Update README.md
Browse files
README.md
CHANGED
|
@@ -27,6 +27,11 @@ We also have other two MiniCheck model variants:
|
|
| 27 |
|
| 28 |
|
| 29 |
### Model Performance
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 30 |
The performance of these models is evaluated on our new collected benchmark (unseen by our models during training), [LLM-AggreFact](https://huggingface.co/datasets/lytang/LLM-AggreFact),
|
| 31 |
from 10 recent human annotated datasets on fact-checking and grounding LLM generations. MiniCheck-DeBERTa-v3-Large outperform all
|
| 32 |
exisiting specialized fact-checkers with a similar scale by a large margin but is 2% worse than our best model MiniCheck-Flan-T5-Large, which
|
|
|
|
| 27 |
|
| 28 |
|
| 29 |
### Model Performance
|
| 30 |
+
|
| 31 |
+
<p align="center">
|
| 32 |
+
<img src="./cost-vs-bacc.png" width="360">
|
| 33 |
+
</p>
|
| 34 |
+
|
| 35 |
The performance of these models is evaluated on our new collected benchmark (unseen by our models during training), [LLM-AggreFact](https://huggingface.co/datasets/lytang/LLM-AggreFact),
|
| 36 |
from 10 recent human annotated datasets on fact-checking and grounding LLM generations. MiniCheck-DeBERTa-v3-Large outperform all
|
| 37 |
exisiting specialized fact-checkers with a similar scale by a large margin but is 2% worse than our best model MiniCheck-Flan-T5-Large, which
|