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/graphs/family_relationships.py

@@ -358,16 +358,14 @@ class FamilyRelationshipsDataset(ProceduralDataset):
 
     def score_answer(self, answer: Optional[str], entry: dict[str, Any]) -> float:
         reward = 0.0
-        if answer is not None:
+        if isinstance(answer, str):
             try:
                 answer_formatted = answer.strip().lower()
-                solved = answer_formatted == entry["answer"].strip().lower()
-                if solved:
+                oracle_answer = entry["answer"].strip().lower()
+                if answer_formatted == oracle_answer:
                     reward = 1.0
-                else:
-                    reward = 0.01
             except:
-                reward = 0.01
+                pass
         return reward
 
 

reasoning_gym/graphs/quantum_lock.py

@@ -169,21 +169,15 @@ Buttons:
 
         The function awards 1.0 for a correct answer and less otherwise.
         """
-        if answer == None:
+        if not isinstance(answer, str):
             return 0.0
 
-        # Get correct solution from metadata
-        correct_solution = entry["metadata"].get("solution_path", [])
-
         # Normalize both answers
-        def normalize_seq(seq):
-            """Handle both string and list inputs by converting to string first"""
-            # Convert sequence to string representation if it's a list
-            input_str = "".join(seq) if isinstance(seq, list) else str(seq or "")
-            return [c.upper() for c in re.findall(r"[A-C]", input_str.upper())]
+        def normalize_seq(seq: str) -> list[str]:
+            return [c.upper() for c in re.findall(r"[A-C]", seq.upper())]
 
         user_sequence = normalize_seq(answer)
-        target_sequence = normalize_seq("".join(correct_solution))
+        target_sequence = normalize_seq(entry["answer"])
 
         # Exact sequence match required
         if user_sequence == target_sequence:
@@ -196,7 +190,7 @@ Buttons:
                 return 1.0  # Different answer, but qually correct
             return 0.5  # Alternative scoring - you're correct, but not optimal
 
-        return 0.1
+        return 0.0
 
     def simulate_sequence(self, metadata: dict, sequence: list[str]) -> int:
         """Simulate button presses to verify solutions"""

reasoning_gym/graphs/shortest_path.py

@@ -125,8 +125,8 @@ class ShortestPathDataset(ProceduralDataset):
 
     def score_answer(self, answer: Optional[str], entry: dict[str, Any]) -> float:
         """Overwrite this method in derived classes if a single oracle answer is not available."""
-        oracle_answer = entry["answer"].strip()
-        if answer is not None and len(answer) > 0:
+        if isinstance(answer, str) and len(answer) > 0:
+            oracle_answer = entry["answer"].strip()
             answer = answer.strip()
 
             # Exact answer
@@ -145,8 +145,6 @@ class ShortestPathDataset(ProceduralDataset):
             elif self._is_valid_path(matrix, answer):
                 return 0.5
 
-            return 0.01
-
         return 0.0
 
     def __getitem__(self, idx: int) -> dict: