TabiBERT: A Large-Scale ModernBERT Foundation Model and A Unified Benchmark for Turkish
TabiBERT is a large-scale Turkish encoder based on ModernBERT architecture, trained on massive text data to improve performance in Turkish NLP. It uses new features like RoPE and FlashAttention for better context handling. TabiBench provides a standardized benchmark to rigorously compare different models, showing TabiBERT achieves state-of-the-art results.