andrewzamai
commited on
Commit
•
7068ade
1
Parent(s):
5dc088a
Update README.md
Browse files
README.md
CHANGED
@@ -9,12 +9,14 @@ pipeline_tag: text-generation
|
|
9 |
|
10 |
# SLIMER: Show Less Instruct More Entity Recognition
|
11 |
|
12 |
-
SLIMER is an instruction-tuned LLM (LLaMA-2-7B based) for NER.
|
13 |
|
14 |
-
Instructed on a reduced number of samples, it is designed to tackle never-seen-before named entity tags by leveraging a prompt enriched with a
|
15 |
|
16 |
<img src="https://huggingface.co/expertai/SLIMER/resolve/main/SLIMER_instruction_prompt.png" width="200">
|
17 |
|
18 |
-
|
|
|
|
|
|
|
19 |
|
20 |
-
<img src="https://huggingface.co/expertai/SLIMER/resolve/main/OOD_evals.png">
|
|
|
9 |
|
10 |
# SLIMER: Show Less Instruct More Entity Recognition
|
11 |
|
12 |
+
SLIMER is an instruction-tuned LLM (LLaMA-2-7B based) for zero-shot NER.
|
13 |
|
14 |
+
Instructed on a reduced number of samples, it is designed to tackle never-seen-before named entity tags by leveraging a prompt enriched with a DEFINITION and GUIDELINES for the NE to be extracted.
|
15 |
|
16 |
<img src="https://huggingface.co/expertai/SLIMER/resolve/main/SLIMER_instruction_prompt.png" width="200">
|
17 |
|
18 |
+
Currently existing LLMs for NER fine-tune on an extensive number of entity classes (around 13K) and assess zero-shot NER capabilities on Out-Of-Distribution input domains.
|
19 |
+
SLIMER performs comparably to these state-of-the-art approaches on OOD input domains, while being trained only a reduced number of samples and a set of NE tags that overlap in lesser degree with test set.
|
20 |
+
|
21 |
+
<img src="https://huggingface.co/expertai/SLIMER/resolve/main/OOD_evals.png">
|
22 |
|
|