcare systems, and software development agents, without any comparative experiment, meta-analysis, or even a table summarizing effect sizes, success rates, and error rates from the cited works. The paper should add a systematic comparison of reported outcomes between HAS and fully autonomous agents in at least one domain; otherwise the central empirical claim remains unsupported.
  4. [Sections 3.1, 5, 6.1, and 8] The paper repeatedly relies on the authors' own prior survey [108] as the foundation for the HAS paradigm, its challenges, and its implementation guidelines. Since [108] is a preprint by the same group, citing it does not provide independent validation. The authors should either substantiate key paradigm-defining claims with independent sources or explicitly flag that the paradigm description is based on their own survey, so readers can judge the degree of external support.
minor comments (5)
  1. [Section 6.1] The sentence 'This framework are partitioned five key domains' should read 'This framework is partitioned into five key domains' for grammatical correctness.
  2. [Section 5 (Data subsection)] The phrase 'flexible frameworks that can adopt to this diversity' should be 'adapt to this diversity'; 'adopt' appears to be a typo.
  3. [Section 2] The list item '1.Reliability, Trust, and Safety.' has a formatting error—there is no space after the number—and should be presented as a properly spaced enumerated list, as in '1. Reliability, Trust, and Safety.'
  4. [Section 7.4] The response states that 'response time is one component of the overall value equation,' but V in Equation (1) is defined as value per successful task, not time; clarify how latency is represented in the utility model.
  5. [Figure 2 and Section 6] Figure 2 is referenced as the implementation guideline diagram, but its caption and in-text description do not clearly map the figure's visual elements (e.g., the five domain boxes) to the subsections of Section 6; consider expanding the caption to aid navigation.

Circularity Check

0 steps flagged · score 2.0 of 10

No equation-level circularity: the utility framework in Section 7.1 is definitional scaffolding, the conclusion about current h* is an asserted empirical premise, and the self-citation footprint is minor and not load-bearing.

full rationale

This is a position paper rather than a derivation, and I find no step where a claimed prediction reduces by construction to its inputs. Section 7.1 defines U = V*S - C_h - C_e and h* as the maximizer of U(h), but the paper never estimates V, S, C_h, or C_e and never derives a numerical prediction from the equations. The statement in Section 7.2 that current technological limitations mean the optimal balance still requires significant human participation is an asserted empirical premise about the unknown parameter values, not a consequence of the definitions. The paper itself concedes in Section 5 that standard metrics for human workload and efficiency are lacking, and Section 7.4 acknowledges that human involvement can be a bottleneck for time-critical applications. Those concessions mean the key premise is unquantified rather than hidden. The main self-citation is [108], the authors' own LLM-HAS survey, which is used to name the paradigm, organize the challenge taxonomy, and specify interaction dimensions. These uses are expository and are surrounded by many independent citations, so the central argument does not reduce to the self-cited survey. Under the strict reduction standard, no equation or fitted parameter is repackaged as a prediction, so there is no genuine circularity. The score of 2 reflects only the minor, non-load-bearing self-citation footprint rather than any structural circularity.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The paper's central claim rests on the unmeasured premise that human collaboration improves reliability sufficiently to offset its costs, plus the definitional utility model in Section 7.1. The utility model's variables are listed as free parameters because no values are assigned, and the qualitative assumptions are listed as axioms. No invented entities appear.

free parameters (4)
  • V (value per successful task)
    Introduced in Definition 7.1; no operationalization or measurement is provided.
  • S (success rate)
    Introduced in Definition 7.1; the paper assumes human involvement raises S but never estimates it.
  • C_h (human costs)
    Definition 7.1; no metric for time/effort is specified.
  • C_e (error costs)
    Definition 7.1; the paper argues autonomous agents have high C_e but does not quantify it.
assumptions (4)
  • domain assumption Human involvement increases the success rate S of LLM-based task completion relative to autonomous operation in complex real-world domains.
    Section 3.2 asserts this as an advantage; Section 4 relies on selected examples but provides no controlled comparison.
  • domain assumption Error costs of autonomous agents currently exceed the human operational costs of collaboration.
    Section 7.2-7.5 argue this qualitatively; it is necessary for h* to be significantly greater than zero.
  • ad hoc to paper The utility function U = V*S - C_h - C_e adequately represents the value trade-off for agent systems.
    Definition 7.1; no justification other than plausibility; all subsequent responses use this model.
  • domain assumption Human feedback, despite being noisy, contains contextual knowledge not present in LLMs.
    Section 7.3 states this; it underpins the rejection of fully autonomous learning.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Call for Collaborative Intelligence: Why Human-Agent Systems Should Precede AI Autonomy." pith.science (2026). https://pith.science/paper/DORHIH7P

@misc{pith2026250609420,
  author       = {Pith},
  title        = {Pith review of: A Call for Collaborative Intelligence: Why Human-Agent Systems Should Precede AI Autonomy},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DORHIH7P}},
  note         = {Machine review of arXiv:2506.09420}
}
read the original abstract

Recent improvements in large language models (LLMs) have led many researchers to focus on building fully autonomous AI agents. This position paper questions whether this approach is the right path forward, as these autonomous systems still have problems with reliability, transparency, and understanding the actual requirements of human. We suggest a different approach: LLM-based Human-Agent Systems (LLM-HAS), where AI works with humans rather than replacing them. By keeping human involved to provide guidance, answer questions, and maintain control, these systems can be more trustworthy and adaptable. Looking at examples from healthcare, finance, and software development, we show how human-AI teamwork can handle complex tasks better than AI working alone. We also discuss the challenges of building these collaborative systems and offer practical solutions. This paper argues that progress in AI should not be measured by how independent systems become, but by how well they can work with humans. The most promising future for AI is not in systems that take over human roles, but in those that enhance human capabilities through meaningful partnership.

Figures

Figures reproduced from arXiv: 2506.09420 by the authors.

Figure 1
Figure 1. From Autonomous Agent Systems to Human-Agent Systems. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Implementation guidelines for the Human-Agent Systems. More details in Section [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. The Boundaries of Automation: A Theory of Persistent Human Participation

    cs.AI 2026-07 conditional novelty 6.0 of 10

    Human–AI collaboration may remain necessary even with highly capable AI because in some tasks the evaluative target is constituted through interaction, not fixed in advance.

