Researchers at MBZUAI release SlimPajama-DC, an empirical analysis of data combinations for pretraining LLMs using the SlimPajama dataset. The study examines the impact of global vs. local deduplication and the proportions of highly-deduplicated multi-source datasets. Results show that increased data diversity after global deduplication is crucial, with the best configuration outperforming models trained on RedPajama.
This paper studies the impact of data scale on Arabic Pretrained Language Models (PLMs). Researchers retrained BERT-base and T5-base models on large Arabic corpora, achieving state-of-the-art results on the ALUE and ORCA benchmarks. The analysis indicates that pretraining data volume is the most important factor for performance. Why it matters: This work provides valuable insights into building effective Arabic language models, emphasizing the importance of large, high-quality datasets for advancing Arabic NLP.
Researchers from the National Center for AI in Saudi Arabia investigated the sensitivity of Large Language Model (LLM) leaderboards to minor benchmark perturbations. They found that small changes, like choice order, can shift rankings by up to 8 positions. The study recommends hybrid scoring and warns against over-reliance on simple benchmark evaluations, providing code for further research.
Researchers at MBZUAI have demonstrated a method called "Data Laundering" to artificially boost language model benchmark scores using knowledge distillation. The technique covertly transfers benchmark-specific knowledge, leading to inflated accuracy without genuine improvements in reasoning. The study highlights a vulnerability in current AI evaluation practices and calls for more robust benchmarks.
This paper introduces two methods for creating Arabic LLM prompts at scale: translating existing English prompt datasets and creating natural language prompts from Arabic NLP datasets. Using these methods, the authors generated over 67.4 million Arabic prompts covering tasks like summarization and question answering. Fine-tuning a 7B Qwen2 model on these prompts outperforms a 70B Llama3 model in handling Arabic prompts. Why it matters: The research provides a cost-effective approach to scaling Arabic LLM training data, potentially improving the performance of smaller, more accessible models for Arabic NLP.