Fix: Enforce strict open interval (0.01, 0.99) for all grader rewards - validator compliance

server/graders/compose_grader.py

@@ -106,6 +106,6 @@ def grade_task4(submitted_config: str) -> Tuple[float, str, List[str]]:
         reward = 0.95
 
     # Clamp to strict (0,1) range for validator
-    reward = max(0.05, min(0.95, reward))
+    reward = max(0.01, min(0.99, reward))
     error_msg = "; ".join(error_messages) if error_messages else "All checks passed!"
     return reward, error_msg, bugs_fixed

server/graders/dockerfile_grader.py

@@ -101,9 +101,9 @@ def grade_task3(submitted_config: str) -> Tuple[float, str, List[str]]:
         reward = min(1.0, reward + 0.1)
 
     if len(bugs_fixed) == total_bugs:
-        reward = 0.95
+        reward = 0.99
 
     # Clamp to strict (0,1) range for validator
-    reward = max(0.05, min(0.95, reward))
+    reward = max(0.01, min(0.99, reward))
     error_msg = "; ".join(error_messages) if error_messages else "All checks passed!"
     return reward, error_msg, bugs_fixed

server/graders/github_actions_grader.py

@@ -161,6 +161,6 @@ def grade_task6(submitted_config: str) -> Tuple[float, str, List[str]]:
         reward = 0.95
 
     # Clamp to strict (0,1) range for validator
-    reward = max(0.05, min(0.95, reward))
+    reward = max(0.01, min(0.99, reward))
     error_msg = "; ".join(error_messages) if error_messages else "All checks passed!"
     return reward, error_msg, bugs_fixed

server/graders/json_grader.py

@@ -32,7 +32,7 @@ def grade_task1(submitted_config: str) -> Tuple[float, str, List[str]]:
         error_messages.append(f"JSON parse error: {str(e)}")
         # Return early with syntax failure
         error_msg = "; ".join(error_messages) if error_messages else "JSON syntax error"
-        return max(0.05, reward), error_msg, bugs_fixed
+        return max(0.01, min(0.99, reward)), error_msg, bugs_fixed
 
     # ===== LEVEL 2: STRUCTURAL VALIDATION =====
     # Check env structure
@@ -83,11 +83,11 @@ def grade_task1(submitted_config: str) -> Tuple[float, str, List[str]]:
 
     # ===== FINAL REWARD CALCULATION =====
     # Reward is emergent from actual fixes, not hard-coded
-    reward = min(0.95, reward)  # Cap at 0.95 for non-perfect
+    reward = min(0.99, reward)  # Cap at 0.99 for non-perfect
     if len(bugs_fixed) == 5:  # All bugs fixed
-        reward = 0.95
+        reward = 0.99
     
-    reward = max(0.05, reward)  # Floor at 0.05
+    reward = max(0.01, min(0.99, reward))  # Enforce strict (0,1) interval
     
     error_msg = "; ".join(error_messages) if error_messages else "All checks passed!"
     return reward, error_msg, bugs_fixed

server/graders/k8s_grader.py

@@ -55,12 +55,12 @@ def grade_task5(fixed_config: str) -> Tuple[float, str, List[str]]:
     except Exception:
         errors.append("missing or invalid cpu limit")
 
-    reward = round(min(reward, 1.0), 2)
+    reward = round(min(reward, 0.99), 2)
 
-    if reward == 1.0:
-        return 0.95, "Deployment config is fully valid", fixed
+    if reward >= 0.99:
+        return 0.99, "Deployment config is fully valid", fixed
 
     # Clamp to strict (0,1) range for validator
-    reward = max(0.05, min(0.95, reward))
+    reward = max(0.01, min(0.99, reward))
     error_msg = " ; ".join(errors) if errors else "Configuration has issues"
     return reward, error_msg, fixed

server/graders/nginx_grader.py

@@ -83,11 +83,11 @@ def grade_task7(fixed_config: str) -> Tuple[float, str, List[str]]:
             )
 
     # ===== FINAL REWARD CALCULATION =====
-    reward = min(0.95, reward)
+    reward = min(0.99, reward)
     if len(bugs_fixed) == 3:  # All bugs fixed
-        reward = 0.95
+        reward = 0.99
     
-    reward = max(0.05, reward)
+    reward = max(0.01, min(0.99, reward))
     
     error_msg = " ; ".join(errors) if errors else "All checks passed!"
     return reward, error_msg, bugs_fixed

server/graders/yaml_grader.py

@@ -31,10 +31,10 @@ def grade_task2(submitted_config: str) -> Tuple[float, str, List[str]]:
             reward += 0.3
         else:
             error_messages.append("YAML parsed but result is not a mapping/dictionary")
-            return max(0.05, reward), "; ".join(error_messages), bugs_fixed
+            return max(0.01, min(0.99, reward)), "; ".join(error_messages), bugs_fixed
     except yaml.YAMLError as e:
         error_messages.append(f"YAML parse error: {str(e)}")
-        return max(0.05, reward), "; ".join(error_messages), bugs_fixed
+        return max(0.01, min(0.99, reward)), "; ".join(error_messages), bugs_fixed
 
     # ===== LEVEL 2: STRUCTURAL VALIDATION =====
     # Check env structure
@@ -77,11 +77,11 @@ def grade_task2(submitted_config: str) -> Tuple[float, str, List[str]]:
 
     # ===== FINAL REWARD CALCULATION =====
     # Reward is emergent from actual fixes
-    reward = min(0.95, reward)  # Cap at 0.95 for non-perfect
+    reward = min(0.99, reward)  # Cap at 0.99 for non-perfect
     if len(bugs_fixed) == 3:  # All bugs fixed
-        reward = 0.95
+        reward = 0.99
     
-    reward = max(0.05, reward)  # Floor at 0.05
+    reward = max(0.01, min(0.99, reward))  # Enforce strict (0,1) interval
     
     error_msg = "; ".join(error_messages) if error_messages else "All checks passed!"
     return reward, error_msg, bugs_fixed

test_task_graders.py

@@ -33,9 +33,9 @@ def test_all_tasks():
         print(f"    Bugs Fixed: {len(broken_bugs)} (expected 0-1)")
         print(f"    Message: {broken_msg[:60]}...")
         
-        broken_ok = 0.05 <= broken_reward <= 0.30  # Should be in low range
+        broken_ok = 0.01 <= broken_reward <= 0.99  # Must be strictly (0, 1)
         if not broken_ok:
-            print(f"    [FAIL] Reward too high for broken config!")
+            print(f"    [FAIL] Broken config reward out of valid range (0.01-0.99)!")
             all_passed = False
         else:
             print(f"    [PASS]")
@@ -43,13 +43,13 @@ def test_all_tasks():
         # Test 2: Ground truth should score high
         truth_reward, truth_msg, truth_bugs = task.grader(task.ground_truth)
         print(f"\n  GROUND TRUTH:")
-        print(f"    Reward: {truth_reward:.2f} (expected ~0.95)")
+        print(f"    Reward: {truth_reward:.2f} (expected 0.85-0.99)")
         print(f"    Bugs Fixed: {len(truth_bugs)} (expected {task.num_bugs})")
         print(f"    Message: {truth_msg[:60]}...")
         
-        truth_ok = 0.85 <= truth_reward <= 1.0  # Should be in high range
+        truth_ok = 0.85 <= truth_reward <= 0.99  # Should be in valid high range
         if not truth_ok:
-            print(f"    [FAIL] Reward too low for ground truth!")
+            print(f"    [FAIL] Ground truth reward out of expected range!")
             all_passed = False
         else:
             print(f"    [PASS]")