Research Track
Track at a Glance
Call · 01
Overview
As AI agents move from controlled demonstrations into live scientific, technical, economic, and social environments, DAI 2026 aims to provide a focused international forum for research that advances the foundations, engineering, evaluation, deployment, and governance of distributed and agentic intelligence.
The Research Track welcomes work across theory, algorithms, systems, applications, and responsible deployment. We especially encourage submissions that connect classical questions in distributed AI and multi-agent systems with emerging challenges in large language model agents, self-improving systems, agent infrastructure, AI for science, embodied multi-agent systems, and human-agent societies.
Call · 02
Important Dates
All deadlines are Anywhere on Earth (AoE), UTC-12.
| Milestone | Date | Notes |
|---|---|---|
| Submission system opens | OpenReview opens for abstract registration and paper uploads | |
| Abstract registration deadline | Required before full-paper submission | |
| Research Track submission deadline | Full research paper due | |
| Review period | Compact review process | |
| Research Track notification | Author decisions | |
| Camera-ready deadline | Accepted papers | |
| Early registration deadline | - | |
| Late registration deadline | - | |
| Conference | City University of Hong Kong |
Call · 03
Submission Format
Research Track submissions must be written in English and submitted as PDFs through OpenReview. Authors should prepare manuscripts using the ACM LaTeX template. Two-column submissions using the ACM sigconf option are acceptable.
Research Track papers should be full papers of up to 8 pages, excluding references. Papers may include an unlimited appendix after the bibliography; appendices are optional reading for reviewers. Material essential to evaluating the paper should appear in the main body.
Submissions should clearly state their contributions, explain their relevance to DAI, situate the work in relation to prior research, and provide appropriate theoretical, empirical, experimental, or system-level evidence.
Call · 04
Scope and Topics of Interest
Topics of interest include, but are not limited to, the following areas. Authors will be asked to select one or more relevant areas during submission.
Agent Engineering & Infrastructure
- Agent frameworks, harnesses, and operating systems: LangGraph, AutoGen, CrewAI, Smolagents, OpenAI Agents SDK
- Memory architectures: long-term, short-term, episodic, factual, and experiential memory
- Skill acquisition and atomic skills
- Tool use: tool selection, grounding, and reliability
- Context engineering: context windows, compression, and selective retrieval
- Agent protocols: MCP, A2A, and interoperability standards
- AgentOps: observability, debugging, evaluation, and failure recovery
- Agent identity, reputation, and provenance
Foundations of Agent Learning
- Reinforcement learning, multi-agent reinforcement learning, and cooperative or competitive learning
- Post-training for agents
- Self-play, curriculum, and open-ended learning
- Continual learning, meta-learning, and transfer
- Reward design and credit assignment
- Distributed, privacy-preserving, and collaborative learning
- Scaling laws and empirical theory of agent learning
Self-Evolving & Self-Improving Agents
- Self-improvement and recursive self-modification
- Meta-reasoning and self-reflection
- Experience distillation: from trajectories to transferable knowledge
- Co-evolution of policies and critics
- Multi-agent evolutionary systems
- Gödel-style self-rewriting agents
- Benchmarks and evaluation for self-evolving systems
Multi-Agent Cooperation & Human-Agent Interaction
- Cooperative multi-agent reinforcement learning, credit assignment, and teamwork
- Communication, language emergence, and negotiation
- LLM-based multi-agent orchestration
- Ad-hoc teamwork and zero-shot coordination
- Coalition formation and distributed problem solving
- Collective intelligence and swarm behavior
- Trust, explainability, and accountability
- AI agents as digital employees, collaborators, and competitors
- Human-agent and human-robot interaction
- Agent-based human interaction analysis
- Agents for enhancing human cooperation
Game Theory, Economics & Agent Markets
- Algorithmic game theory and equilibrium computation
- Mechanism and market design, auctions, and social choice
- Strategic behavior of LLM agents and algorithmic collusion
- Machine-payable APIs and agent-to-agent transactions
- Contract theory and principal-agent models for AI
- Blockchain economics and decentralized systems
- Behavioral game models and bounded rationality
- Security games
Embodied Multi-Agent Systems
- Multi-robot learning, coordination, and swarms
- Vision-Language-Action models for agent teams
- World models for multi-agent planning
- Sim-to-real transfer in multi-agent settings
- Heterogeneous embodied teams
- Safety layers and hardware-software co-design for physical agents
Science of AI & AI for Science
Science of AI
- Evaluation, benchmarking, and reproducibility of agent systems
- Interpretability of multi-agent LLM systems
- Emergent behavior, scaling laws, and phase transitions
- Failure modes, red-teaming, and safety evaluation
- Theoretical foundations of agentic AI
AI for Science
- AI agents for scientific discovery
- AI agents in mathematics, physics, chemistry, biology, and materials
- Automated experiment design and execution
- Scientific literature understanding and hypothesis generation
- Human-agent collaborative research
- Agent-based simulation of societies
- Policy, governance, and alignment of agent collectives
Submissions that do not fit neatly into these categories but advance the goals of distributed and agentic AI are welcome.
