Language agents use auction to cut communication costs and boost reasoning

Researchers have developed a new framework called DALA (Dynamic Auction-based Language Agent) to improve communication efficiency in multi-agent systems powered by large language models. This system treats communication as a scarce resource, using an auction-based mechanism where agents bid to send messages based on their predicted value. Experiments show DALA achieves state-of-the-art results on benchmarks like MMLU and HumanEval while significantly reducing token usage compared to existing methods. The framework also encourages agents to strategically remain silent when communication is not valuable. AI

影响 Introduces a novel, cost-effective communication protocol for LLM agents that could reduce operational expenses and improve performance.

排序理由 Academic paper introducing a novel framework for LLM-based multi-agent systems.

在 arXiv cs.AI 阅读 →

• arXiv
• GSM8K
• HumanEval
• LLM
• MMLU

• 论文
• 模型发布

AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →