Pica Code IR-2026-078 Explainable Entity Resolution with LLMs

Development of explainable artificial intelligence methods for Large LanguageModel-based Entity Resolution systems. The activity will define attribution modelsto assess the contribution of individual records, record interactions, and semanticinformation to clustering decisions. Techniques based on surrogate representations,perturbation analysis, and Shapley values will be implemented and validated. Theexperimental evaluation will assess explanation fidelity, interpretability, efficiency,and scalability.