Retrieval-Augmented Large Language Models for Climate Change and Renewable Energy Knowledge Synthesis
気候変動および再生可能エネルギー知識統合のための検索拡張型大規模言語モデル (AI 翻訳)
Orabi A, Harb H, Khedr A
🤖 gxceed AI 要約
日本語
本論文は、気候変動と再生可能エネルギーに関する知識統合のための検索拡張生成（RAG）フレームワークを提案する。FAISSとHNSWを用いた意味検索とLLMを組み合わせ、IPCCやIEAなどの信頼できる文書から情報を取得する。実験では87%の精度と41%の幻覚削減を達成し、静的モデルより優れる。
English
This paper proposes a Retrieval-Augmented Generation (RAG) framework for climate change and renewable energy knowledge synthesis. It combines dense retrieval (FAISS, HNSW) with LLMs to ground responses in authoritative sources like IPCC and IEA reports. Experiments show 87% accuracy and 41% reduction in hallucination compared to parametric-only baselines.
Unofficial AI-generated summary based on the public title and abstract. Not an official translation.
📝 gxceed 編集解説 — Why this matters
日本のGX文脈において
日本でも、気候変動や再エネに関する膨大な情報を効率的に統合するニーズが高まっている。本フレームワークは、日本企業や研究機関が国内外の政策文書や科学報告を横断的に分析する基盤となり得る。
In the global GX context
As climate and energy knowledge grows exponentially, this RAG framework offers a scalable method for synthesizing information from diverse authoritative sources. It addresses the critical challenge of factual grounding in LLMs, relevant for global climate reporting and policy analysis.
👥 読者別の含意
🔬研究者:Demonstrates a domain-specific RAG architecture that reduces hallucination and improves factual accuracy for climate knowledge tasks.
🏢実務担当者:Could be used by sustainability teams to quickly retrieve and summarize relevant climate reports and data.
🏛政策担当者:Enables evidence-based policymaking by providing grounded summaries of complex climate and energy documents.
📄 Abstract（原文）
<title>Abstract</title> <p>Climate change mitigation and renewable energy transitions require intelligent systems capable of reasoning over large, continuously evolving bodies of scientific, technical, and policy knowledge. While transformer-based Large Language Models (LLMs) demonstrate strong generative capabilities, their reliance on static parametric knowledge limits factual grounding, transparency, and adaptability to newly emerging climate evidence. This paper proposes a domain-specific Retrieval-Augmented Generation (RAG) framework for climate change and renewable energy intelligence. The system integrates a transformer-based decoder architecture with dense semantic retrieval using FAISS and HNSW indexing. Authoritative climate documents, including reports from the Intergovernmental Panel on Climate Change, International Energy Agency, United Nations Framework Convention on Climate Change, and International Renewable Energy Agency indexed using transformer-based embeddings . We evaluate both retrieval and generation components using Recall@K, MRR, nDCG, ROUGE-L, and hallucination rate metrics. Experimental results show 87% accuracy and a 41% reduction in hallucination compared to parametric-only baselines.</p>
🔗 Provenance — このレコードを発見したソース
- Research Square https://doi.org/10.21203/rs.3.rs-10042165/v1first seen 2026-06-20 04:52:22 · last seen 2026-06-30 04:55:17
🔔 こうした論文の新着を逃したくない方は キーワードアラート に登録（無料・3キーワードまで）。
gxceed は公開メタデータに基づく研究支援データセットです。要約・翻訳・解説は AI 支援で生成されています。 最終的な解釈・検証は利用者が原典資料に基づいて行うことを前提とします。