Interpretation Meets Safety: A Survey on Interpretation Methods and Tools for Improving LLM Safety

2025-06-05

AI Summary

Unsafe behaviors of large language models (LLMs) in real-world deployment pose critical risks, yet existing interpretability methods lack a unified, security-oriented analytical framework spanning the full model workflow—from input to reasoning to output. Method: We conduct a systematic review of over 70 works and propose the first security-aware interpretability taxonomy grounded in LLM workflow stages. Integrating literature analysis, cross-method comparison, security-interpretability joint modeling, and toolchain mapping, we construct a structured knowledge graph that uncovers causal pathways through which explanation techniques enhance safety. Contributions/Results: (1) We establish a theoretical paradigm for co-optimizing interpretability and safety; (2) we identify key technical gaps hindering practical adoption; and (3) we deliver an actionable, dual-audience guideline—tailored for both researchers and practitioners—to bridge theory and implementation in secure LLM development.

Technology Category

• Natural Language Processing: Safety and Robustness
• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Interpretability, Explainability, and Transparency

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Security and Privacy: Large-scale security measurements

Abstract

As large language models (LLMs) see wider real-world use, understanding and mitigating their unsafe behaviors is critical. Interpretation techniques can reveal causes of unsafe outputs and guide safety, but such connections with safety are often overlooked in prior surveys. We present the first survey that bridges this gap, introducing a unified framework that connects safety-focused interpretation methods, the safety enhancements they inform, and the tools that operationalize them. Our novel taxonomy, organized by LLM workflow stages, summarizes nearly 70 works at their intersections. We conclude with open challenges and future directions. This timely survey helps researchers and practitioners navigate key advancements for safer, more interpretable LLMs.

Problem

• Understanding and mitigating unsafe behaviors in LLMs
• Bridging interpretation techniques with LLM safety enhancements
• Providing a unified framework for safety-focused interpretation methods

Innovation

• Unified framework connects interpretation and safety
• Novel taxonomy organizes 70 works by stages
• Survey guides safer interpretable LLM advancements