diff --git a/README.md b/README.md
index de175c7..f2da4eb 100644
--- a/README.md
+++ b/README.md
@@ -80,7 +80,7 @@ To evaluate a submission in [Kaggle ARC format](https://www.kaggle.com/competiti
 ```python
 >>> eval_set.score_submission(
     'submission.csv', # Submission with two columns output_id,output in Kaggle fomrat
-    topn=3,           # How many predictions to consider (default: 3)
+    topn=2,           # How many predictions to consider (default: 2)
     return_correct=False # Whether to return a list of which tasks were solved
     )
 ```
diff --git a/arckit/data.py b/arckit/data.py
index 0380922..67008aa 100644
--- a/arckit/data.py
+++ b/arckit/data.py
@@ -185,7 +185,7 @@ def __iter__(self):
     def __repr__(self):
         return f"<TaskSet: {len(self.tasks)} tasks>"
 
-    def score_submission(self,fn: str, topn=3, return_correct=False) -> int:
+    def score_submission(self,fn: str, topn=2, return_correct=False) -> int:
         """
         Score a submission file, in Kaggle csv format
         Two columns: output_id,output
@@ -303,4 +303,4 @@ def load_single(id: str, version='latest') -> Task:
     for i in range(10):
         train_tasks[i].show()
 
-    print(train_tasks["08ed6ac7"].gpt_prompt("gpt3"))
\ No newline at end of file
+    print(train_tasks["08ed6ac7"].gpt_prompt("gpt3"))