Efficiency at What Cost? Safety and Fairness in Parameter-Efficient Fine-Tuning of LLMs

Summary

This study tests whether benign parameter-efficient fine-tuning of LLMs affects safety and fairness. Across 235 conversational variants of four instruction-tuned families (Meta-Llama-3-8B, Qwen2.5-7B, Mistral-7B, Gemma-7B) using LoRA, IA3, Prompt-Tuning and P-Tuning, plus 96 coding-task models, benign PEFT produced alignment shifts. Adapter-based methods were generally safer, and Gemma showed the steepest safety decline.

Mitigation

The source directs practitioners to start from a well-aligned base model, favor adapter-based PEFT, and audit safety and fairness at the category level.

Related items

• InfoUber only fulfilled 4 rides from AI agents last quarterSame vendor · The Verge (AI)
• InfoCrowdStrike to acquire Fiddler, Dreadnode to extend AI security across agent lifecycleSame vendor · CSO Online
• MediumGoogle: LLMjacking attacks rack up $1 million in cloud AI chargesSame vendor · BleepingComputer
• InfoMeta seeks to raise up to $25 billion in debt sale as AI spending jumpsSame vendor · CNBC Technology
• InfoOpenAI says ChatGPT is catching up to Google in searchSame vendor · The Verge (AI)