Commit
•
233b549
1
Parent(s):
2da0010
Update README.md
Browse files
README.md
CHANGED
|
@@ -20,6 +20,8 @@ Classifies sentences as "Likely" or "Unlikely" biased/disinformation (max token
|
|
| 20 |
|
| 21 |
Fine-tuned BERT ([bert-base-uncased](https://huggingface.co/google-bert/bert-base-uncased)) on the `headline` and `text_label` fields in the [News Media Bias Plus Dataset](https://huggingface.co/datasets/vector-institute/newsmediabias-plus).
|
| 22 |
|
|
|
|
|
|
|
| 23 |
### Metics
|
| 24 |
|
| 25 |
*Evaluated on a 0.1 random sample of the NMB+ dataset, unseen during training*
|
|
|
|
| 20 |
|
| 21 |
Fine-tuned BERT ([bert-base-uncased](https://huggingface.co/google-bert/bert-base-uncased)) on the `headline` and `text_label` fields in the [News Media Bias Plus Dataset](https://huggingface.co/datasets/vector-institute/newsmediabias-plus).
|
| 22 |
|
| 23 |
+
**This model was trained without weighted sampling, and the dataset contains 81.9% 'Likely' and 18.1% 'Unlikely' examples.** The same model trained with weighted sampling preformed better when evaluated by gpt-4o-mini as a judge and is available [here](https://huggingface.co/maximuspowers/nmbp-bert-headlines-balanced).
|
| 24 |
+
|
| 25 |
### Metics
|
| 26 |
|
| 27 |
*Evaluated on a 0.1 random sample of the NMB+ dataset, unseen during training*
|