Update README.md

README.md

@@ -72,7 +72,9 @@ The dataset contains four splits:
   - **op5_train**: five-option multiple-choice QA (choices A-E)
   - **op5_eval**: five-option multiple-choice QA (choices A-E)
 
-The train/eval splits for a given number of options are 80/20, where the last 20% of the original, ordered dataset is used for evals.
+`op5` splits contain the same content as `op4` splits, but with one additional answer choice to increase difficulty. Note that while the content is the same, the letter choice corresponding to the correct answer is sometimes different between these splits.
+
+The train/eval splits for a given number of options are 80/20, where the last 20% of the original, complete dataset is used for evals.
 
 ### Dataset Sources
 
@@ -105,18 +107,10 @@ if __name__ == "__main__":
     # load only op5 splits
     op5_train, op5_eval = load_dataset("mkieffer/Medbullets", split=["op5_train", "op5_eval"])
 
-    # optionally, combine them into a single dataset
-    dataset = DatasetDict({
-        "op4_train": op4_train,
-        "op4_eval":  op4_eval,
-        "op5_train": op5_train,
-        "op5_eval":  op5_eval,
-    })
-
-    print("\nop4_train:\n", json.dumps(dataset["op4_train"][0], indent=2))
-    print("\nop4_eval:\n", json.dumps(dataset["op4_eval"][0], indent=2))
-    print("\nop5_train:\n", json.dumps(dataset["op5_train"][0], indent=2))
-    print("\nop5_eval:\n", json.dumps(dataset["op5_eval"][0], indent=2))
+    print("\nop4_train:\n", json.dumps(op4_train[0], indent=2))
+    print("\nop4_eval:\n", json.dumps(op4_eval[0], indent=2))
+    print("\nop5_train:\n", json.dumps(op5_train[0], indent=2))
+    print("\nop5_eval:\n", json.dumps(op5_eval[0], indent=2))
 ```