metadata
license: cc-by-nc-4.0
language:
- en
RA-IT-NER-8B
Description: The RA-IT-NER-8B model is trained from LLama3-8B using the proposed Retrieval Augmented Instruction Tuning (RA-IT) approach. This model can be used for English Open NER with and without RAG. The training data is the Pile-NER-type presented by UniversalNER.
Check our paper for more information. Check our github repo about how to use the model.
Inference
The template for inference instances is as follows:
Prompting template:
USER: Here are some examples of named entity recognition: {Fill the NER examples here}
ASSISTANT: I’ve read these examples.
USER: Text: {Fill the input text here}
ASSISTANT: I’ve read this text.
USER: What describes {Fill the entity type here} in the text?
ASSISTANT: (model's predictions in JSON format)
USER: Here are some examples of named entity recognition: {Fill the NER examples here}
ASSISTANT: I’ve read these examples.
USER: Text: {Fill the input text here}
ASSISTANT: I’ve read this text.
USER: What describes {Fill the entity type here} in the text?
ASSISTANT: (model's predictions in JSON format)
Note:
- The model can conduct inference with and without NER examples. If you want to conduct inference without examples, just start from the third line in the above template by directly inputting "Text: {input text}" in the "USER" role.
- Inferences are based on one entity type at a time. For multiple entity types, create separate instances for each type.
License
This model is released under the CC BY-NC 4.0 license. It is primarily used for research purposes.