Update README.md
Browse files
README.md
CHANGED
|
@@ -2,11 +2,6 @@
|
|
| 2 |
|
| 3 |
This model is a dual-head reward model trained for literary prose. It features a style head for ranking prose quality and a faithfulness head used as a veto mechanism.
|
| 4 |
|
| 5 |
-
## Training Metrics
|
| 6 |
-
- **Pair Accuracy**: 0.939
|
| 7 |
-
- **Style Pair Accuracy**: 0.939
|
| 8 |
-
- **Faithfulness Veto Accuracy**: 1.000
|
| 9 |
-
|
| 10 |
## Architecture
|
| 11 |
- **Backbone**: answerdotai/ModernBERT-large
|
| 12 |
- **Objective**: style_primary_faith_veto_flat_copy_length_gm_v3
|
|
|
|
| 2 |
|
| 3 |
This model is a dual-head reward model trained for literary prose. It features a style head for ranking prose quality and a faithfulness head used as a veto mechanism.
|
| 4 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 5 |
## Architecture
|
| 6 |
- **Backbone**: answerdotai/ModernBERT-large
|
| 7 |
- **Objective**: style_primary_faith_veto_flat_copy_length_gm_v3
|