Skip to content
GCC AI Research

GATech at AbjadMed: Bidirectional Encoders vs. Causal Decoders: Insights from 82-Class Arabic Medical Classification

arXiv · · Notable

NLP Research Arabic AI Healthcare LLM

Summary

Researchers from Georgia Tech explored Arabic medical text classification using 82 categories from the AbjadMed dataset. They compared fine-tuned AraBERTv2 encoders with hybrid pooling against multilingual encoders and large causal decoders like Llama 3.3 70B and Qwen 3B. The study found that bidirectional encoders outperformed causal decoders in capturing semantic boundaries for fine-grained medical text classification. Why it matters: The research provides insights into optimal model selection for specialized Arabic NLP tasks, specifically highlighting the effectiveness of fine-tuned encoders for medical text categorization.

Keywords

Arabic NLP · medical text classification · AraBERT · causal decoders · AbjadMed

Read original article →

Get the weekly digest

Top AI stories from the GCC region, every week.

Related

AraModernBERT: Transtokenized Initialization and Long-Context Encoder Modeling for Arabic

arXiv ·

The paper introduces AraModernBERT, an adaptation of the ModernBERT encoder architecture for Arabic, focusing on transtokenized embedding initialization and long-context modeling up to 8,192 tokens. Transtokenization is shown to be crucial for Arabic language modeling, significantly enhancing masked language modeling performance. The model demonstrates stable and effective long-context modeling, improving intrinsic language modeling performance at extended sequence lengths. Why it matters: This research provides practical insights for adapting modern encoder architectures to Arabic and other languages using Arabic-derived scripts, advancing Arabic NLP.

Benchmarking the Medical Understanding and Reasoning of Large Language Models in Arabic Healthcare Tasks

arXiv ·

This paper benchmarks the performance of large language models (LLMs) on Arabic medical natural language processing tasks using the AraHealthQA dataset. The study evaluated LLMs in multiple-choice question answering, fill-in-the-blank, and open-ended question answering scenarios. The results showed that a majority voting solution using Gemini Flash 2.5, Gemini Pro 2.5, and GPT o3 achieved 77% accuracy on MCQs, while other LLMs achieved a BERTScore of 86.44% on open-ended questions. Why it matters: The research highlights both the potential and limitations of current LLMs in Arabic clinical contexts, providing a baseline for future improvements in Arabic medical AI.