Advancing ESG Intelligence: An Expert-level Agent and Comprehensive Benchmark for Sustainable Finance

2026-01-13

AI Summary

This work addresses the challenges posed by fragmented unstructured data and the need for rigorous, multi-step auditing in ESG (Environmental, Social, and Governance) analysis, which current large language models struggle to handle effectively. To this end, we propose ESGAgent—the first expert-level, hierarchical multi-agent framework tailored for sustainable finance—that integrates retrieval-augmented generation, real-time web search, and domain-specific tools to support end-to-end analysis, ranging from atomic-level question answering to comprehensive report generation. We also introduce the first three-tier ESG capability evaluation benchmark, spanning foundational comprehension to professional reporting. Experimental results demonstrate that ESGAgent achieves an average accuracy of 84.15% on atomic tasks, significantly outperforming leading closed-source models, and is capable of producing professional-grade reports complete with visualizations and verifiable citations.

Technology Category

• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Multiagent Systems: Modeling other Agents
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI

Abstract

Environmental, social, and governance (ESG) criteria are essential for evaluating corporate sustainability and ethical performance. However, professional ESG analysis is hindered by data fragmentation across unstructured sources, and existing large language models (LLMs) often struggle with the complex, multi-step workflows required for rigorous auditing. To address these limitations, we introduce ESGAgent, a hierarchical multi-agent system empowered by a specialized toolset, including retrieval augmentation, web search and domain-specific functions, to generate in-depth ESG analysis. Complementing this agentic system, we present a comprehensive three-level benchmark derived from 310 corporate sustainability reports, designed to evaluate capabilities ranging from atomic common-sense questions to the generation of integrated, in-depth analysis. Empirical evaluations demonstrate that ESGAgent outperforms state-of-the-art closed-source LLMs with an average accuracy of 84.15% on atomic question-answering tasks, and excels in professional report generation by integrating rich charts and verifiable references. These findings confirm the diagnostic value of our benchmark, establishing it as a vital testbed for assessing general and advanced agentic capabilities in high-stakes vertical domains.

Problem

• ESG analysis
• data fragmentation
• large language models
• multi-step workflows
• sustainable finance

Innovation

• ESGAgent
• multi-agent system
• retrieval-augmented generation
• sustainable finance benchmark
• domain-specific LLM evaluation