Enhancing Language Models for Robust Greenwashing Detection

TL;DR

This paper introduces a robust, parameter-efficient NLP framework that improves greenwashing detection in sustainability reports by combining contrastive learning, ordinal ranking, and feature modulation, enhancing generalization and robustness.

Contribution

It presents a novel structured LLM approach integrating contrastive learning and ranking objectives, with gating and stabilization techniques for better greenwashing detection.

Findings

Outperforms standard baselines in cross-category robustness

Reveals a trade-off between model rigidity and generalization

Demonstrates effectiveness in real-world ESG report analysis

Abstract

Sustainability reports are critical for ESG assessment, yet greenwashing and vague claims often undermine their reliability. Existing NLP models lack robustness to these practices, typically relying on surface-level patterns that generalize poorly. We propose a parameter-efficient framework that structures LLM latent spaces by combining contrastive learning with an ordinal ranking objective to capture graded distinctions between concrete actions and ambiguous claims. Our approach incorporates gated feature modulation to filter disclosure noise and utilizes MetaGradNorm to stabilize multi-objective optimization. Experiments in cross-category settings demonstrate superior robustness over standard baselines while revealing a trade-off between representational rigidity and generalization.