Reproduce “Enhancing Retrieval-Augmented Generation for Electric Power Industry Customer Support” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Hei Yu Chan, Kuok Tou Ho, Chenglong Ma, Yujing Si, Hok Lai Lin, Sa Lei Lam

Pachira (International) Technology Ltd. developed an optimized Retrieval-Augmented Generation (RAG) system for electric power industry customer support, achieving 97.9% answer accuracy on a diverse query dataset and 89.6% on a FAQ dataset. The system integrates a graph-based RAG framework with query rewriting, RAG Fusion, context reranking, and intent recognition to deliver precise, contextually relevant, and actionable responses.

Key insights

  • Graph-based RAG frameworks significantly outperform vector-store baselines in complex, technical domains by recognizing entities and relationships more effectively.
  • Combining query rewriting, RAG Fusion, context reranking, and intent recognition is crucial for robust RAG system performance in challenging, real-world applications.
  • Keyword augmentation can negatively impact retrieval accuracy if keywords are not well-aligned with the query, highlighting the need for careful design of augmentation strategies.
  • Context reranking, particularly reversing the order of top-ranked documents, is effective in mitigating the 'Lost In the Middle' problem, where LLMs overlook crucial information within long contexts.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.