Depth of LoRA Adapter Injection in Lugha-Llama for Cross-Lingual Alignment in Swahili-English Translation

Large Language Models (LLMs) have demonstrated remarkable capabilities, yet their performance in low-resource languages (LRLs), such as Swahili, often lags due to data scarcity and underrepresentation in pre-training. A key challenge is achieving robust cross-lingual lexical alignment, crucial for tasks like translation and cross-lingual information retrieval. This paper introduces Targeted Lexical Injection (TLI), a novel and efficient fine-tuning approach. We first demonstrate that Lugha-Llama-8B-wura, a Swahili-centric LLM, exhibits strong, near-perfect lexical alignment for Swahili-English

Research goal: How does varying the depth of LoRA adapter injection in Lugha-Llama affect cross-lingual alignment accuracy on low-resource Swahili-English translation pairs compared to early-layer-only strategies?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 7.5/10.