Call · 05
Review Process
The Research Track will use double-blind peer review. Reviewers will not know the identities of authors during review, and authors will not know reviewer identities. Submissions must be anonymous: authors should remove names, affiliations, acknowledgments, and other identifying information from the submitted paper and supplementary material. Papers containing identifying information after the submission deadline may be rejected without review.
- Review period: 6 weeks
- Final decisions will be made by the program committee based on reviewer assessments, meta-reviews, and program-level deliberation.
Submissions will be evaluated according to criteria including originality, soundness, relevance to DAI, significance, quality of presentation, understanding of the state of the art, reproducibility, and ethical and societal considerations.
Submissions will be managed through OpenReview. Detailed author instructions and reviewer guidelines will be provided in the Author Guide and Reviewer Guide.
Call · 06
Publication and Presentation
Accepted Research Track papers will be included in the DAI 2026 proceedings, published through the ACM International Conference Proceedings Series (ICPS), and made available in the ACM Digital Library.
The ACM open access publishing model for ICPS applies. Authors based at institutions that are not part of the ACM Open program and who do not qualify for a waiver may be required to pay an article processing charge (APC) to publish their ICPS article in the ACM Digital Library.
Authors should consult ACM's ICPS author guidance, the ICPS publishing model FAQ, and the ACM Open program details for current publication and APC information. Questions about the ACM publishing model should be directed to [email protected].
Accepted papers will be presented at the conference as oral presentations, spotlight presentations, posters, or other formats determined by the program committee. At least one author of each accepted paper must register for the conference and present the work.
Call · 07
Key Policies
- Submissions must present original work that has not appeared in, been accepted to, or be under review at another archival venue.
- Simultaneous submissions of substantially similar work to other archival conferences or journals are not allowed during the DAI 2026 review period.
- Submissions must follow the Research Track double-blind review-anonymity policy.
- Authors must disclose relevant conflicts of interest in the submission system.
- Authors are responsible for the correctness, originality, and integrity of all submitted content.
- The use of generative AI tools must be disclosed when they materially contribute to writing, research design, code generation, data analysis, experiments, or other substantive parts of the work.
- Prompt injection, hidden instructions to reviewers or automated review tools, reviewer manipulation, plagiarism, fabricated results, and collusion are prohibited.
- Submissions should discuss ethical, legal, and societal considerations whenever they are relevant to the work, including implications for fairness, accountability, transparency, misuse, governance, labor, scientific practice, and the broader impacts of distributed or agentic AI systems.
- Submissions involving human subjects, sensitive or proprietary data, privacy or security risks, autonomous decision-making, multi-agent collusion or deception, scientific automation, safety-critical deployment, or high-impact domains must include an appropriate discussion of risks, safeguards, consent, approval processes, and legal or institutional compliance.
Call · 08
Contact
Program Chair
Ying Wen, [email protected]