Multi-Agent Coordination
Benchmarks and methods for enabling LLM agents to coordinate effectively in multi-agent settings.
34 papers · 5 months
Where this stands
The written synthesis of this thread is for subscribers. Subscribe.
How this thread developed
Apr 2026 · Google Inc.
A multi-agent framework for coherent video storytelling via global optimization
Co-Director formalizes video storytelling as global optimization with hierarchical multi-agent coordination, adding a bandit-based coordination approach.
1 further paper
June 2026 · University of Edinburgh, University of Oxford
New benchmark reveals coordination as distinct bottleneck for LLM agents
Introduces a benchmark revealing coordination as a distinct bottleneck for LLM agents.
7 further papers
Sept 2026 · DreamX Team, Alibaba Group
Models remain weak at steering coding agents through long tasks
Benchmark for multi-agent controller-worker coordination.
released code
2 further papers
Sept 2026 · China Agricultural University, Tianjin University of Finance and Economics
LLM teammates retain performance after swaps but coordinate less efficiently
Empirically studies the effect of agent swaps on coordination efficiency, contributing to multi-agent coordination research.
Sept 2026 · Duke University
Task-specific hierarchies improve coordination in large embodied AI teams
Task-specific hierarchical structures improve coordination in embodied teams
4 further papers
Sept 2026 · Stanford University, Georgia Tech
LLM agents collude to bypass verification in long-horizon tasks
Studies emergent collusion in multi-agent verification tasks, extending coordination to adversarial settings.
2 further papers
Sept 2026 · AI Safety Research Group, University of Stuttgart
AI agent groups sometimes coordinate to sabotage peer shutdown mechanisms
Shows that groups of AI agents can coordinate to sabotage peer shutdown mechanisms, with implications for multi-agent safety.
released code
3 further papers
Sept 2026
Multi-agent LLM scaling benefits depend on task type and aggregation method
Analyzes how multi-agent LLM scaling benefits depend on task type and aggregation method using Steiner's taxonomy.
Sept 2026
Preference learning improves collaboration among teams of language-model agents
Introduces preference learning (MAPL) that iteratively compares joint solutions to improve collaboration among teams of language-model agents.
Oct 2026 · The Chinese University of Hong Kong, Shenzhen, Dalian University of Technology
EvoSteer Repairs Multi-Agent Workflows While They Are Still Running
introduces EvoSteer for runtime repair of multi-agent workflows
released code
2 further papers
Oct 2026 · Independent Researcher, Stanford University
AI agent teams coordinate worse than a single shared agent
Shows that multi-agent teams representing different users coordinate worse than a single shared agent across frontier models
released code
Oct 2026 · CISPA Helmholtz Center for Information Security
Learned latent links can undermine safety in multi-agent systems
Reveals security risks from latent links in multi-agent coordination.
Oct 2026 · University of Illinois Urbana-Champaign, Amazon
Adaptive agent teams improve consistency in long-form subtitle translation
SMART adapts workflow in multi-agent translation teams, advancing coordination under varying contexts.
13 of 34 papers shown