Enhancing Language Models for Robust Greenwashing Detection

2026-01-29

AI Summary

This study addresses the limited robustness of current natural language processing models in detecting "greenwashing" in corporate sustainability reports, particularly their difficulty in distinguishing substantive actions from vague or unsubstantiated claims. To this end, the authors propose a parameter-efficient framework that structures the latent representation space of large language models by integrating contrastive learning with an ordinal ranking objective. The approach further incorporates a gated feature modulation mechanism to filter disclosure noise and employs MetaGradNorm to stabilize multi-objective optimization during training. Experimental results demonstrate that the method significantly outperforms baseline models under cross-category settings, enhancing robustness against semantic noise while uncovering a critical trade-off between representational rigidity and generalization capability.

Technology Category

• Natural Language Processing: Language Grounding & Multi-modal NLP
• Machine Learning: Efficient ML / Green AI
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

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.

Problem

• greenwashing
• sustainability reports
• robustness
• NLP
• ESG

Innovation

• contrastive learning
• ordinal ranking
• gated feature modulation
• MetaGradNorm
• greenwashing detection