English

EmoRAG: Evaluating RAG Robustness to Symbolic Perturbations

Cryptography and Security 2025-12-02 v1 Artificial Intelligence Computation and Language

Abstract

Retrieval-Augmented Generation (RAG) systems are increasingly central to robust AI, enhancing large language model (LLM) faithfulness by incorporating external knowledge. However, our study unveils a critical, overlooked vulnerability: their profound susceptibility to subtle symbolic perturbations, particularly through near-imperceptible emoticon tokens such as "(@_@)" that can catastrophically mislead retrieval, termed EmoRAG. We demonstrate that injecting a single emoticon into a query makes it nearly 100% likely to retrieve semantically unrelated texts that contain a matching emoticon. Our extensive experiment across general question-answering and code domains, using a range of state-of-the-art retrievers and generators, reveals three key findings: (I) Single-Emoticon Disaster: Minimal emoticon injections cause maximal disruptions, with a single emoticon almost 100% dominating RAG output. (II) Positional Sensitivity: Placing an emoticon at the beginning of a query can cause severe perturbation, with F1-Scores exceeding 0.92 across all datasets. (III) Parameter-Scale Vulnerability: Counterintuitively, models with larger parameters exhibit greater vulnerability to the interference. We provide an in-depth analysis to uncover the underlying mechanisms of these phenomena. Furthermore, we raise a critical concern regarding the robustness assumption of current RAG systems, envisioning a threat scenario where an adversary exploits this vulnerability to manipulate the RAG system. We evaluate standard defenses and find them insufficient against EmoRAG. To address this, we propose targeted defenses, analyzing their strengths and limitations in mitigating emoticon-based perturbations. Finally, we outline future directions for building robust RAG systems.

Keywords

Cite

@article{arxiv.2512.01335,
  title  = {EmoRAG: Evaluating RAG Robustness to Symbolic Perturbations},
  author = {Xinyun Zhou and Xinfeng Li and Yinan Peng and Ming Xu and Xuanwang Zhang and Miao Yu and Yidong Wang and Xiaojun Jia and Kun Wang and Qingsong Wen and XiaoFeng Wang and Wei Dong},
  journal= {arXiv preprint arXiv:2512.01335},
  year   = {2025}
}

Comments

Accepted to ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) 2026

R2 v1 2026-07-01T08:03:08.379Z