Zang Li

SPARD: Defending Harmful Fine-Tuning Attack via Safety Projection with Relevance–Diversity Data Selection

Improving Reasoning Ability of Large Language Models via Iterative Uncertainty-based Preference Optimization