TFlow framework enables LLM agents to communicate via weight updates

Researchers have developed TFlow, a novel framework for multi-agent LLM collaboration that utilizes weight perturbations instead of traditional text-based messaging. This approach compiles sender agents' internal states into transient, low-rank adaptations for the receiver model, reducing computational overhead and memory usage. Experiments with Qwen3-4B agents demonstrated TFlow's ability to improve accuracy and significantly decrease processed tokens and inference time compared to both standalone models and text-based communication methods. AI

影响 Introduces a more efficient communication method for multi-agent LLM systems, potentially reducing costs and improving performance.

排序理由 The cluster contains an academic paper detailing a new method for LLM agent communication. [lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.CL 阅读 →

• Qwen3-4B
• TFlow

• 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 →