Datasets we use for pretraining large language models
AI & ML interests
We are an interdisciplinary research group based in Türkiye, committed to advancing Natural Language Processing (NLP), Large Language Models (LLM), and Artificial Intelligence (AI) with a strong focus on Turkish and other low-resource languages. Our team brings together researchers from multiple academic institutions, PhD and undergraduate students, and global collaborators from both academia and industry. By combining expertise in linguistics, machine learning, and software engineering, we drive projects that bridge rigorous academic research with real-world applications. Through open science principles, we create open-source tools, large-scale datasets, and practical AI solutions. Our goal is to establish a robust research ecosystem that accelerates innovation and enhances the accessibility and impact of AI technologies for Turkish and similar languages worldwide.
-
Tokens with Meaning: A Hybrid Tokenization Approach for NLP
Paper • 2508.14292 • Published • 1 -
Doğal Dil İşlemede Tokenizasyon Standartları ve Ölçümü: Türkçe Üzerinden Büyük Dil Modellerinin Karşılaştırmalı Analizi
Paper • 2508.13058 • Published • 1 -
Büyük Dil Modelleri için TR-MMLU Benchmarkı: Performans Değerlendirmesi, Zorluklar ve İyileştirme Fırsatları
Paper • 2508.13044 • Published • 1 -
Tokenization Standards for Linguistic Integrity: Turkish as a Benchmark
Paper • 2502.07057 • Published
Datasets we use for pretraining large language models
-
Tokens with Meaning: A Hybrid Tokenization Approach for NLP
Paper • 2508.14292 • Published • 1 -
Doğal Dil İşlemede Tokenizasyon Standartları ve Ölçümü: Türkçe Üzerinden Büyük Dil Modellerinin Karşılaştırmalı Analizi
Paper • 2508.13058 • Published • 1 -
Büyük Dil Modelleri için TR-MMLU Benchmarkı: Performans Değerlendirmesi, Zorluklar ve İyileştirme Fırsatları
Paper • 2508.13044 • Published • 1 -
Tokenization Standards for Linguistic Integrity: Turkish as a Benchmark
Paper • 2502.07057 • Published