Science AI Journal
Computer Science · Research topic

Open research questions in Multi-Agent Systems and Negotiation

117 unresolved questions extracted from the limitations and future-work sections of 620 Multi-Agent Systems and Negotiation papers in our library. Each links back to the study that raised it.

What the literature leaves open

  • Initial implementation burden. System integration, as trials often span multiple legacy platforms that lack interoperability. Human judgment boundaries, as there is a need to specify which tasks agents may perform autonomously and which require human oversight.

    Towards a multi-agent Clinical Trial Center framework to support modern clinical trials · 2026 · DOI

  • Traditional workflows in Clinical Trial Centers are labor-intensive and error-prone. Existing AI handles narrow and siloed tasks. There is a need for a framework that can streamline trial execution, reduce coordination burden, and enhance data quality while maintaining human oversight and regulatory compliance.

    Towards a multi-agent Clinical Trial Center framework to support modern clinical trials · 2026 · DOI

  • Secure multi-party computation problems and their applications: a review and open problems.

    Trustworthy Agent Network: Trust in Agent Networks Must Be Baked In, Not Bolted On · 2026

  • • If no consensus exists yet, provide your own reasoning, but state that you are open to changing your mind.

    Conflict-Resilient Multi-Agent Reasoning via Signed Graph Modeling · 2026

  • Current studies mostly focus on isolated attack strategies, lacking multi-agent coordination mechanisms. There is a need to explore attack methods capable of handling cross-agent behaviors and model-specific vulnerabilities. The existing methods have limited attack effectiveness and fail to fully exploit the potential security vulnerabilities of LLMs.

    Breaking Barriers: Multi-Agent Prompt Fusion for Automated LLM Jailbreaks · 2026 · DOI

  • Current agent systems research has focused primarily on agent reasoning capabilities, with less attention paid to the engineering challenges of building reliable production pipelines. The deployment of autonomous AI agents in enterprise environments presents significant architectural challenges.

    Design and Implementation of Agentic AI Pipelines for Enterprise Decision-Making Architecture Patterns for Production Systems · 2026 · DOI

  • Further development of ACI and its components, such as QBD and CNS. Exploration of the applications and implications of ACI in various domains.

    Agentic Convergence Infrastructure™ (ACI™) · 2026 · DOI

  • The industry is recognizing the need for competing trust and authorization models, verifiable intent, and cross-platform orchestration. Autonomous systems are scaling faster than the coordination structures required to govern them.

    Agentic Convergence Infrastructure™ (ACI™) · 2026 · DOI

  • The lack of interpretability and verifiable grounding in large language models is a significant challenge in legal document analysis. Traditional keyword search systems are insufficient for semantic reasoning.

    A Multi-Agent Retrieval-Augmented Generation Framework for Context-Aware Legal Document Analysis · 2026 · DOI

  • The current framework depends on the quality of embed-dings and the availability of legal datasets. External precedent retrieval may be affected by API limitations. Future work includes expanding datasets, optimizing agent communication, and integrating domain-specific fine-tuned models.

    A Multi-Agent Retrieval-Augmented Generation Framework for Context-Aware Legal Document Analysis · 2026 · DOI

  • Existing evals have expressivity gaps in domain-customized datasets and domain-specific query types. Existing data analysis agents fail on stateful and incident-specific queries.

    Generating Expressive and Customizable Evals for Timeseries Data Analysis Agents with AgentFuel · 2026 · DOI

  • We view AgentFuel as a first but significant step in advancing evals for timeseries analysis agents. We conclude by acknowledging several limitations and directions for improvement. First, we focus only on the analysis workflow and ignore other issues in data 9 wrangling or data cleaning. Second, we focus on one-shot agents that answer the question as posed; a natural direction is to extend AgentFuel to agents that can ask questions to refine the intent. Third, we showed the value of AgentFuel as a testing framework. As future work, we plan to provide more fine-grained debugging (e.g., possibly connecting with backend tracing) to understand failure patterns and continuous training.

