Brick-DICL: Dynamic In-Context Learning for Automated Brick Schema Classification

2026-06-16 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Internet of Things (IoT) & Connected Devices · Depth: Advanced, quick

Summary

Brick-DICL is a novel two-stage dynamic in-context learning framework designed for automated Brick schema classification, addressing the lack of standardization in Building Management Systems (BMS) points. It tackles challenges such as the extensive 936 Brick classes, limited domain-specific knowledge in large language models (LLMs), and substantial manual verification effort. The framework comprises metadata-RAG, which retrieves relevant examples to enhance LLM domain knowledge, and class-RAG, which narrows down potential Brick classes. Additionally, Brick-DICL incorporates a multi-LLM filtering mechanism to compare predictions across multiple models, flagging low-confidence classifications for human review. This approach is applicable to any BMS regardless of manufacturer or metadata format, achieves significant classification accuracy improvements over existing methods, and efficiently reduces manual verification, accelerating digital building onboarding.

Key takeaway

For Machine Learning Engineers developing automated classification systems for extensive and complex taxonomies, such as building ontologies, Brick-DICL demonstrates a robust framework. You should consider integrating dynamic in-context learning with RAG-based context retrieval and multi-LLM filtering. This approach can significantly improve classification accuracy and reduce the manual verification effort typically associated with onboarding diverse datasets, accelerating the path to standardized, interoperable systems.

Key insights

Brick-DICL automates Brick schema classification using dynamic in-context learning, RAG for domain knowledge, and multi-LLM filtering for accuracy.

Principles

• Standardized ontologies enhance system interoperability.
• Context retrieval boosts LLM domain knowledge.
• Multi-LLM filtering improves classification confidence.

Method

Brick-DICL employs a two-stage dynamic in-context learning framework: metadata-RAG retrieves examples, and class-RAG narrows Brick classes. A multi-LLM filtering mechanism compares predictions to flag low-confidence classifications.

In practice

• Integrate RAG for LLM domain knowledge.
• Employ multi-LLM consensus for critical classifications.
• Narrow classification space for large ontologies.

Topics

• Brick Schema Classification
• Building Management Systems
• In-Context Learning
• Large Language Models
• Retrieval-Augmented Generation
• Ontology Mapping

Best for: AI Scientist, Machine Learning Engineer, Research Scientist

Related on AIssential

• Retrieval-Augmented Generation as a Bridge Between Classical NLP Pipelines and Modern LLM Systems — Retrieval-augmented generation (RAG) addresses the limitation of static knowledge in large language models (LLMs) by connecting them to external, up-to-date knowledge sources.
• Towards Generalization of Block Attention via Automatic Segmentation and Block Distillation — Semantic segmentation and an efficient distillation framework enable practical, scalable block attention for long-context LLMs.
• How SHACL Makes Your LLMs Hum — Integrating SHACL-defined knowledge graphs with LLMs significantly reduces hallucinations and improves response accuracy.
• Multi-Agent Architecture with RAG and Dynamic Context Windows for Text-to-SQL Optimization — A multi-agent Text-to-SQL system uses RAG and dynamic context to reduce token consumption.
• LLMs on Tabular Data with Limited Semantics: Evidence from Industrial Car Retrofit Prediction — LLMs complement, but do not replace, strong tabular models on privacy-constrained industrial data lacking semantic cues.
• Scaling AI Knowledge Systems: Lessons from the DHTMLX MCP Server — Scaling RAG systems for multiple distinct products requires intelligent query routing and optimized knowledge structuring.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.