expertai/SLIMER-PARALLEL-LLaMA3

expertai/SLIMER-PARALLEL-LLaMA3 is an 8 billion parameter LLaMA-3 based language model specifically instructed for zero-shot Named Entity Recognition (NER) on English text. It significantly improves over previous SLIMER models, achieving a +17% performance increase and enabling parallel extraction of up to 16 Named Entities per prompt. This model excels at identifying never-seen-before Named Entity tags by leveraging prompt-enriched definitions and guidelines, making it highly effective for out-of-distribution NER tasks.