IREA: Intermediate Representation-based Embedding Alignment for Normative RAG

2026-10-04

AI Summary

This study addresses the query-text asymmetry challenge in norm retrieval for ethical judgment in large language models by proposing an Intermediate Representation-based Embedding Alignment (IREA) method. IREA pioneers the mapping of both queries and normative texts into a shared situational-behavioral representation space, thereby overcoming the limitations of conventional unilateral expansion and effectively optimizing Retrieval-Augmented Generation (RAG) frameworks for norm retrieval. Experimental results demonstrate that IREA significantly improves norm retrieval accuracy and enhances downstream ethical judgment performance, validating the effectiveness of the proposed bidirectional alignment strategy.

Abstract

Large language models (LLMs) have shown strong performance across various tasks, but they still struggle with questions involving ethical judgment. Previous studies have attempted to train LLMs on ethical standards, but the diversity and relativity of ethical norms make them difficult to fully internalize in model parameters. As an alternative, we introduce normative RAG, a retrieval-augmented approach that supports ethical judgment using external normative knowledge. Normative retrieval involves a distinct asymmetry between context rich narrative queries and generalized normative statements. Existing factual retrieval methods rely on query-only expansion into a document-like form, making them insufficient for resolving this asymmetry. Therefore, we propose Intermediate Representation-based Embedding Alignment (IREA), a bidirectional alignment method that maps both text types into a shared situation-behavior representation. This representation captures ethically salient contextual and behavioral information in a normalized form, reducing surface-level discrepancies and improving alignment in the embedding space. Experimental results show that IREA improves normative retrieval and downstream ethical judgment across multiple settings, demonstrating the effectiveness of bidirectional alignment for normative RAG.

Problem

• Ethical judgment
• Normative RAG
• Retrieval asymmetry
• Large language models
• Embedding alignment

Innovation

• Normative RAG
• Embedding Alignment
• Intermediate Representation
• Bidirectional Alignment
• Ethical Judgment