#model-safety

[ follow ]
Bootstrapping
fromHackernoon
5 months ago

Comprehensive Detection of Untrained Tokens in Language Model Tokenizers | HackerNoon

Glitch tokens in LLMs lead to unwanted behaviors.
Effective methods are needed to identify problematic tokens.
An analysis of tokenizers reveals their role in model safety.
fromHackernoon
7 months ago

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

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.
Data science
[ Load more ]