Increased LLM Vulnerabilities from Fine-tuning and Quantization: Experiment Set-up & Results | HackerNoon
Briefly

from Hackernoon3 months ago
The testing on different downstream tasks, including fine-tuning and quantization, shows that while fine-tuning can improve task effectiveness, it can simultaneously increase jailbreaking vulnerabilities in LLMs.
Hackernoonhttps://hackernoon.com/increased-llm-vulnerabilities-from-fine-tuning-and-quantization-experiment-set-up-and-results
Our experiments reveal that foundational models, when fine-tuned, tend to lose their safety alignment, making them more susceptible to jailbreaking, a crucial finding for model security.
Hackernoonhttps://hackernoon.com/increased-llm-vulnerabilities-from-fine-tuning-and-quantization-experiment-set-up-and-results
Read at Hackernoon
#llms #fine-tuning #vulnerabilities #quantization #model-safety
[
|
]