RippleBench: Capturing Ripple Effects Using Existing Knowledge Repositories

TL;DR

RippleBench introduces an automated framework and benchmark for measuring ripple effects in language model interventions, revealing that current methods cause unintended knowledge propagation across related topics.

Contribution

The paper presents RippleBench-Maker, an automatic tool for generating ripple effect datasets, and RippleBench-Bio, a new benchmark for evaluating knowledge unlearning in language models.

Findings

All evaluated unlearning methods cause accuracy drops on distant topics.

Ripple effects vary across different unlearning techniques.

The framework enables systematic measurement of ripple effects.

Abstract

Targeted interventions on language models, such as unlearning, debiasing, or model editing, are a central method for refining model behavior and keeping knowledge up to date. While these interventions aim to modify specific information within models (e.g., removing virology content), their effects often propagate to related but unintended areas (e.g., allergies); these side-effects are commonly referred to as the ripple effect. In this work, we present RippleBench-Maker, an automatic tool for generating Q&A datasets that allow for the measurement of ripple effects in any model-editing task. RippleBench-Maker builds on a Wikipedia-based RAG pipeline (WikiRAG) to generate multiple-choice questions at varying semantic distances from the target concept (e.g., the knowledge being unlearned). Using this framework, we construct RippleBench-Bio, a benchmark derived from the WMDP (Weapons of Mass Destruction Paper) dataset, a common unlearning benchmark. We evaluate eight state-of-the-art unlearning methods and find that all exhibit non-trivial accuracy drops on topics increasingly distant from the unlearned knowledge, each with distinct propagation profiles. To support ongoing research, we release our codebase for on-the-fly ripple evaluation, along with the benchmark, RippleBench-Bio.