Add 5MT-3digit config to dataset card

README.md

@@ -12,6 +12,12 @@ configs:
   data_files:
   - split: train
     path: 10BT/train-*.parquet
+- config_name: 5MT-3digit
+  data_files:
+  - split: train
+    path: 5MT-3digit/train-*.parquet
+  - split: test
+    path: 5MT-3digit/test-*.parquet
 ---
 
 # Addition Dataset
@@ -23,11 +29,13 @@ Addition problems in the format `{a} + {b} = {c}` where operands have stratified
 - **test**: 5K held-out evaluation examples (operands >= 10, i.e. min 2 digits)
 - **1BT**: ~85M training examples (~1 billion tokens under Llama-3 tokenizer)
 - **10BT**: ~850M training examples (~10 billion tokens)
+- **5MT-3digit**: Exhaustive 3-digit addition (all 900x900 = 810K ordered pairs, ~5.35M tokens). Symmetry-safe train/test split (10% test).
 
 ## Deduplication
 
 - Commutative dedup: if `a + b = c` exists, `b + a = c` is excluded
 - Test exclusion: both orderings of test-set pairs are excluded from train splits
+- 5MT-3digit: both orderings always in the same split (no commutative leakage)
 
 ## Usage
 
@@ -37,4 +45,8 @@ from datasets import load_dataset
 train = load_dataset("deqing/addition_dataset", "1BT", split="train")
 train_10b = load_dataset("deqing/addition_dataset", "10BT", split="train")
 test = load_dataset("deqing/addition_dataset", "test", split="test")
+
+# 3-digit exhaustive
+train_3d = load_dataset("deqing/addition_dataset", "5MT-3digit", split="train")
+test_3d = load_dataset("deqing/addition_dataset", "5MT-3digit", split="test")
 ```