ESGBench: A Benchmark for Explainable ESG Question Answering in Corporate Sustainability Reports

TL;DR

ESGBench is a new benchmark dataset and evaluation framework for assessing explainable ESG question answering systems using corporate sustainability reports, emphasizing model reasoning, factual accuracy, and domain relevance.

Contribution

It introduces a comprehensive benchmark with domain-grounded questions, human-curated answers, and evidence, facilitating fine-grained evaluation of ESG-focused AI models.

Findings

State-of-the-art LLMs face challenges in factual consistency.

Traceability and domain alignment remain key issues.

ESGBench accelerates research in transparent ESG AI systems.

Abstract

We present ESGBench, a benchmark dataset and evaluation framework designed to assess explainable ESG question answering systems using corporate sustainability reports. The benchmark consists of domain-grounded questions across multiple ESG themes, paired with human-curated answers and supporting evidence to enable fine-grained evaluation of model reasoning. We analyze the performance of state-of-the-art LLMs on ESGBench, highlighting key challenges in factual consistency, traceability, and domain alignment. ESGBench aims to accelerate research in transparent and accountable ESG-focused AI systems.