Reference graph

Works this paper leans on

110 extracted references · 21 canonical work pages · cited by 1 Pith paper

  1. [108]

    A survey on large language model based human-agent systems.arXiv preprint arXiv:2505.00753, 2025

    Henry Peng Zou, Wei-Chieh Huang, Yaozu Wu, Yankai Chen, Chunyu Miao, Hoang Nguyen, Yue Zhou, Weizhi Zhang, Liancheng Fang, Langzhou He, et al. A survey on large language model based human-agent systems.arXiv preprint arXiv:2505.00753, 2025. 16

    arXiv 2025

  2. [1]

    tAIfa: Enhancing Team Effectiveness and Cohesion with AI-Generated Automated Feedback

    Mohammed Almutairi, Charles Chiang, Yuxin Bai, and Diego Gomez-Zara. taifa: Enhanc- ing team effectiveness and cohesion with ai-generated automated feedback.arXiv preprint arXiv:2504.14222, 2025

    work page Pith review arXiv 2025

  3. [2]

    Self-rag: Learning to retrieve, generate, and critique through self-reflection

    Akari Asai, Zeqiu Wu, Yizhong Wang, Avirup Sil, and Hannaneh Hajishirzi. Self-rag: Learning to retrieve, generate, and critique through self-reflection. InThe Twelfth International Conference on Learning Representations, 2023

    2023

  4. [3]

    A general language assistant as a laboratory for alignment.arXiv preprint arXiv:2112.00861, 2021

    Amanda Askell, Yuntao Bai, Anna Chen, Dawn Drain, Deep Ganguli, Tom Henighan, Andy Jones, Nicholas Joseph, Ben Mann, Nova DasSarma, et al. A general language assistant as a laboratory for alignment.arXiv preprint arXiv:2112.00861, 2021

    arXiv 2021

  5. [4]

    Augmenting human-annotated training data with large language model generation and distilla- tion in open-response assessment.arXiv preprint arXiv:2501.09126, 2025

    Conrad Borchers, Danielle R Thomas, Jionghao Lin, Ralph Abboud, and Kenneth R Koedinger. Augmenting human-annotated training data with large language model generation and distilla- tion in open-response assessment.arXiv preprint arXiv:2501.09126, 2025

    arXiv 2025

  6. [5]

    Infrastructure for ai agents.arXiv preprint arXiv:2501.10114, 2025

    Alan Chan, Kevin Wei, Sihao Huang, Nitarshan Rajkumar, Elija Perrier, Seth Lazar, Gillian K Hadfield, and Markus Anderljung. Infrastructure for ai agents.arXiv preprint arXiv:2501.10114, 2025

    arXiv 2025

  7. [6]

    Partnr: A benchmark for planning and reasoning in embodied multi-agent tasks.arXiv preprint arXiv:2411.00081, 2024

    Matthew Chang, Gunjan Chhablani, Alexander Clegg, Mikael Dallaire Cote, Ruta Desai, Michal Hlavac, Vladimir Karashchuk, Jacob Krantz, Roozbeh Mottaghi, Priyam Parashar, et al. Partnr: A benchmark for planning and reasoning in embodied multi-agent tasks.arXiv preprint arXiv:2411.00081, 2024

    arXiv 2024

  8. [7]

    Moore, Frederica F

    Dian Chen, Han Jun Yoon, Zelin Wan, Nithin Alluru, Sang Won Lee, Richard He, Terrence J. Moore, Frederica F. Nelson, Sunghyun Yoon, Hyuk Lim, Dan Dongseong Kim, and Jin-Hee Cho. Advancing human-machine teaming: Concepts, challenges, and applications, 2025

    2025

Show all 110 references
  1. [8]

    End-to-end autonomous driving: Challenges and frontiers.IEEE TPAMI, 46(12):10164–10183, 2024

    Li Chen, Penghao Wu, Kashyap Chitta, Bernhard Jaeger, Andreas Geiger, and Hongyang Li. End-to-end autonomous driving: Challenges and frontiers.IEEE TPAMI, 46(12):10164–10183, 2024

    2024

  2. [9]

    Evaluating large language models trained on code.arXiv preprint arXiv:2107.03374, 2021

    Mark Chen, Jerry Tworek, Heewoo Jun, Qiming Yuan, Henrique Ponde De Oliveira Pinto, Jared Kaplan, Harri Edwards, Yuri Burda, Nicholas Joseph, Greg Brockman, et al. Evaluating large language models trained on code.arXiv preprint arXiv:2107.03374, 2021

    2021 arXiv

  3. [10]

    Xtragpt: Llms for human-ai collaboration on controllable academic paper revision.arXiv preprint arXiv:2505.11336, 2025

    Nuo Chen, Andre Lin HuiKai, Jiaying Wu, Junyi Hou, Zining Zhang, Qian Wang, Xidong Wang, and Bingsheng He. Xtragpt: Llms for human-ai collaboration on controllable academic paper revision.arXiv preprint arXiv:2505.11336, 2025

    2025 arXiv

  4. [11]

    Evaluating human trust in llm-based planners: A preliminary study.arXiv preprint arXiv:2502.20284, 2025

    Shenghui Chen, Yunhao Yang, Kayla Boggess, Seongkook Heo, Lu Feng, and Ufuk Topcu. Evaluating human trust in llm-based planners: A preliminary study.arXiv preprint arXiv:2502.20284, 2025

    2025 arXiv

  5. [12]

    Reinforcing clinical decision sup- port through multi-agent systems and ethical ai governance.arXiv preprint arXiv:2504.03699, 2025

    Ying-Jung Chen, Chi-Sheng Chen, and Ahmad Albarqawi. Reinforcing clinical decision sup- port through multi-agent systems and ethical ai governance.arXiv preprint arXiv:2504.03699, 2025. 10

    2025

  6. [13]

    An exploratory study on how ai awareness impacts human-ai design collaboration

    Zhuoyi Cheng, Pei Chen, Wenzheng Song, Hongbo Zhang, Zhuoshu Li, and Lingyun Sun. An exploratory study on how ai awareness impacts human-ai design collaboration. InProceedings of the 30th International Conference on Intelligent User Interfaces, pages 157–172, 2025

    2025

  7. [14]

