Remove the requested metric configuration paragraph from the model card
Browse files
README.md
CHANGED
|
@@ -75,11 +75,6 @@ ties. No reference-based tuning was performed.
|
|
| 75 |
| SPICE | 0.205827 |
|
| 76 |
| BERTScore F1 | 0.939820 |
|
| 77 |
|
| 78 |
-
COCO metrics use Stanford PTB tokenization and pycocoevalcap 1.2. BERTScore
|
| 79 |
-
uses RoBERTa-large layer 17, slow tokenization, no IDF or baseline rescaling,
|
| 80 |
-
and all available raw reference captions. Exact metric configurations, package
|
| 81 |
-
versions and evaluated-input hashes are included in `evaluation/`.
|
| 82 |
-
|
| 83 |
Additional tests:
|
| 84 |
|
| 85 |
- **41,849 held-out pseudo-labeled pairs:** accuracy 0.890893, F1 0.904035,
|
|
|
|
| 75 |
| SPICE | 0.205827 |
|
| 76 |
| BERTScore F1 | 0.939820 |
|
| 77 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 78 |
Additional tests:
|
| 79 |
|
| 80 |
- **41,849 held-out pseudo-labeled pairs:** accuracy 0.890893, F1 0.904035,
|