English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Recent advances in large language models (LLMs) have opened up new paradigmsfor accessing the knowledge stored in their parameters. One critical challengethat has emerged is the presence of hallucinations in LLM outputs due to falseor outdated knowledge. Since retraining LLMs with updated information isresource-intensive, there has been a growing interest in model editing.However, many model editing methods, while effective in various scenarios, tendto overemphasize aspects such as efficacy, generalization, and locality inediting performance, often overlooking potential side effects on the generalabilities of LLMs. In this paper, we raise concerns that the improvement ofmodel factuality may come at the cost of a significant degradation of thesegeneral abilities, which is not conducive to the sustainable development ofLLMs. Systematically, we analyze side effects by evaluating four popularediting methods on two LLMs across eight representative task categories.Extensive empirical research reveals that model editing does improve modelfactuality but at the expense of substantially impairing general abilities.Therefore, we advocate for more research efforts to minimize the loss ofgeneral abilities acquired during LLM pre-training and to ultimately preservethem during model editing.

Model Editing Can Hurt General Abilities of Large Language Models