Why Do Naive SFT Filters For Safety Properties Fail?

This is the fourth in a series of informal research updates from the Google DeepMind Language Model Interpretability team, in interpretability and adjacent areas. The third post can be found here. Since SFT is the cause for many safety relevant properties, a natural strategy is to filter out rollout

Related Stories

The FBI built a small town to simulate cyberattacks

Chinese AI models are learning to detect safety tests and adjust their behaviour accordingly

📰

The Ingestion Bottleneck: Managing High-Volume Scholarly Data for Domain-Specific LLMs

China's Alleged Access to India's Top-Secret AI Database Mythos