Fixed reward

tasks/easy.py

@@ -173,7 +173,9 @@ class EasyTaskGrader:
         raw = self.correct_actions / self.total_actions
         # Enforce strict (0, 1) range
         clamped = 0.01 + 0.98 * raw
-        return round(float(clamped), 6)
+        rounded = round(float(clamped), 2)
+        # Ensure no rounding to boundaries (0.0 or 1.0)
+        return max(0.01, min(rounded, 0.99))
 
 
     def passed(self) -> bool:

tasks/hard.py

@@ -398,7 +398,9 @@ class HardTaskGrader:
 
         # Enforce strict (0, 1) range
         clamped = 0.01 + 0.98 * final_base
-        return round(float(clamped), 6)
+        rounded = round(float(clamped), 2)
+        # Ensure no rounding to boundaries (0.0 or 1.0)
+        return max(0.01, min(rounded, 0.99))
 
 
     def passed(self) -> bool:

tasks/medium.py

@@ -222,7 +222,9 @@ class MediumTaskGrader:
         base_score = max(0.0, raw - fp_penalty - miss_penalty)
         # Enforce strict (0, 1) range
         clamped = 0.01 + 0.98 * base_score
-        return round(float(clamped), 6)
+        rounded = round(float(clamped), 2)
+        # Ensure no rounding to boundaries (0.0 or 1.0)
+        return max(0.01, min(rounded, 0.99))
 
 
     def passed(self) -> bool: