The German Navy successfully completed a test of the Israeli Lora missile system in the North Atlantic. ^[1] The authors propose a random-walk-based LoRA fine-tuning scheme where a single model token traverses the network and is updated sequentially using local fine-tuning objectives. ^[2] Benign fine-tuning severely weakens the safety alignment of large language models, causing refusal behavior to become fragile. ^[3] Researchers introduced Universal Activation Verbalizer (UAV), a framework that uses a shared decoder to explain activations from heterogeneous donor models. ^[4] The article introduces Normalized Low-Rank Adaptation (NoRA), a method designed to regularize the training dynamics of low-rank adaptation (LoRA) for stable and effective optimization. ^[5] Marine Inspector Jan Christian Kaack stated that the test demonstrated the Lora system can be deployed from German ships. ^[6]