    Interpretable risk mitigation in llm agent systems.arXiv preprint arXiv:2505.10670, 2025

    Jan Chojnacki. Interpretable risk mitigation in llm agent systems.arXiv preprint arXiv:2505.10670, 2025

    2025 arXiv

  8. [15]

    Measuring ai agent au- tonomy: Towards a scalable approach with code inspection.arXiv preprint arXiv:2502.15212, 2025

    Peter Cihon, Merlin Stein, Gagan Bansal, Sam Manning, and Kevin Xu. Measuring ai agent au- tonomy: Towards a scalable approach with code inspection.arXiv preprint arXiv:2502.15212, 2025

    2025 arXiv

  9. [16]

    Drive as you speak: Enabling human-like interaction with large language models in autonomous vehicles

    Can Cui, Yunsheng Ma, Xu Cao, Wenqian Ye, and Ziran Wang. Drive as you speak: Enabling human-like interaction with large language models in autonomous vehicles. InProceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, pages 902–909, 2024

    2024

  10. [17]

    The art of designing socially intelligent agents: Science, fiction, and the human in the loop.Applied artificial intelligence, 12(7-8):573–617, 1998

    Kerstin Dautenhahn. The art of designing socially intelligent agents: Science, fiction, and the human in the loop.Applied artificial intelligence, 12(7-8):573–617, 1998

    1998

  11. [18]

    Deconstructing the ethics of large language models from long-standing issues to new-emerging dilemmas: A survey.arXiv preprint arXiv:2406.05392, 2024

    Chengyuan Deng, Yiqun Duan, Xin Jin, Heng Chang, Yijun Tian, Han Liu, Yichen Wang, Kuofeng Gao, Henry Peng Zou, Yiqiao Jin, et al. Deconstructing the ethics of large language models from long-standing issues to new-emerging dilemmas: A survey.arXiv preprint arXiv:2406.05392, 2024

    2024 arXiv

  12. [19]

    Evaluating llms in experiential context: Insights from a survey of recent chi publications

    Christine Dierk, Jennifer Healey, and Mustafa Doga Dogan. Evaluating llms in experiential context: Insights from a survey of recent chi publications. 2025

    2025

  13. [20]

    Llm-evaluation tropes: Perspectives on the validity of llm-evaluations.arXiv preprint arXiv:2504.19076, 2025

    Laura Dietz, Oleg Zendel, Peter Bailey, Charles Clarke, Ellese Cotterill, Jeff Dalton, Faegheh Hasibi, Mark Sanderson, and Nick Craswell. Llm-evaluation tropes: Perspectives on the validity of llm-evaluations.arXiv preprint arXiv:2504.19076, 2025

    2025

  14. [21]

    Citations and trust in llm generated responses

    Yifan Ding, Matthew Facciani, Ellen Joyce, Amrit Poudel, Sanmitra Bhattacharya, Balaji Veeramani, Sal Aguinaga, and Tim Weninger. Citations and trust in llm generated responses. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 23787–23795, 2025

    2025

  15. [22]

    Yu, and Wenpeng Yin

    Jiangshu Du, Yibo Wang, Wenting Zhao, Zhongfen Deng, Shuaiqi Liu, Renze Lou, Henry Peng Zou, Pranav Narayanan Venkit, Nan Zhang, Mukund Srinath, Haoran Ranran Zhang, Vipul Gupta, Yinghui Li, Tao Li, Fei Wang, Qin Liu, Tianlin Liu, Pengzhi Gao, Congying Xia, Chen Xing, Cheng Ji...

    2024

  16. [23]

    A survey on the optimization of large language model-based agents.arXiv preprint arXiv:2503.12434, 2025

    Shangheng Du, Jiabao Zhao, Jinxin Shi, Zhentao Xie, Xin Jiang, Yanhong Bai, and Liang He. A survey on the optimization of large language model-based agents.arXiv preprint arXiv:2503.12434, 2025

    2025

  17. [24]

    Problem solving through human-ai preference-based cooperation.arXiv preprint arXiv:2408.07461, 2024

    Subhabrata Dutta, Timo Kaufmann, Goran Glavaš, Ivan Habernal, Kristian Kersting, Frauke Kreuter, Mira Mezini, Iryna Gurevych, Eyke Hüllermeier, and Hinrich Schuetze. Problem solving through human-ai preference-based cooperation.arXiv preprint arXiv:2408.07461, 2024

    2024 arXiv

  18. [25]

    Large language model-based human-agent collaboration for complex task solving

    Xueyang Feng, Zhi-Yuan Chen, Yujia Qin, Yankai Lin, Xu Chen, Zhiyuan Liu, and Ji-Rong Wen. Large language model-based human-agent collaboration for complex task solving. In Yaser Al-Onaizan, Mohit Bansal, and Yun-Nung Chen, editors,Findings of the Association for Computational...

    2024

  19. [26]

    Large language model-based human-agent collaboration for complex task solving, 2024

    Xueyang Feng, Zhi-Yuan Chen, Yujia Qin, Yankai Lin, Xu Chen, Zhiyuan Liu, and Ji-Rong Wen. Large language model-based human-agent collaboration for complex task solving, 2024. 11

    2024

  20. [27]

    Association for Computational Linguistics

  21. [28]

    From llm reasoning to autonomous ai agents: A comprehensive review.arXiv preprint arXiv:2504.19678, 2025

    Mohamed Amine Ferrag, Norbert Tihanyi, and Merouane Debbah. From llm reasoning to autonomous ai agents: A comprehensive review.arXiv preprint arXiv:2504.19678, 2025

    2025 arXiv

  22. [29]

    Patrick Fernandes, Aman Madaan, Emmy Liu, António Farinhas, Pedro Henrique Martins, Amanda Bertsch, José G. C. de Souza, Shuyan Zhou, Tongshuang Wu, Graham Neubig, and André F. T. Martins. Bridging the gap: A survey on integrating (human) feedback for natural language generati...

