Update README.md

README.md

@@ -19,9 +19,9 @@ for **machine unlearning** experiments.
 ## Source
 
 - Original dataset: `roupenminassian/twitter-misinformation`
-- Label:
-  - `0` → factual
-  - `1` → misinformation
+- Fields:
+- `text` (string): tweet text.
+- `label` (int): `0` for factual, `1` for misinformation.
 
 ## Splits
 
@@ -33,15 +33,9 @@ We construct three splits from the **test** partition of the original dataset:
 
 - `retain1`:
   - Contains `label = 0` (factual) examples.
-  - Size ≈ `2 × |forget|`.
-
+    
 - `retain2`:
   - Contains `label = 0` (factual) examples.
-  - Size ≈ `2 × |forget|`.
-
-Overall ratio (if enough data is available) is approximately:
-
-`|forget| : |retain1| : |retain2| = 1 : 2 : 2`.
 
 ## Intended Use
 
@@ -50,9 +44,3 @@ Overall ratio (if enough data is available) is approximately:
   - Forget all misinformation (`forget`).
   - Preserve utility on factual content (`retain1`, `retain2`).
 
-## Fields
-
-- `text` (string): tweet text.
-- `label` (int): `0` for factual, `1` for misinformation.
-- `split_role` (string): one of `forget`, `retain1`, `retain2`.
-