# GLiNER: Generalist Model for Named Entity Recognition using Bidirectional Transformer

**Authors:** Urchade Zaratiana, Nadi Tomeh, Pierre Holat, Thierry Charnois  
**arXiv:** [2311.08526](https://arxiv.org/abs/2311.08526)  
**Date:** 2023-11-14  
**Summary page:** https://www.summarizepaper.com/en/paper/2311.08526v1/

## Key points

- Growing interest in fine-tuning open source language models like LLaMa and InstructUIE for named entity recognition tasks
- Introduction of GLiNER, a compact NER model trained to efficiently identify any type of entity
- Leveraging a bidirectional transformer encoder architecture for parallel entity extraction
- Outperforming traditional LLMs in zero-shot evaluations
- Comparison with recent models designed for open-type NER tasks using prompting techniques from previous studies

## Summary

, , , , 


In recent research efforts, there has been a growing interest in fine-tuning open source language models like LLaMa and InstructUIE for named entity recognition tasks. These approaches have shown promising results, but also have limitations that need to be addressed. This paper introduces GLiNER, a compact NER model trained to efficiently identify any type of entity. Leveraging a bidirectional transformer encoder architecture, GLiNER enables parallel entity extraction and outperforms traditional LLMs in zero-shot evaluations. Additionally, the study compares GLiNER with recent models designed for open-type NER tasks using prompting techniques from previous studies. By addressing key limitations and showcasing strong performance results, GLiNER stands out as a promising solution for efficient and effective named entity recognition in natural language processing applications.