We describe our submission to the WMT26 Low-Resource Arabic–Asian Language Translation shared task, covering ten translation directions between Arabic and five Asian languages. Our central question is whether a post-hoc controller can improve translation quality beyond a fine-tuned backbone without modifying any model weights. For the five into-Arabic directions, our primary system jointly fine-tunes facebook/nllb-200-3.3B on all source languages and augments it at inference time with per-direction NTK-Mirror controllers, each containing fewer than 6K parameters, that rescale frozen transformer output channels. For the five from-Arabic directions, we use Qwen2.5-7B-Instruct as the backbone, adapted via QLoRA and likewise augmented with NTK-Mirror. In the official evaluation, our system ranked first on Bengali-to-Arabic and Indonesian-to-Arabic, second on Hindi-to-Arabic and Urdu-to-Arabic, and third on English-to-Arabic, placing in the top three across all five into-Arabic directions. Post-submission analysis indicates that the weaker results in the from-Arabic directions stem primarily from a model-selection decision made at submission time.