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

Optimizing large language models (LLMs) for downstream use cases ofteninvolves the customization of pre-trained LLMs through further fine-tuning.Meta's open release of Llama models and OpenAI's APIs for fine-tuning GPT-3.5Turbo on custom datasets also encourage this practice. But, what are the safetycosts associated with such custom fine-tuning? We note that while existingsafety alignment infrastructures can restrict harmful behaviors of LLMs atinference time, they do not cover safety risks when fine-tuning privileges areextended to end-users. Our red teaming studies find that the safety alignmentof LLMs can be compromised by fine-tuning with only a few adversariallydesigned training examples. For instance, we jailbreak GPT-3.5 Turbo's safetyguardrails by fine-tuning it on only 10 such examples at a cost of less than$0.20 via OpenAI's APIs, making the model responsive to nearly any harmfulinstructions. Disconcertingly, our research also reveals that, even withoutmalicious intent, simply fine-tuning with benign and commonly used datasets canalso inadvertently degrade the safety alignment of LLMs, though to a lesserextent. These findings suggest that fine-tuning aligned LLMs introduces newsafety risks that current safety infrastructures fall short of addressing --even if a model's initial safety alignment is impeccable, it is not necessarilyto be maintained after custom fine-tuning. We outline and critically analyzepotential mitigations and advocate for further research efforts towardreinforcing safety protocols for the custom fine-tuning of aligned LLMs.

Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To!