Aligning Greener Electronics Recommendations with EU Policy Using Retrieval-Augmented Generation
Abstract
Mapping sustainability recommendations for greener electronics to European Union legislation requires navigating a broad regulatory corpus. This thesis investigates whether a retrieval-augmented generation pipeline can automate this alignment task by evaluating retrieval effectiveness, comparing retrieval configurations and assessing an instruction-tuned language model's classification of recommendation-evidence alignment. The retrieval stage is evaluated using BM25, dense retrieval with BGE-M3 and E5-Large-v2, SPLADE, Reciprocal Rank Fusion and cross-encoder reranking on a gold standard of 130 queries linked to 24 EU documents and the MTEB MuPLeR-retrieval benchmark. SPLADE with reranking achieved the highest gold-standard NDCG@10 (0.575), but BGE-M3 with dense retrieval and reranking was selected for its higher Hit@10 (0.746) and stronger MuPLeR chunk-level performance (0.955 vs. 0.865), improving gold-standard NDCG@10 by 0.160 over BM25. Applied to whitepaper recommendations, the classifier assigned 25.0% as Aligned, 68.8% as Conditional, and 6.2% as having No explicit legal basis. A domain expert rated 94.9% of evaluated outputs as generally or partially correct, with evidence Precision@3 of 0.917. These results demonstrate that the pipeline is viable as a first-pass screening tool, while highlighting the need for larger annotated benchmarks and stronger classifiers for legal reasoning.
Read the paper
Related studies
- Knowledge Management Perspective of Generative Artificial Intelligence
- Navigating Generative Artificial Intelligence Promises and Perils for Knowledge and Creative Work
- Characterizing the evolving research on enterprise content management
- Large language models and multimodal foundation models for precision oncology
- From Generative AI to Extended Reality: Multidisciplinary Perspectives on the Challenges, Opportunities, and Future of Educational Computing
- Generative Artificial Intelligence: A New Engine for Advancing Environmental Science and Engineering
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.