    2023

  23. [30]

    A taxonomy for human-llm interaction modes: An initial exploration

    Jie Gao, Simret Araya Gebreegziabher, Kenny Tsu Wei Choo, Toby Jia-Jun Li, Simon Tangi Perrault, and Thomas W Malone. A taxonomy for human-llm interaction modes: An initial exploration. InExtended Abstracts of the CHI Conference on Human Factors in Computing Systems, pages 1–11, 2024

    2024

  24. [31]

    Inherent and emergent liability issues in llm-based agentic systems: a principal-agent perspective.arXiv preprint arXiv:2504.03255, 2025

    Garry A Gabison and R Patrick Xian. Inherent and emergent liability issues in llm-based agentic systems: a principal-agent perspective.arXiv preprint arXiv:2504.03255, 2025

    2025 arXiv

  25. [32]

    Creative problem solving in artificially intelligent agents: A survey and framework.Journal of Artificial Intelligence Research, 75:857–911, 2022

    Evana Gizzi, Lakshmi Nair, Sonia Chernova, and Jivko Sinapov. Creative problem solving in artificially intelligent agents: A survey and framework.Journal of Artificial Intelligence Research, 75:857–911, 2022

    2022

  26. [33]

    Concept-guided llm agents for human-ai safety codesign

    Florian Geissler, Karsten Roscher, and Mario Trapp. Concept-guided llm agents for human-ai safety codesign. InProceedings of the AAAI Symposium Series, volume 3, pages 100–104, 2024

    2024

  27. [34]

    Convcodeworld: Bench- marking conversational code generation in reproducible feedback environments, 2025

    Hojae Han, Seung won Hwang, Rajhans Samdani, and Yuxiong He. Convcodeworld: Bench- marking conversational code generation in reproducible feedback environments, 2025

    2025

  28. [35]

    Convcodeworld: Bench- marking conversational code generation in reproducible feedback environments.arXiv preprint arXiv:2502.19852, 2025

    Hojae Han, Seung-won Hwang, Rajhans Samdani, and Yuxiong He. Convcodeworld: Bench- marking conversational code generation in reproducible feedback environments.arXiv preprint arXiv:2502.19852, 2025

    2025 arXiv

  29. [36]

    Overtrust in ai recom- mendations about whether or not to kill: Evidence from two human-robot interaction studies

    Colin Holbrook, Daniel Holman, Joshua Clingo, and Alan R Wagner. Overtrust in ai recom- mendations about whether or not to kill: Evidence from two human-robot interaction studies. Scientific reports, 14(1):19751, 2024

    2024

  30. [37]

    Llm multi-agent systems: Challenges and open problems.arXiv preprint arXiv:2402.03578, 2024

    Shanshan Han, Qifan Zhang, Yuhang Yao, Weizhao Jin, Zhaozhuo Xu, and Chaoyang He. Llm multi-agent systems: Challenges and open problems.arXiv preprint arXiv:2402.03578, 2024

    2024 arXiv

  31. [38]

    A survey on hallucination in large language models: Principles, taxonomy, challenges, and open questions.ACM Transactions on Information Systems, 43(2):1–55, 2025

    Lei Huang, Weijiang Yu, Weitao Ma, Weihong Zhong, Zhangyin Feng, Haotian Wang, Qiang- long Chen, Weihua Peng, Xiaocheng Feng, Bing Qin, et al. A survey on hallucination in large language models: Principles, taxonomy, challenges, and open questions.ACM Transactions on Informati...

    2025

  32. [39]

    Unified mind model: Reimagining autonomous agents in the llm era.arXiv preprint arXiv:2503.03459, 2025

    Pengbo Hu and Xiang Ying. Unified mind model: Reimagining autonomous agents in the llm era.arXiv preprint arXiv:2503.03459, 2025

    2025 arXiv

  33. [40]

    Beavertails: Towards improved safety alignment of llm via a human-preference dataset.Advances in Neural Information Processing Systems, 36:24678–24704, 2023

    Jiaming Ji, Mickel Liu, Josef Dai, Xuehai Pan, Chi Zhang, Ce Bian, Boyuan Chen, Ruiyang Sun, Yizhou Wang, and Yaodong Yang. Beavertails: Towards improved safety alignment of llm via a human-preference dataset.Advances in Neural Information Processing Systems, 36:24678–24704, 2023

    2023

  34. [41]

    Unpacking dpo and ppo: Disentangling best practices for learning from preference feedback.Advances in neural information processing systems, 37:36602–36633, 2024

    Hamish Ivison, Yizhong Wang, Jiacheng Liu, Zeqiu Wu, Valentina Pyatkin, Nathan Lambert, Noah A Smith, Yejin Choi, and Hanna Hajishirzi. Unpacking dpo and ppo: Disentangling best practices for learning from preference feedback.Advances in neural information processing systems, ...

    2024

  35. [42]

    A review on trust in human-robot interaction.arXiv preprint arXiv:2105.10045, 2021

    Zahra Rezaei Khavas. A review on trust in human-robot interaction.arXiv preprint arXiv:2105.10045, 2021. 12

    2021 arXiv

  36. [43]

    Exploring the impact of an llm-powered teachable agent on learning gains and cognitive load in music education

    Lingxi Jin, Baicheng Lin, Mengze Hong, Kun Zhang, and Hyo-Jeong So. Exploring the impact of an llm-powered teachable agent on learning gains and cognitive load in music education. arXiv preprint arXiv:2504.00636, 2025

    2025 arXiv

  37. [44]

    Beyond turn-taking: Introducing text-based overlap into human-llm interactions.arXiv preprint arXiv:2501.18103, 2025

    JiWoo Kim, Minsuk Chang, and JinYeong Bak. Beyond turn-taking: Introducing text-based overlap into human-llm interactions.arXiv preprint arXiv:2501.18103, 2025

    2025 arXiv

  38. [45]

    Meganno+: A human-llm collaborative annotation system.arXiv preprint arXiv:2402.18050, 2024

    Hannah Kim, Kushan Mitra, Rafael Li Chen, Sajjadur Rahman, and Dan Zhang. Meganno+: A human-llm collaborative annotation system.arXiv preprint arXiv:2402.18050, 2024

    2024 arXiv

  39. [46]

    Mitchell

    Benjamin Klieger, Charis Charitsis, Miroslav Suzara, Sierra Wang, Nick Haber, and John C. Mitchell. Chatcollab: Exploring collaboration between humans and ai agents in software teams, 2024

    2024

  40. [47]

    Applying the gricean maxims to a human-llm interaction cycle: Design insights from a participatory approach.arXiv preprint arXiv:2503.00858, 2025

