Search

Results for "localization"

AceGPT, Localizing Large Language Models in Arabic

arXiv · Sep 21

Researchers introduce AceGPT, a localized large language model (LLM) specifically for Arabic, addressing cultural sensitivity and local values not well-represented in mainstream models. AceGPT incorporates further pre-training with Arabic texts, supervised fine-tuning using native Arabic instructions and GPT-4 responses, and reinforcement learning with AI feedback using a reward model attuned to local culture. Evaluations demonstrate that AceGPT achieves state-of-the-art performance among open Arabic LLMs across several benchmarks. Why it matters: This work advances culturally-aware AI development for Arabic-speaking communities, providing a valuable resource and benchmark for future research.

Language and Planning in Robotic Navigation: A Multilingual Evaluation of State-of-the-Art Models

arXiv · Jan 7

This paper introduces Arabic language integration into Vision-and-Language Navigation (VLN) in robotics, evaluating multilingual SLMs like GPT-4o mini, Llama 3 8B, Phi-3 14B, and Jais using the NavGPT framework. The study uses the R2R dataset to assess the impact of language on navigation reasoning through zero-shot sequential action prediction. Results show the framework enables high-level planning in both English and Arabic, though some models face challenges with Arabic due to reasoning limitations and parsing issues. Why it matters: This work highlights the need to improve language model planning and reasoning for effective navigation, especially to unlock the potential of Arabic-language models in real-world applications.