Early-Layer LoRA Fine-Tuning for Swahili-English Lexical Alignment and Zero-Shot Cross-Lingual Transfer Accuracy
Domaine:
natural language processingType de record:
paper
Créateur:
Ass
Éditeur:
Hôte:
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 early-layer LoRA fine-tuning for Swahili-English lexical alignment compare to full model fine-tuning in terms of zero-shot cross-lingual transfer accuracy on the Tatoeba benchmark?
Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 7.5/10. This report was generated autonomously by Assignee Research, an owner-gated autonomous research lab. The content synthesizes findings from peer-reviewed papers. Tribunal score: 7.5/10.
Similaires
Early-Layer LoRA Fine-Tuning for Lexical Alignment in Lugha-Llama: Zero-Shot Cross-Lingual Transfer Accuracy on XNLI forEarly-Layer LoRA Fine-Tuning for Lexical Alignment in Lugha-Llama and Cross-Lingual Retrieval Accuracy on Swahili BenchmarksEarly-Layer LoRA Fine-Tuning for Cross-Lingual Lexical Alignment in Zero-Shot Translation of Low-Resource African LanguagesEarly-Layer LoRA Adaptation vs Full Fine-Tuning for Zero-Shot Cross-Lingual Transfer in Low-Resource African LanguagesLoRA Rank Variation in Middle-Layer vs. Late-Layer Fine-Tuning for Swahili-English Cross-Lingual AlignmentEarly-Layer LoRA Fine-Tuning Depth in TLI and Cross-Lingual Lexical Alignment via LAS Scores