    Generating Expressive and Customizable Evals for Timeseries Data Analysis Agents with AgentFuel · 2026 · DOI

  • The paper identifies the need for governance by construction to ensure safety and compliance. The gap is the lack of a principled governance layer for generalist computer-using agents.

    Governance by Construction for Generalist Agents · 2026 · DOI

  • To validate SEED as a finished system - To estimate population effects - To develop governance principles for responsible use of SEED-assisted design

    Agents for Experiments, Experiments for Agents: A Design Grammar for AI-Enabled Experimental Science · 2026

  • The lack of a compact language for describing experimental conditions - The need for a framework that can guide AI-assisted generation of candidate experimental conditions

    Agents for Experiments, Experiments for Agents: A Design Grammar for AI-Enabled Experimental Science · 2026

  • We further outline open challenges for harness engineering, including evaluation beyond final task success, verification under incomplete feedback, regression-free harness improvement, consistent shared state across multiple agents, human oversight for safety-critical actions, and extensions to multimodal environments.

    Code as Agent Harness · 2026

  • Yet the reliability of these judges for deep research agents remains poorly understood, posing a critical meta-evaluation problem: before deploying LLM judges to supervise research agents, we must first evaluate the judges themselves.

    Time to REFLECT: Can We Trust LLM Judges for Evidence-based Research Agents? · 2026

  • The absence of unified, role-structured orchestration frameworks. The lack of formal collaboration methods and clear task assignments. The need for cooperative AI systems for software engineering.

    Open AGI — Multi-Agent Automation Framework for AI-Orchestrated Software Development · 2026 · DOI

  • The need for modular procedural memory in multi-agent LLM systems. The lack of a systematic study of procedural memory in multi-agent systems.

    LEGOMem: Modular Procedural Memory for Multi-agent LLM Systems for Workflow Automation · 2026 · DOI

  • LegoNE currently depends on human-curated building blocks, - the framework’s scope is limited to algorithm analysis problems where the instantiation and forgetting principles apply

    Discovering expert-level Nash equilibrium algorithms with large language models · 2026 · DOI

  • developing more expressive languages that allow algorithms to be analyzed from first principles, - broadening the range of problems amenable to automated analysis

    Discovering expert-level Nash equilibrium algorithms with large language models · 2026 · DOI

  • Current agent architectures lack explicit mechanisms to autonomously evolve their own requirements and code. Limited support for structured, long-term evolution.

