Minor question template & score_answer improvements (#261)

* math prompt improvements
* ignore brackets in complex_arithmetic results
* improve additional instruction in prompt of polynomial_equations
* more strict tests for score_answer in polynomial_equations
* simplify special reward handling
* fix test_intermediate_integration
* fix sokoban dataset
* add common dataset score_answer consistency test

reasoning_gym/code/bf.py

@@ -121,20 +121,23 @@ int main() {{
             float: The computed score between 0.0 and 1.0.
         """
 
-        if answer == None:
+        if not isinstance(answer, str):
             return 0.0
-        if answer != entry["answer"]:
-            if entry["answer"] in answer.splitlines():
-                # We can be quite confident that the correct answer was given
-                # It was likely just given alongside an explanation
-                return max(0.9 * len(answer) / len(entry["answer"]), 0.1)
-            if entry["answer"] in answer:
-                # Since answers are English words, some risk of the response coincidentally containing the answer
-                return max(0.5 * len(answer) / len(entry["answer"]), 0.1)
-            return 0.01
-        else:
+
+        if answer == entry["answer"]:
             return 1.0  # Yay
 
+        if entry["answer"] in answer.splitlines():
+            # We can be quite confident that the correct answer was given
+            # It was likely just given alongside an explanation
+            return max(0.9 * len(answer) / len(entry["answer"]), 0.1)
+
+        if entry["answer"] in answer:
+            # Since answers are English words, some risk of the response coincidentally containing the answer
+            return max(0.5 * len(answer) / len(entry["answer"]), 0.1)
+
+        return 0.0
+
 
 # Register the dataset
 register_dataset("bf", BFDataset, BFConfig)