Replace fold-0 numbers with the pooled 5-fold cross-validation

README.md

@@ -49,17 +49,22 @@ Script: `scripts/multitask_train.py` in the narrative-lens repository; provenanc
 
 ## Measurements
 
-Unseen authors (fold 0 of an author-disjoint 5-fold split over LitBank and PDNC):
+Unseen authors (pooled over an author-disjoint 5-fold split over LitBank, PDNC, and the PDNC alias-list
+documents; each fold's teacher also structures ~29,200 windows from the 100 unlabelled Gutenberg novels):
 
-| | student | teacher |
+| | student (pooled) | teacher (range across folds) |
 | --- | ---: | ---: |
-| Proper person spans, exact F1 | 0.812 | 0.829 |
-| Nominal mentions, exact F1 | 0.704 | 0.641 |
-| Antecedent accuracy, pronouns (within window) | 0.908 | 0.923 |
-| Quotation speaker accuracy | 0.814 | 0.863 |
+| Proper person spans, exact F1 | 0.822 | 0.633-0.906 |
+| Nominal mentions, exact F1 | 0.708 | 0.604-0.686 |
+| Antecedent accuracy, proper mentions (within window) | 0.937 | - |
+| Antecedent accuracy, nominal mentions (within window) | 0.859 | - |
+| Antecedent accuracy, pronouns (within window) | 0.935 | 0.925-0.954 |
+| Quotation speaker accuracy | 0.854 | 0.849-0.898 |
 
-The fold-0 figures above are for the v1 student (same recipe without alias lists and Gutenberg self-training);
-the v2 five-fold cross-validation is in progress and will replace them. LitBank development split through the
+The student is close to or ahead of its own teacher on every head except speaker accuracy, where the
+teacher stays a few points ahead. The teacher's token-head scores vary more across folds because
+LitBank's entity and nominal labels are its only supervision, trained on one window grid for two
+epochs against the student's two grids with LitBank repeated four times. LitBank development split through the
 narrative-lens pipeline (the model trained on these works, so this shows the integration, not generalization):
 supplied-cast pronoun resolution F1 0.848 (precision 0.861), explicit-mode speaker F1 0.699 (precision 0.787),
 against 0.786 and 0.657 for v1.