Create README.md

README.md

@@ -1,28 +1,85 @@
 ---
-dataset_info:
-  features:
-  - name: facts
-    dtype: string
-  - name: formula
-    dtype: string
-  - name: num_operators
-    dtype: int64
-  - name: gold_formula
-    dtype: bool
-  splits:
-  - name: train
-    num_bytes: 3078137
-    num_examples: 15941
-  - name: test
-    num_bytes: 764386
-    num_examples: 3985
-  download_size: 1229290
-  dataset_size: 3842523
-configs:
-- config_name: default
-  data_files:
-  - split: train
-    path: data/train-*
-  - split: test
-    path: data/test-*
+annotations_creators:
+- no-annotation
+language_creators:
+- found
+language:
+- en
+license:
+- apache-2.0
+multilinguality:
+- monolingual
+size_categories:
+- 10K<n<100K
+source_datasets:
+- generated/synthetic
+task_categories:
+- text-classification
+- other
 ---
+
+
+# Implications
+
+## Dataset Description
+
+**Implications** is a synthetic dataset of logical formulas with Boolean variables and logical operators (`and`, `or`, `implies`).  
+Each example contains:
+
+- `facts`: truth values assigned to variables (e.g., `a is true, b is false, ...`)  
+- `formula`: a linear formula without parentheses (e.g., `a and b implies c or d`)  
+- `num_operators`: the number of logical operators in the formula  
+- `gold_formula`: the overall truth value of the formula (`True` or `False`)
+
+All duplicate `(facts, formula)` pairs have been removed to ensure that each combination is unique.
+
+The dataset is designed for training models on **logical reasoning**, **linear probing**, or **LLMs** to predict the truth value of formulas.
+
+---
+
+## Data Fields
+
+| Column          | Type    | Description |
+|-----------------|---------|-------------|
+| facts           | string  | Boolean values assigned to variables |
+| formula         | string  | Linear formula without parentheses |
+| num_operators   | int     | Number of logical operators in the formula |
+| gold_formula    | bool    | Truth value of the full formula |
+
+---
+
+## Example
+
+```csv
+facts,formula,num_operators,gold_formula
+"a is true, b is false, c is true","a and b or c implies d",5,True
+"p is false, q is true","p or q implies r",2,False
+
+
+### `num_operators` counts (both in train and in test)
+
+| num_operators | train_count | test_count |
+|---------------|------------|-----------|
+| 1  | 730 | 196 |
+| 2  | 775 | 225 |
+| 3  | 811 | 189 |
+| 4  | 814 | 186 |
+| 5  | 788 | 212 |
+| 6  | 799 | 201 |
+| 7  | 788 | 212 |
+| 8  | 800 | 200 |
+| 9  | 808 | 192 |
+| 10 | 795 | 205 |
+| 11 | 806 | 194 |
+| 12 | 787 | 213 |
+| 13 | 817 | 183 |
+| 14 | 806 | 194 |
+| 15 | 817 | 183 |
+| 16 | 819 | 181 |
+| 17 | 815 | 185 |
+| 18 | 793 | 207 |
+| 19 | 778 | 222 |
+| 20 | 795 | 205 |
+
+**Note:** All duplicate `(facts, formula)` pairs have been removed to ensure each example is unique.
+