The Unlearning Sensitive Content from Large Language Models task aims to remove targeted datapoints from trained models while minimally affecting their general knowledge. In our work, we leverage parameter-efficient, gradient-based unlearning using low-rank (LoRA) adaptation and layer-focused fine-tuning. To further enhance unlearning effectiveness, we employ data chunking, splitting forget data into disjoint partitions and merging them with cyclically sampled retain samples at a pre-defined ratio. Our task-agnostic method achieves an outstanding forget-retain balance, ranking first on leaderboards and significantly outperforming baselines and competing systems.

Fox News' William La Jeunesse reports on wait times as long as 26 years to enter the U.S. legally. The U.S. government has started requiring migrants without passports to submit to facial recognition technology to take domestic flights under a change that prompted confusion this week among immigrants and advocacy groups in Texas. It is not clear exactly when the change took effect, but several migrants with flights out of South Texas on Tuesday told advocacy groups that they thought they were being turned away. The migrants included people who had used the government's online appointment system to pursue their immigration cases. Advocates were also concerned about migrants who had crossed the U.S.-Mexico border illegally before being processed by Border Patrol agents and released to pursue their immigration cases.