Update README.md
Browse files
README.md
CHANGED
@@ -19,6 +19,8 @@ inference: false
|
|
19 |
|
20 |
This model provides the best embedding for the Entity Recognition task in English.
|
21 |
|
|
|
|
|
22 |
This is the model from our [**Paper**](https://arxiv.org/abs/2402.15343): **NuNER: Entity Recognition Encoder Pre-training via LLM-Annotated Data**
|
23 |
|
24 |
|
@@ -36,7 +38,7 @@ Read more about evaluation protocol & datasets in our [paper](https://arxiv.org/
|
|
36 |
|
37 |
Here is the aggregated performance of the models over several datasets.
|
38 |
|
39 |
-
k=X means that as
|
40 |
|
41 |
| Model | k=1 | k=4 | k=16 | k=64 |
|
42 |
|----------|----------|----------|----------|----------|
|
@@ -44,7 +46,7 @@ k=X means that as a training data for this evaluation, we took only X examples f
|
|
44 |
| RoBERTa-base + NER-BERT pre-training | 32.3 | 50.9 | 61.9 | 67.6 |
|
45 |
| NuNER v1.0 | **39.4** | **59.6** | **67.8** | **71.5** |
|
46 |
|
47 |
-
NuNER v1.0 has similar performance to 7B LLMs (70 times bigger
|
48 |
|
49 |
| Model | k=8~16| k=64~128 |
|
50 |
|----------|----------|----------|
|
|
|
19 |
|
20 |
This model provides the best embedding for the Entity Recognition task in English.
|
21 |
|
22 |
+
We suggest using **newer version of this model: [NuNER v2.0](https://huggingface.co/numind/NuNER-v2.0)**
|
23 |
+
|
24 |
This is the model from our [**Paper**](https://arxiv.org/abs/2402.15343): **NuNER: Entity Recognition Encoder Pre-training via LLM-Annotated Data**
|
25 |
|
26 |
|
|
|
38 |
|
39 |
Here is the aggregated performance of the models over several datasets.
|
40 |
|
41 |
+
k=X means that as training data for this evaluation, we took only X examples for each class, trained the model, and evaluated it on the full test set.
|
42 |
|
43 |
| Model | k=1 | k=4 | k=16 | k=64 |
|
44 |
|----------|----------|----------|----------|----------|
|
|
|
46 |
| RoBERTa-base + NER-BERT pre-training | 32.3 | 50.9 | 61.9 | 67.6 |
|
47 |
| NuNER v1.0 | **39.4** | **59.6** | **67.8** | **71.5** |
|
48 |
|
49 |
+
NuNER v1.0 has similar performance to 7B LLMs (70 times bigger than NuNER v1.0) created specifically for the NER task.
|
50 |
|
51 |
| Model | k=8~16| k=64~128 |
|
52 |
|----------|----------|----------|
|