The Technology Innovation Institute (TII) in Abu Dhabi has launched Falcon LLM, a 40 billion parameter large language model (LLM). Developed by the AI and Digital Science Research Center (AIDRC), Falcon LLM outperforms GPT-3 while using less training compute. The model was trained on one trillion tokens and will impact fields like education and healthcare. Why it matters: This release establishes the UAE as a major player in generative AI and supports the nation's AI strategy.
The paper introduces Ara-HOPE, a human-centric post-editing evaluation framework for Dialectal Arabic to Modern Standard Arabic (DA-MSA) translation. Ara-HOPE includes a five-category error taxonomy and a decision-tree annotation protocol designed to address the challenges of dialect-specific MT errors. Evaluation of Jais, GPT-3.5, and NLLB-200 shows dialect-specific terminology and semantic preservation remain key challenges. Why it matters: The new framework and public dataset will help improve the evaluation and development of dialect-aware MT systems for Arabic.
Researchers introduce two new benchmarks, derived from the Qiyas exam, to evaluate mathematical reasoning and language understanding in Arabic. They tested ChatGPT-3.5-turbo and ChatGPT-4, which achieved 49% and 64% accuracy respectively. The new benchmarks aim to address the lack of resources for evaluating Arabic language models.
MBZUAI researchers introduce PG-Video-LLaVA, a large multimodal model with pixel-level grounding capabilities for videos, integrating audio cues for enhanced understanding. The model uses an off-the-shelf tracker and grounding module to localize objects in videos based on user prompts. PG-Video-LLaVA is evaluated on video question-answering and grounding benchmarks, using Vicuna instead of GPT-3.5 for reproducibility.
This paper introduces a predictive analysis of Arabic court decisions, utilizing 10,813 real commercial court cases. The study evaluates LLaMA-7b, JAIS-13b, and GPT3.5-turbo models under zero-shot, one-shot, and fine-tuned training paradigms, also experimenting with summarization and translation. GPT-3.5 models significantly outperformed others, exceeding JAIS model performance by 50%, while also demonstrating the unreliability of most automated metrics. Why it matters: This research bridges computational linguistics and Arabic legal analytics, offering insights for enhancing judicial processes and legal strategies in the Arabic-speaking world.
MBZUAI has released Jais and Jais-chat, two new open generative large language models (LLMs) with a focus on Arabic. The 13 billion parameter models are based on the GPT-3 architecture and pretrained on Arabic, English, and code. Evaluation shows state-of-the-art Arabic knowledge and reasoning, with competitive English performance.
LAraBench introduces a benchmark for Arabic NLP and speech processing, evaluating LLMs like GPT-3.5-turbo, GPT-4, BLOOMZ, Jais-13b-chat, Whisper, and USM. The benchmark covers 33 tasks across 61 datasets, using zero-shot and few-shot learning techniques. Results show that SOTA models generally outperform LLMs in zero-shot settings, though larger LLMs with few-shot learning reduce the gap. Why it matters: This benchmark helps assess and improve the performance of LLMs on Arabic language tasks, highlighting areas where specialized models still excel.
MBZUAI President Eric Xing led a global collaboration to develop Vicuna, an LLM alternative to GPT-3 addressing the unsustainable costs of training LLMs. OpenAI CEO Sam Altman acknowledged Abu Dhabi's role in the global AI conversation, building off of achievements like Vicuna. Xing and colleagues are publishing research at MLSys 2023 on "cross-mesh resharding" to improve computer communication in deep learning, aiming for low-carbon, affordable, and miniaturized AI. Why it matters: This research signals a push towards sustainable AI development in the region, emphasizing efficiency and reduced environmental impact.