    Yoonsu Kim, Brandon Chin, Kihoon Son, Seoyoung Kim, and Juho Kim. Applying the gricean maxims to a human-llm interaction cycle: Design insights from a participatory approach.arXiv preprint arXiv:2503.00858, 2025

    2025 arXiv

  41. [48]

    Rlaif: Scaling reinforcement learning from human feedback with ai feedback

    Harrison Lee, Samrat Phatale, Hassan Mansoor, Kellie Ren Lu, Thomas Mesnard, Johan Ferret, Colton Bishop, Ethan Hall, Victor Carbune, and Abhinav Rastogi. Rlaif: Scaling reinforcement learning from human feedback with ai feedback. 2023

    2023

  42. [49]

    Human preferences for constructive interactions in language model alignment.arXiv preprint arXiv:2503.16480, 2025

    Yara Kyrychenko, Jon Roozenbeek, Brandon Davidson, Sander van der Linden, and Ramit Debnath. Human preferences for constructive interactions in language model alignment.arXiv preprint arXiv:2503.16480, 2025

    2025 arXiv

  43. [50]

    Llm-based automated grading with human-in-the-loop.arXiv preprint arXiv:2504.05239, 2025

    Hang Li, Yucheng Chu, Kaiqi Yang, Yasemin Copur-Gencturk, and Jiliang Tang. Llm-based automated grading with human-in-the-loop.arXiv preprint arXiv:2504.05239, 2025

    2025

  44. [51]

    Mmedagent: Learning to use medical tools with multi-modal agent.arXiv preprint arXiv:2407.02483, 2024

    Binxu Li, Tiankai Yan, Yuanting Pan, Jie Luo, Ruiyang Ji, Jiayuan Ding, Zhe Xu, Shilong Liu, Haoyu Dong, Zihao Lin, et al. Mmedagent: Learning to use medical tools with multi-modal agent.arXiv preprint arXiv:2407.02483, 2024

    2024 arXiv

  45. [52]

    Fb-bench: A fine-grained multi-task benchmark for evaluating llms’ responsiveness to human feedback.arXiv preprint arXiv:2410.09412, 2024

    Youquan Li, Miao Zheng, Fan Yang, Guosheng Dong, Bin Cui, Weipeng Chen, Zenan Zhou, and Wentao Zhang. Fb-bench: A fine-grained multi-task benchmark for evaluating llms’ responsiveness to human feedback.arXiv preprint arXiv:2410.09412, 2024

    arXiv

  46. [53]

    Vlfeedback: A large-scale ai feedback dataset for large vision-language models alignment.arXiv preprint arXiv:2410.09421, 2024

    Lei Li, Zhihui Xie, Mukai Li, Shunian Chen, Peiyi Wang, Liang Chen, Yazheng Yang, Benyou Wang, Lingpeng Kong, and Qi Liu. Vlfeedback: A large-scale ai feedback dataset for large vision-language models alignment.arXiv preprint arXiv:2410.09421, 2024

    arXiv

  47. [54]

    Llm-based human-robot collaboration framework for manipulation tasks.arXiv preprint arXiv:2308.14972, 2023

    Haokun Liu, Yaonan Zhu, Kenji Kato, Izumi Kondo, Tadayoshi Aoyama, and Yasuhisa Hasegawa. Llm-based human-robot collaboration framework for manipulation tasks.arXiv preprint arXiv:2308.14972, 2023

    2023 arXiv

  48. [55]

    Du, and Natasha Jaques

    Yancheng Liang, Daphne Chen, Abhishek Gupta, Simon S. Du, and Natasha Jaques. Learning to cooperate with humans using generative agents, 2024

    2024

  49. [56]

    Large language model agent: A survey on methodology, applications and challenges.arXiv preprint arXiv:2503.21460, 2025

    Junyu Luo, Weizhi Zhang, Ye Yuan, Yusheng Zhao, Junwei Yang, Yiyang Gu, Bohan Wu, Binqi Chen, Ziyue Qiao, Qingqing Long, et al. Large language model agent: A survey on methodology, applications and challenges.arXiv preprint arXiv:2503.21460, 2025

    2025 arXiv

  50. [57]

    Large language model-based agents for software engineering: A survey.arXiv preprint arXiv:2409.02977, 2024

    Junwei Liu, Kaixin Wang, Yixuan Chen, Xin Peng, Zhenpeng Chen, Lingming Zhang, and Yiling Lou. Large language model-based agents for software engineering: A survey.arXiv preprint arXiv:2409.02977, 2024

    2024 arXiv

  51. [58]

    Learning autonomous driving tasks via human feedbacks with large language models

    Yunsheng Ma, Xu Cao, Wenqian Ye, Can Cui, Kai Mei, and Ziran Wang. Learning autonomous driving tasks via human feedbacks with large language models. InProc. of EMNLP (Findings), pages 4985–4995, 2024

    2024

  52. [59]

    Sphere: An evaluation card for human-ai systems.arXiv preprint arXiv:2504.07971, 2025

    Qianou Ma, Dora Zhao, Xinran Zhao, Chenglei Si, Chenyang Yang, Ryan Louie, Ehud Reiter, Diyi Yang, and Tongshuang Wu. Sphere: An evaluation card for human-ai systems.arXiv preprint arXiv:2504.07971, 2025

    2025 arXiv

  53. [60]

    Clarigen: Bridging instruction gaps via interactive clarification in code generation

    Chunyu Miao, Yibo Wang, Langzhou He, Liancheng Fang, and Philip S Yu. Clarigen: Bridging instruction gaps via interactive clarification in code generation. InAAAI 2025 Workshop on Preventing and Detecting LLM Misinformation (PDLM)

    2025

  54. [61]

    Improving grounded language understanding in a collaborative environment by interacting with agents through help feedback

    Nikhil Mehta, Milagro Teruel, Xin Deng, Sergio Figueroa Sanz, Ahmed Awadallah, and Julia Kiseleva. Improving grounded language understanding in a collaborative environment by interacting with agents through help feedback. In Yvette Graham and Matthew Purver, editors, Findings ...

