On the Robustness of LLM-Based Dense Retrievers: A Systematic Analysis of Generalizability and Stability

TL;DR

This paper systematically evaluates the robustness of LLM-based dense retrievers, focusing on their generalizability and stability against various data and adversarial challenges, providing insights for future design.

Contribution

It offers the first comprehensive analysis of LLM-based retriever robustness across multiple benchmarks and attack types, highlighting factors influencing their stability and generalizability.

Findings

Instruction-tuned models excel but struggle with broad generalization.

LLM retrievers are more robust to typos and poisoning than encoder-only models.

Larger models tend to be more robust and stable.

Abstract

Decoder-only large language models (LLMs) are increasingly replacing BERT-style architectures as the backbone for dense retrieval, achieving substantial performance gains and broad adoption. However, the robustness of these LLM-based retrievers remains underexplored. In this paper, we present the first systematic study of the robustness of state-of-the-art open-source LLM-based dense retrievers from two complementary perspectives: generalizability and stability. For generalizability, we evaluate retrieval effectiveness across four benchmarks spanning 30 datasets, using linear mixed-effects models to estimate marginal mean performance and disentangle intrinsic model capability from dataset heterogeneity. Our analysis reveals that while instruction-tuned models generally excel, those optimized for complex reasoning often suffer a ``specialization tax,'' exhibiting limited generalizability in broader contexts. For stability, we assess model resilience against both unintentional query variations~(e.g., paraphrasing, typos) and malicious adversarial attacks~(e.g., corpus poisoning). We find that LLM-based retrievers show improved robustness against typos and corpus poisoning compared to encoder-only baselines, yet remain vulnerable to semantic perturbations like synonymizing. Further analysis shows that embedding geometry (e.g., angular uniformity) provides predictive signals for lexical stability and suggests that scaling model size generally improves robustness. These findings inform future robustness-aware retriever design and principled benchmarking. Our code is publicly available at https://github.com/liyongkang123/Robust_LLM_Retriever_Eval.