Get Started
Home
Topics
Search
Library
Research questionWhen teams develop private coordination conventions, can role-matched agents be replaced without increasing communication or hurting task success?Agents may learn coordination conventions with particular teammates over repeated interactions. Replacing one agent can therefore change communication costs even when the replacement fills the same role and task performance remains similar.
AI
AI Agents
Evaluation & Benchmarks
Multi-agent Systems
Latest papersRecent research connected to this question, newest first.Testing Interchangeability in LLM Agent TeamsThe evidence concerns independently formed teams of agents from one base model, with private notebooks maintained across ten formation episodes. Role-matched agents were swapped on held-out tasks in Hanabi and Collab-Overcooked, with comparisons across base models, decoding temperatures, and formation lengths. The findings support conclusions about these tested settings rather than interchangeability across all multi-agent systems.research paper · Sep 4, 2026
Related questions
How can LLM orchestrators preserve continuous state when collaborating with non-language agents?How can multimodal agents collaborate asynchronously under time pressure with split information and imperfect communication?How does LLM-agent proportion affect consensus and the legitimacy of resulting norms in mixed human–AI groups?How can agent harnesses adapt across tasks and models without manual redesign?