docs: fix duplicated 'to' typo in finetuning guidance

FINETUNING.md

@@ -60,7 +60,7 @@ Notes:
 Canonical harness scripts live in `examples/scripts/finetuning/`:
 
 1. `finetune_secret_demo.sh`
-   - The demo dataset contains strings redacted as <ACCOUNT_NUMBER> by the baseline model, but with ground truths labeled as <SECRET> instead. It illustrates how the model can be retrained to adapt to  to the category policy change.
+   - The demo dataset contains strings redacted as <ACCOUNT_NUMBER> by the baseline model, but with ground truths labeled as <SECRET> instead. It illustrates how the model can be retrained to adapt to the category policy change.
 2. `finetune_custom_label_demo.sh`
    - The demo defines a new label space consisting only of the background class ("O") and a single, newly defined "custom_secret" category. It illustrates how the model can be adapted to recognize this new category instead of its original categories.
 

examples/scripts/finetuning/finetune_secret_demo.sh

@@ -137,7 +137,7 @@ for split in train validation test; do
 done
 
 echo "Finetuning Privacy Filter: Demo 1."
-echo "The demo dataset contains strings redacted as <ACCOUNT_NUMBER> by the baseline model, but with ground truths labeled as <SECRET> instead. It illustrates how the model can be retrained to adapt to  to the category policy change."
+echo "The demo dataset contains strings redacted as <ACCOUNT_NUMBER> by the baseline model, but with ground truths labeled as <SECRET> instead. It illustrates how the model can be retrained to adapt to the category policy change."
 echo "Training data: ${DATA_DIR}/train.jsonl"
 echo "Validation data: ${DATA_DIR}/validation.jsonl"
 echo "Test data: ${DATA_DIR}/test.jsonl"