The field of large language models (LLMs) is moving towards more flexible and scalable knowledge aggregation, with a focus on adaptive selection and fusion of multiple LLMs to build stronger models. This direction is driven by the need to overcome the limitations of traditional fine-tuning and ensemble methods, which require substantial memory and struggle to adapt to changing data environments. Recent innovations include the development of edge-centric multimodal frameworks that integrate LLMs with edge computing for real-time decision-making, as well as comprehensive system-level analyses of AI agents that highlight the need for compute-efficient reasoning. Additionally, there is a growing trend towards democratizing agricultural intelligence through the use of domain-specific foundation models, neural knowledge graphs, and multi-agent reasoning. Noteworthy papers include:

• Enabling Flexible Multi-LLM Integration for Scalable Knowledge Aggregation, which proposes a framework for adaptively selecting and aggregating knowledge from diverse LLMs.
• Farm-LightSeek, which presents an edge-centric multimodal agricultural IoT data analytics framework that integrates LLMs with edge computing.
• OpenAg, which introduces a comprehensive framework for advancing agricultural artificial general intelligence through the combination of domain-specific foundation models, neural knowledge graphs, and multi-agent reasoning.