Search papers, labs, and topics across Lattice.
This paper introduces CONCAT, a training-free framework for efficient LLM-based multi-agent collaboration that leverages consensus and confidence to reduce communication overhead. CONCAT clusters agents based on initial answers, selects leaders based on confidence, and uses a Theory of Mind-based heuristic to predict collaboration benefits between leaders. Experiments across multiple LLMs and benchmarks demonstrate that CONCAT achieves up to 2.02x higher efficiency than LLM-Debate and reduces average latency by 50.1% on Qwen2.5-14B-Instruct, without task-specific training.
Slash LLM multi-agent system latency by up to 50% without any training using a clever consensus- and confidence-driven ad hoc teaming strategy.
Although large language model (LLM) based multi-agent systems (MAS) show their capability to solve complex tasks and achieve higher performance over single agent systems, they lead to huge computational overheads because of heavy communication between agents. Previous research has made efforts to train a sparse multi-agent graph or fine-tune a planner to orchestrate the workflow better. However, such extra training processes introduce computational costs and limit MAS to specific domains, therefore compromising their generalizability. In this paper, we propose CONCAT, a training-free multi-agent collaboration framework based on CONsensus and Confidence-driven Ad hoc Teaming to efficiently organize agent interactions. Specifically, agents are clustered based on their initial answers, and leaders of each cluster are selected based on the agents'confidence. Then, a heuristic function based on the Theory of Mind is designed to predict the collaboration benefits between every two leaders according to their answers and confidence. Finally, an ad hoc multi-agent network is organized after evicting a percentage of communications based on the predicted benefits. Experiments across three LLMs and three benchmarks show that CONCAT achieves up to 2.02x higher efficiency (accuracy/latency ratio) than LLM-Debate and outperforms training-aware methods such as AgentDropout, while reducing average latency by 50.1% on Qwen2.5-14B-Instruct, without any task-specific training.