    Self-Evolving Software Agents · 2026 · DOI

  • This paper introduced a framework for self-evolving software agents that integrates automated software evolution principles within a BDI architecture augmented by LLMs. By separating runtime reasoning from an explicit evolution module, the proposed approach enables agents to autonomously revise goals, reasoning structures, and executable actions, moving beyond traditional notions of behavioural adaptation. The prototype and preliminary evaluation demonstrate the feasibility of autonomous evolution in dynamic multi-agent environments, while also revealing current limitations in behavioural inheritance, stability, and scalability. These challenges point to the need for reinforcement mechanisms, memory consolidation, and more robust selection strategies. Future work will focus on strengthening inheritance and longterm consistency, extending the evaluation to more complex multiagent scenarios, and exploring collective and cooperative forms of software evolution among agents, as well as leveraging retrievalaugmented generation to enhance LLM reasoning with external knowledge. REFERENCES L. Bettini. 2015. Implementing Domain-Specific Languages with Xtext and Xtend. Packt Publishing, Birmingham, UK. B. W. Boehm. 1988. A spiral model of software development and enhancement. ACM SIGSOFT Software Engineering Notes 11, 4 (1988), 14–24. M. Böhm and A. Zimmermann. 2020. The Autonomous System Dilemma: Balancing Adaptability and Predictability. IEEE Software 37, 4 (2020), 44–49. R. et al. Bommasani. 2021. On the opportunities and risks of foundation models. arXiv preprint arXiv:2108.07258 (2021), e220119. Tom B Brown et al. 2020. Language Models are Few-Shot Learners. Advances in Neural Information Processing Systems 33 (2020), 1877–1901. J. M. Burge and D. C. Brown. 1999. Software change: Cost, causes, and complexity. Software Engineering Journal 14, 3 (1999), 180–190. Mark Chen et al. 2021. Evaluating Large Language Models Trained on Code. arXiv preprint arXiv:2107.03374 (2021). B. H. Cheng, H. Giese, P. Inverardi, and J. Magee. 2009. Software Engineering for Self-Adaptive Systems: A Research Roadmap. Software Engineering for Self- Adaptive Systems (2009), 1–26. T. H. Davenport and R. Kalakota. 2019. The potential for artificial intelligence in healthcare. Future Healthcare Journal 6, 2 (2019), 94–98. R. de Lemos, H. Giese, H. A. Müller, and M. Shaw. 2001. Self-adaptive software: Landscape and research challenges. ACM Transactions on Autonomous and Adaptive Systems 4, 2 (2001), 1–25. Juan Fernandez-Ramil, Dewayne Perry, and Nazim H. Madhavji (Eds.). 2006. Software Evolution and Feedback: Theory and Practice. Wiley, Chichester. S. Franklin and A. Graesser. 1996. Is it an agent, or just a program?: A taxonomy for autonomous agents. In Proceedings of the International Workshop on Agent Theories, Architectures, and Languages. Springer, Berlin, Heidelberg, 21–35. D.

    Self-Evolving Software Agents · 2026 · DOI

  • The paper suggests that future research could focus on applying the splitting approach to other types of argumentation frameworks. The paper notes that future research could also explore the relationship between the splitting approach and other techniques for addressing computational intractability.

    Splitting Assumption-Based Argumentation Frameworks · 2026 · DOI

  • The paper identifies the computational intractability of core reasoning tasks in ABA as a research gap. The paper notes that existing approaches to addressing this gap are limited.

    Splitting Assumption-Based Argumentation Frameworks · 2026 · DOI

Most-cited papers in Multi-Agent Systems and Negotiation

Most recent work

Find a gap in your own Multi-Agent Systems and Negotiation sub-topic

This page shows what the Multi-Agent Systems and Negotiation literature already flags as unresolved. To narrow it to your specific question, run the guided finder — it searches the gap library on demand and checks candidates against 250M+ OpenAlex works.

Open the Research Gap Finder →

Related topics in Computer Science

Online Learning and Analytics
1,106 open questions
AI in Service Interactions
861 open questions
Quantum Computing Algorithms and Architecture
645 open questions
Teaching and Learning Programming
623 open questions
Digital literacy in education
617 open questions
Music Technology and Sound Studies
576 open questions
Virtual Reality Applications and Impacts
496 open questions
Topic Modeling
476 open questions
117 open questions have been extracted from the limitations and future-work passages of 620 Multi-Agent Systems and Negotiation papers in our library. Each one below links back to the study that raised it, so you can read the original claim in context.

Tools for your next paper

Gaps & Literature SearchFind open questions in any field — 120,000+ open gaps mined from our own library.Pre-CheckIs your paper ready? A Tier 1–5 acceptance probability in seconds.Journal FinderA ranked shortlist of target journals from a 17,500-venue index.Duplicate Publication CheckerCheck prior-publication & salami-slicing risk across 8 sources.Citation GeneratorFreeA DOI or title → APA, MLA, Vancouver, Chicago, BibTeX, RIS. No signup.Graphical Abstract MakerFreeTurn your findings into a colorblind-safe graphical abstract.

Compare the category — Honest roundups of the AI research tools, ours listed alongside the alternatives.

Command palette

Jump anywhere, run any action.