Team Alkhder's submission to the AlexandriaX-2026 shared task addresses Subtask 1: translating English dialogue turns into thirteen Arabic dialects. The system is built by fine-tuning NileChat-3B-Base, a 3-billion-parameter model, with QLoRA.
The paper's title describes the approach as context-aware, indicating that the model is adapted for dialogue rather than isolated sentences. The abstract does not report specific evaluation results, so the effectiveness of the method is not detailed in the available text.
The work is notable for attempting to cover thirteen Arabic dialects with a single 3B model. Whether that proves sufficient is left for the full paper to show, as the abstract stops short of reporting results.