A new arXiv paper explores how prompt design and demonstration selection impact the machine translation capabilities of local large language models (LLMs). The study evaluated models like Llama3.2 3B, mistral:latest, and Qwen2.5:14b on translating between English and nine European Union languages, comparing them against specialized MT systems such as OPUS-MT and NLLB-200. Findings indicate that while dedicated MT systems generally perform better, few-shot prompting can benefit some LLMs, and embedding-similarity retrieval for demonstrations offers a modest advantage. The research also highlights the feasibility and limitations of family-scope prompting for multi-language translation tasks with LLMs. AI
IMPACT This research provides insights into optimizing LLM performance for translation tasks by refining prompt engineering and data selection strategies.
RANK_REASON The cluster contains a single academic paper detailing research findings on LLM capabilities. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
