Multi-Agent Framework for Low-Resource Dialect Translation

This approach reframes neural machine translation as a structured decision space explored by autonomous agents rather than a single output. By modeling distinct pathways—including zero-shot, fine-tuned, and pivot translation—the method treats inter-agent divergence as an interpretable signal for Turkish-to-Syrian Arabic translation.