RADAR: Defending RAG Dynamically against Retrieval Corruption

TL;DR

RADAR is a novel framework that enhances the robustness of RAG systems against retrieval corruption by dynamically modeling context selection as a graph-based energy minimization problem, balancing stability and adaptability.

Contribution

It introduces a graph-based energy minimization approach with Bayesian memory for dynamic, robust context selection in RAG systems, reducing storage costs and improving resilience.

Findings

RADAR outperforms baselines in robustness and response quality.

It maintains minimal storage overhead.

Experiments on a new dynamic dataset validate effectiveness.

Abstract

While RAG systems are increasingly deployed in dynamic web search, temporal volatility amplifies their vulnerability to adversarial attacks. Existing static-oriented defenses struggle to handle evolving threats and incur prohibitive storage costs in dynamic settings. We propose RADAR, a framework that models reliable context selection as a graph-based energy minimization problem, solved exactly via Max-Flow Min-Cut. By incorporating a Bayesian memory node, RADAR recursively updates a belief state instead of archiving raw historical documents, effectively balancing stability against attacks with adaptability to genuine knowledge shifts. Experiments on a novel dynamic dataset show that RADAR achieves superior robustness and response quality with minimal storage overhead compared to the baselines.