expertai/SLIMER-LLaMA3

Hugging Face
TEXT GENERATIONPricing:Input $0.2 / Cached $0.028 / Output $0.32Concurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Oct 8, 2024License:llama3.1Architecture:Transformer0.0K Featherless Exclusive Warm

expertai/SLIMER-LLaMA3 is an 8 billion parameter LLaMA-3 based model developed by expertai, specifically instruction-tuned for zero-shot Named Entity Recognition (NER) on English text. It leverages a prompt enriched with definitions and guidelines for target NEs, enabling it to effectively extract never-seen-before entity tags. This model performs comparably to state-of-the-art NER models on out-of-distribution input domains while being trained on a reduced number of samples, excelling particularly with novel financial entities.

Loading preview...

SLIMER-LLaMA3: Zero-Shot Named Entity Recognition

expertai/SLIMER-LLaMA3 is an 8 billion parameter model built on the LLaMA-3 architecture, specifically designed for zero-shot Named Entity Recognition (NER) in English. This model distinguishes itself by its instruction-tuning methodology, which utilizes prompts enriched with explicit definitions and guidelines for the Named Entity types to be extracted. This approach allows SLIMER-LLaMA3 to effectively identify and extract novel, previously unseen entity tags.

Key Capabilities

  • Zero-Shot NER: Excels at extracting Named Entities without prior examples for specific entity types, leveraging detailed instructions.
  • Robustness to Novel Entities: Demonstrates strong performance on out-of-distribution (OOD) input domains, particularly with challenging financial entities (BUSTER benchmark).
  • Efficient Instruction Tuning: Achieves competitive results with a reduced number of training samples compared to models fine-tuned on extensive entity classes.
  • JSON Output: Designed to return extracted entities as a JSON list, facilitating programmatic use.

Good For

  • Applications requiring flexible and adaptable Named Entity Recognition for diverse and evolving entity types.
  • Scenarios where traditional NER models struggle with unseen or domain-specific entities.
  • Developers needing a robust zero-shot NER solution that can be guided by clear definitions and rules.
  • Research into instruction-tuned LLMs for information extraction tasks.

Finetunes
4 models