    2024

  55. [62]

    An empirical study of the role of incompleteness and ambiguity in interactions with large language models.arXiv preprint arXiv:2503.17936, 2025

    Riya Naik, Ashwin Srinivasan, Estrid He, and Swati Agarwal. An empirical study of the role of incompleteness and ambiguity in interactions with large language models.arXiv preprint arXiv:2503.17936, 2025

    2025 arXiv

  56. [63]

    Fully au- tonomous ai agents should not be developed.arXiv preprint arXiv:2502.02649, 2025

    Margaret Mitchell, Avijit Ghosh, Alexandra Sasha Luccioni, and Giada Pistilli. Fully au- tonomous ai agents should not be developed.arXiv preprint arXiv:2502.02649, 2025

    2025

  57. [64]

    Law-following ai: Designing ai agents to obey human laws

    Cullen O’Keefe, Ketan Ramakrishnan, Janna Tay, and Christoph Winter. Law-following ai: Designing ai agents to obey human laws. 2025

    2025

  58. [65]

    Human-in-the-loop or ai-in-the-loop? automate or collaborate? InProceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 28594–28600, 2025

    Sriraam Natarajan, Saurabh Mathur, Sahil Sidheekh, Wolfgang Stammer, and Kristian Kersting. Human-in-the-loop or ai-in-the-loop? automate or collaborate? InProceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 28594–28600, 2025

    2025

  59. [66]

    Agentcoord: Visually exploring coordination strategy for llm-based multi-agent collaboration.arXiv preprint arXiv:2404.11943, 2024

    Bo Pan, Jiaying Lu, Ke Wang, Li Zheng, Zhen Wen, Yingchaojie Feng, Minfeng Zhu, and Wei Chen. Agentcoord: Visually exploring coordination strategy for llm-based multi-agent collaboration.arXiv preprint arXiv:2404.11943, 2024

    2024 arXiv

  60. [67]

    Ethical and regulatory challenges of large language models in medicine.The Lancet Digital Health, 6(6):e428–e432, 2024

    Jasmine Chiat Ling Ong, Shelley Yin-Hsi Chang, Wasswa William, Atul J Butte, Nigam H Shah, Lita Sui Tjien Chew, Nan Liu, Finale Doshi-Velez, Wei Lu, Julian Savulescu, et al. Ethical and regulatory challenges of large language models in medicine.The Lancet Digital Health, 6(6):...

    2024

  61. [68]

    Emoagent: Assessing and safeguarding human-ai interaction for mental health safety.arXiv preprint arXiv:2504.09689, 2025

    Jiahao Qiu, Yinghui He, Xinzhe Juan, Yiming Wang, Yuhan Liu, Zixin Yao, Yue Wu, Xun Jiang, Ling Yang, and Mengdi Wang. Emoagent: Assessing and safeguarding human-ai interaction for mental health safety.arXiv preprint arXiv:2504.09689, 2025

    2025 arXiv

  62. [69]

    Ai agents in modern healthcare: From foundation to pioneer–a comprehensive review and implementation roadmap for impact and integration in clinical settings

    Dhavalkumar Patel, Ganesh Raut, Satya Narayan Cheetirala, Benjamin Glicksberg, Matthew A Levin, Girish Nadkarni, Robert Freeman, Eyal Klang, and Prem Timsina. Ai agents in modern healthcare: From foundation to pioneer–a comprehensive review and implementation roadmap for impac...

    2025

  63. [70]

    Personagym: Evaluating persona agents and llms.arXiv preprint arXiv:2407.18416, 2024

    Vinay Samuel, Henry Peng Zou, Yue Zhou, Shreyas Chaudhari, Ashwin Kalyan, Tanmay Rajpurohit, Ameet Deshpande, Karthik Narasimhan, and Vishvak Murahari. Personagym: Evaluating persona agents and llms.arXiv preprint arXiv:2407.18416, 2024

    2024 arXiv

  64. [71]

    Llm-based agentic systems in medicine and healthcare.Nature Machine Intelligence, 6(12):1418–1420, 2024

    Jianing Qiu, Kyle Lam, Guohao Li, Amish Acharya, Tien Yin Wong, Ara Darzi, Wu Yuan, and Eric J Topol. Llm-based agentic systems in medicine and healthcare.Nature Machine Intelligence, 6(12):1418–1420, 2024

    2024

  65. [72]

    Collaborative gym: A framework for enabling and evaluating human-agent collaboration.arXiv preprint arXiv:2412.15701, 2024

    Yijia Shao, Vinay Samuel, Yucheng Jiang, John Yang, and Diyi Yang. Collaborative gym: A framework for enabling and evaluating human-agent collaboration.arXiv preprint arXiv:2412.15701, 2024

    2024

  66. [73]

    Reveca: Adaptive planning and trajectory-based validation in cooperative language agents using information relevance and relative proximity

    SeungWon Seo, SeongRae Noh, Junhyeok Lee, SooBin Lim, Won Hee Lee, and HyeongYeop Kang. Reveca: Adaptive planning and trajectory-based validation in cooperative language agents using information relevance and relative proximity. InProceedings of the AAAI Conference on Artifici...

    2025

  67. [74]

    Role & responsibility charting (raci)

    Michael L Smith, James Erwin, and Sandra Diaferio. Role & responsibility charting (raci). In Project Management Forum (PMForum), volume 5, page 12, 2005

    2005

  68. [75]

    From mind to machine: The rise of manus ai as a fully autonomous digital agent.arXiv preprint arXiv:2505.02024, 2025

    Minjie Shen and Qikai Yang. From mind to machine: The rise of manus ai as a fully autonomous digital agent.arXiv preprint arXiv:2505.02024, 2025

    2025 arXiv

  69. [76]

    Aligning large multimodal models with factually augmented rlhf

    Zhiqing Sun, Sheng Shen, Shengcao Cao, Haotian Liu, Chunyuan Li, Yikang Shen, Chuang Gan, Liangyan Gui, Yu-Xiong Wang, Yiming Yang, et al. Aligning large multimodal models with factually augmented rlhf. InACL (Findings), 2024

    2024

  70. [77]

    Prototyp- ing with prompts: Emerging approaches and challenges in generative ai design for collaborative software teams

    Hari Subramonyam, Divy Thakkar, Andrew Ku, Juergen Dieber, and Anoop K Sinha. Prototyp- ing with prompts: Emerging approaches and challenges in generative ai design for collaborative software teams. InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems...

