Evaluating the Ripple Effects of Knowledge Editing in Language Models

by   Roi Cohen, et al.
Tel Aviv University

Modern language models capture a large body of factual knowledge. However, some facts can be incorrectly induced or become obsolete over time, resulting in factually incorrect generations. This has led to the development of various editing methods that allow updating facts encoded by the model. Evaluation of these methods has primarily focused on testing whether an individual fact has been successfully injected, and if similar predictions for other subjects have not changed. Here we argue that such evaluation is limited, since injecting one fact (e.g. “Jack Depp is the son of Johnny Depp”) introduces a “ripple effect” in the form of additional facts that the model needs to update (e.g.“Jack Depp is the sibling of Lily-Rose Depp”). To address this issue, we propose a novel set of evaluation criteria that consider the implications of an edit on related facts. Using these criteria, we then construct , a diagnostic benchmark of 5K factual edits, capturing a variety of types of ripple effects. We evaluate prominent editing methods on , showing that current methods fail to introduce consistent changes in the model's knowledge. In addition, we find that a simple in-context editing baseline obtains the best scores on our benchmark, suggesting a promising research direction for model editing.


page 1

page 4


Eva-KELLM: A New Benchmark for Evaluating Knowledge Editing of LLMs

Large language models (LLMs) possess a wealth of knowledge encoded in th...

Measuring and Manipulating Knowledge Representations in Language Models

Neural language models (LMs) represent facts about the world described b...

Detecting Edit Failures In Large Language Models: An Improved Specificity Benchmark

Recent model editing techniques promise to mitigate the problem of memor...

Locating and Editing Factual Knowledge in GPT

We investigate the mechanisms underlying factual knowledge recall in aut...

Propagating Knowledge Updates to LMs Through Distillation

Modern language models have the capacity to store and use immense amount...

ReFACT: Updating Text-to-Image Models by Editing the Text Encoder

Text-to-image models are trained on extensive amounts of data, leading t...

Can We Edit Factual Knowledge by In-Context Learning?

Previous studies have shown that large language models (LLMs) like GPTs ...

Please sign up or login with your details

Forgot password? Click here to reset