Benchmark Leakage Trap: Can We Trust LLM-based Recommendation?

TL;DR

This paper uncovers and analyzes the impact of benchmark data leakage in LLM-based recommender systems, demonstrating how it can artificially inflate or degrade performance metrics depending on domain relevance, thus questioning the reliability of current evaluation methods.

Contribution

It identifies and experimentally validates the issue of benchmark data leakage in LLM-based recommendation, highlighting its effects on performance measurement and evaluation reliability.

Findings

Leaked domain-relevant data causes inflated performance metrics.

Leaked domain-irrelevant data can reduce recommendation accuracy.

Data leakage significantly affects the trustworthiness of LLM-based recommendation evaluations.

Abstract

The expanding integration of Large Language Models (LLMs) into recommender systems poses critical challenges to evaluation reliability. This paper identifies and investigates a previously overlooked issue: benchmark data leakage in LLM-based recommendation. This phenomenon occurs when LLMs are exposed to and potentially memorize benchmark datasets during pre-training or fine-tuning, leading to artificially inflated performance metrics that fail to reflect true model performance. To validate this phenomenon, we simulate diverse data leakage scenarios by conducting continued pre-training of foundation models on strategically blended corpora, which include user-item interactions from both in-domain and out-of-domain sources. Our experiments reveal a dual-effect of data leakage: when the leaked data is domain-relevant, it induces substantial but spurious performance gains, misleadingly exaggerating the model's capability. In contrast, domain-irrelevant leakage typically degrades recommendation accuracy, highlighting the complex and contingent nature of this contamination. Our findings reveal that data leakage acts as a critical, previously unaccounted-for factor in LLM-based recommendation, which could impact the true model performance. We release our code at https://github.com/yusba1/LLMRec-Data-Leakage.