Update grading logic to strict (0,1) ratio fallback

server/grader.py

@@ -1,24 +1,18 @@
 from .models import ExecutionResult, TaskInfo
 
 def calculate_reward(exec_result: ExecutionResult, task_info: TaskInfo) -> float:
-    """
-    Computes reward using the requested formula:
-    0.3 * compile_score + 0.4 * test_pass_ratio + 0.3 * efficiency_score
-    """
-    compile_score = 1.0 if exec_result.compile_success else 0.0
+    passed_tests = exec_result.test_passed
+    total_tests = exec_result.test_total
     
-    test_pass_ratio = 0.0
-    if exec_result.test_total > 0:
-        test_pass_ratio = exec_result.test_passed / exec_result.test_total
+    if total_tests > 0:
+        reward = passed_tests / total_tests
+    else:
+        reward = 0.0
+
+    # enforce valid range
+    if reward <= 0:
+        reward = 0.1
+    elif reward >= 1:
+        reward = 0.9
         
-    efficiency_score = 0.0
-    # Efficiency is only tracked if tests completely pass to prevent 1-line failure from being "efficient"
-    if exec_result.test_total > 0 and exec_result.test_passed == exec_result.test_total:
-        if exec_result.execution_time_seconds <= task_info.optimal_time_seconds:
-            efficiency_score = 1.0
-        else:
-            # Proportional efficiency dropoff
-            efficiency_score = task_info.optimal_time_seconds / max(exec_result.execution_time_seconds, 0.001)
-            
-    reward = (0.3 * compile_score) + (0.4 * test_pass_ratio) + (0.3 * efficiency_score)
-    return max(0.0, min(1.0, reward))
+    return reward