    2025

  71. [78]

    To help or not to help: Llm- based attentive support for human-robot group interactions

    Daniel Tanneberg, Felix Ocker, Stephan Hasler, Joerg Deigmoeller, Anna Belardinelli, Chao Wang, Heiko Wersing, Bernhard Sendhoff, and Michael Gienger. To help or not to help: Llm- based attentive support for human-robot group interactions. In2024 IEEE/RSJ International Confere...

    2024

  72. [79]

    Multi-agent collaboration: Harnessing the power of intelligent llm agents.arXiv preprint arXiv:2306.03314, 2023

    Yashar Talebirad and Amirhossein Nadiri. Multi-agent collaboration: Harnessing the power of intelligent llm agents.arXiv preprint arXiv:2306.03314, 2023

    2023 arXiv

  73. [80]

    Cybermetric: A benchmark dataset for evaluating large language models knowledge in cybersecurity.arXiv preprint arXiv:2402.07688, 2024

    Norbert Tihanyi, Mohamed Amine Ferrag, Ridhi Jain, and Merouane Debbah. Cybermetric: A benchmark dataset for evaluating large language models knowledge in cybersecurity.arXiv preprint arXiv:2402.07688, 2024

    2024 arXiv

  74. [81]

    Moral alignment for llm agents.arXiv preprint arXiv:2410.01639, 2024

    Elizaveta Tennant, Stephen Hailes, and Mirco Musolesi. Moral alignment for llm agents.arXiv preprint arXiv:2410.01639, 2024

    2024 arXiv

  75. [82]

    A survey on responsible llms: Inherent risk, malicious use, and mitigation strategy.arXiv preprint arXiv:2501.09431, 2025

    Huandong Wang, Wenjie Fu, Yingzhou Tang, Zhilong Chen, Yuxi Huang, Jinghua Piao, Chen Gao, Fengli Xu, Tao Jiang, and Yong Li. A survey on responsible llms: Inherent risk, malicious use, and mitigation strategy.arXiv preprint arXiv:2501.09431, 2025

    2025 arXiv

  76. [83]

    Internalized self-correction for large language models.arXiv preprint arXiv:2412.16653, 2024

    Nishanth Upadhyaya and Raghavendra Sridharamurthy. Internalized self-correction for large language models.arXiv preprint arXiv:2412.16653, 2024

    2024 arXiv

  77. [84]

    A survey of llm-based agents in medicine: How far are we from baymax?arXiv preprint arXiv:2502.11211, 2025

    Wenxuan Wang, Zizhan Ma, Zheng Wang, Chenghan Wu, Wenting Chen, Xiang Li, and Yixuan Yuan. A survey of llm-based agents in medicine: How far are we from baymax?arXiv preprint arXiv:2502.11211, 2025

    2025 arXiv

  78. [85]

    A survey on large language model based autonomous agents.Frontiers of Computer Science, 18(6):186345, 2024

    Lei Wang, Chen Ma, Xueyang Feng, Zeyu Zhang, Hao Yang, Jingsen Zhang, Zhiyuan Chen, Jiakai Tang, Xu Chen, Yankai Lin, et al. A survey on large language model based autonomous agents.Frontiers of Computer Science, 18(6):186345, 2024

    2024

  79. [86]

    MINT: Evaluating LLMs in multi-turn interaction with tools and language feedback

    Xingyao Wang, Zihan Wang, Jiateng Liu, Yangyi Chen, Lifan Yuan, Hao Peng, and Heng Ji. MINT: Evaluating LLMs in multi-turn interaction with tools and language feedback. InThe Twelfth International Conference on Learning Representations, 2024

    2024

  80. [87]

    Mint: Evaluating llms in multi-turn interaction with tools and language feedback.arXiv preprint arXiv:2309.10691, 2023

    Xingyao Wang, Zihan Wang, Jiateng Liu, Yangyi Chen, Lifan Yuan, Hao Peng, and Heng Ji. Mint: Evaluating llms in multi-turn interaction with tools and language feedback.arXiv preprint arXiv:2309.10691, 2023

    2023 arXiv

  81. [88]

    Llm agents making agent tools.arXiv preprint arXiv:2502.11705, 2025

    Georg Wölflein, Dyke Ferber, Daniel Truhn, Ognjen Arandjelovi´c, and Jakob Nikolas Kather. Llm agents making agent tools.arXiv preprint arXiv:2502.11705, 2025

    2025 arXiv

  82. [89]

    Is rlhf more difficult than standard rl? a theoretical perspective.Advances in Neural Information Processing Systems, 36:76006–76032, 2023

    Yuanhao Wang, Qinghua Liu, and Chi Jin. Is rlhf more difficult than standard rl? a theoretical perspective.Advances in Neural Information Processing Systems, 36:76006–76032, 2023

    2023

  83. [90]

    Multi-agent autonomous driving systems with large language models: A survey of recent advances.arXiv preprint arXiv:2502.16804, 2025

    Yaozu Wu, Dongyuan Li, Yankai Chen, Renhe Jiang, Henry Peng Zou, Liancheng Fang, Zhen Wang, and Philip S Yu. Multi-agent autonomous driving systems with large language models: A survey of recent advances.arXiv preprint arXiv:2502.16804, 2025

    2025

  84. [91]

    Ai chains: Transparent and controllable human-ai interaction by chaining large language model prompts

    Tongshuang Wu, Michael Terry, and Carrie Jun Cai. Ai chains: Transparent and controllable human-ai interaction by chaining large language model prompts. InProceedings of the 2022 CHI Conference on Human Factors in Computing Systems, CHI ’22, New York, NY , USA,

    2022

  85. [92]

    Infllm: Training-free long-context extrapolation for llms with an efficient context memory

    Chaojun Xiao, Pengle Zhang, Xu Han, Guangxuan Xiao, Yankai Lin, Zhengyan Zhang, Zhiyuan Liu, and Maosong Sun. Infllm: Training-free long-context extrapolation for llms with an efficient context memory. InThe Thirty-eighth Annual Conference on Neural Information Processing Syst...

    2024

  86. [93]

    Llm a*: Human in the loop large language models enabled a* search for robotics.arXiv preprint arXiv:2312.01797, 2023

    Hengjia Xiao and Peng Wang. Llm a*: Human in the loop large language models enabled a* search for robotics.arXiv preprint arXiv:2312.01797, 2023

    2023 arXiv

  87. [94]

    An evaluation-driven approach to designing llm agents: Process and architecture.arXiv preprint arXiv:2411.13768, 2024

    Boming Xia, Qinghua Lu, Liming Zhu, Zhenchang Xing, Dehai Zhao, and Hao Zhang. An evaluation-driven approach to designing llm agents: Process and architecture.arXiv preprint arXiv:2411.13768, 2024

    2024

  88. [95]

    Finarena: A human-agent collaboration framework for financial market analysis and forecasting.arXiv preprint arXiv:2503.02692, 2025

    Congluo Xu, Zhaobin Liu, and Ziyang Li. Finarena: A human-agent collaboration framework for financial market analysis and forecasting.arXiv preprint arXiv:2503.02692, 2025

    2025 arXiv

  89. [96]

    Nuclear deployed: Analyzing catastrophic risks in decision-making of autonomous llm agents.arXiv preprint arXiv:2502.11355, 2025

    Rongwu Xu, Xiaojian Li, Shuo Chen, and Wei Xu. Nuclear deployed: Analyzing catastrophic risks in decision-making of autonomous llm agents.arXiv preprint arXiv:2502.11355, 2025

    2025 arXiv

  90. [97]

    Large multimodal agents: A survey.arXiv preprint arXiv:2402.15116, 2024

    Junlin Xie, Zhihong Chen, Ruifei Zhang, Xiang Wan, and Guanbin Li. Large multimodal agents: A survey.arXiv preprint arXiv:2402.15116, 2024

    2024 arXiv

  91. [98]

    The lighthouse of language: Enhancing llm agents via critique-guided improvement.arXiv preprint arXiv:2503.16024, 2025

    Ruihan Yang, Fanghua Ye, Jian Li, Siyu Yuan, Yikai Zhang, Zhaopeng Tu, Xiaolong Li, and Deqing Yang. The lighthouse of language: Enhancing llm agents via critique-guided improvement.arXiv preprint arXiv:2503.16024, 2025

    2025

  92. [99]

    Wong, Junchao Wu, and Lidia S

    Xinyi Yang, Runzhe Zhan, Derek F. Wong, Junchao Wu, and Lidia S. Chao. Human-in-the-loop machine translation with large language model, 2023

    2023

  93. [100]

    Supercorrect: Advancing small llm reasoning with thought template distillation and self-correction.arXiv preprint arXiv:2410.09008, 2024

    Ling Yang, Zhaochen Yu, Tianjun Zhang, Minkai Xu, Joseph E Gonzalez, Bin Cui, and Shuicheng Yan. Supercorrect: Advancing small llm reasoning with thought template distillation and self-correction.arXiv preprint arXiv:2410.09008, 2024

    arXiv

  94. [101]

    Yangyang Yu, Zhiyuan Yao, Haohang Li, Zhiyang Deng, Yuechen Jiang, Yupeng Cao, Zhi Chen, Jordan Suchow, Zhenyu Cui, Rong Liu, et al. Fincon: A synthesized llm multi-agent system with conceptual verbal reinforcement for enhanced financial decision making.Advances in Neural Info...

    2024

  95. [102]

    Leveraging dual process theory in lan- guage agent framework for real-time simultaneous human-ai collaboration.arXiv preprint arXiv:2502.11882, 2025

    Shao Zhang, Xihuai Wang, Wenhao Zhang, Chaoran Li, Junru Song, Tingyu Li, Lin Qiu, Xuezhi Cao, Xunliang Cai, Wen Yao, et al. Leveraging dual process theory in lan- guage agent framework for real-time simultaneous human-ai collaboration.arXiv preprint arXiv:2502.11882, 2025

    2025 arXiv

  96. [103]

    Aligning multimodal llm with human preference: A survey.arXiv preprint arXiv:2503.14504, 2025

    Tao Yu, Yi-Fan Zhang, Chaoyou Fu, Junkang Wu, Jinda Lu, Kun Wang, Xingyu Lu, Yunhang Shen, Guibin Zhang, Dingjie Song, et al. Aligning multimodal llm with human preference: A survey.arXiv preprint arXiv:2503.14504, 2025

    2025 arXiv

  97. [104]

    A survey on the memory mechanism of large language model based agents.arXiv preprint arXiv:2404.13501, 2024

    Zeyu Zhang, Xiaohe Bo, Chen Ma, Rui Li, Xu Chen, Quanyu Dai, Jieming Zhu, Zhenhua Dong, and Ji-Rong Wen. A survey on the memory mechanism of large language model based agents.arXiv preprint arXiv:2404.13501, 2024

    2024 arXiv

  98. [105]

    Lifelong learning of large language model based agents: A roadmap.arXiv preprint arXiv:2501.07278, 2025

    Junhao Zheng, Chengming Shi, Xidi Cai, Qiuke Li, Duzhen Zhang, Chenxing Li, Dong Yu, and Qianli Ma. Lifelong learning of large language model based agents: A roadmap.arXiv preprint arXiv:2501.07278, 2025

    2025

  99. [106]

    Characterizing unintended consequences in human-gui agent collaboration for web browsing.arXiv preprint arXiv:2505.09875, 2025

    Shuning Zhang, Jingruo Chen, Jiajing Gao, Zhiqi Gao, Xin Yi, and Hewu Li. Characterizing unintended consequences in human-gui agent collaboration for web browsing.arXiv preprint arXiv:2505.09875, 2025

    2025 arXiv

  100. [107]

    Large language models are involuntary truth-tellers: Exploiting fallacy failure for jailbreak attacks

    Yue Zhou, Henry Zou, Barbara Di Eugenio, and Yang Zhang. Large language models are involuntary truth-tellers: Exploiting fallacy failure for jailbreak attacks. InProceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 13293–13304, 2024

    2024

  101. [109]

    Sweet-rl: Training multi-turn llm agents on collaborative reasoning tasks, 2025

    Yifei Zhou, Song Jiang, Yuandong Tian, Jason Weston, Sergey Levine, Sainbayar Sukhbaatar, and Xian Li. Sweet-rl: Training multi-turn llm agents on collaborative reasoning tasks, 2025

    2025

  102. [2022]

    Association for Computing Machinery

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.