graders result clampped between 0 and 1

environment.py

@@ -1,4 +1,3 @@
-
 import json
 import pandas as pd
 import numpy as np
@@ -454,7 +453,7 @@ class DataCleaningEnv:
         self.actions_taken = []  # Reset action history
 
         obs    = self._get_observation("Environment reset. Start cleaning!")
-        reward = Reward(total=0.0)
+        reward = Reward(total=0.0001)
 
         return StepResult(
             observation=obs,
@@ -468,7 +467,7 @@ class DataCleaningEnv:
             obs = self._get_observation("Episode already done. Call reset().")
             return StepResult(
                 observation=obs,
-                reward=Reward(total=0.0),
+                reward=Reward(total=0.0001),
                 done=True,
                 info={"warning": "Episode already done"}
             )
@@ -542,4 +541,4 @@ class DataCleaningEnv:
             "duplicate_count": int(self.current_df.duplicated().sum()),
             "reward_history": self.reward_history,
             "dtypes":         {c: str(t) for c, t in self.current_df.dtypes.items()}
-        }
+        }

inference.py

@@ -155,9 +155,6 @@ def run_task(
     success    = False
     last_error = None
 
-    def clamp_score(value: float) -> float:
-        return max(0.0001, min(0.9999, round(value, 4)))
-
     while not done and step < MAX_STEPS:
         step += 1
         last_error = None
@@ -196,7 +193,7 @@ def run_task(
             result  = env.step(action)
             obs     = result.observation.model_dump()
             done    = result.done
-            reward  = clamp_score(result.reward.total)
+            reward  = result.reward.total
             rewards.append(reward)
 
             # Check if operation failed
@@ -206,7 +203,7 @@ def run_task(
 
         except Exception as e:
             last_error = str(e)
-            reward = 0.0001
+            reward = 0.0
             rewards.append(reward)
             done = True
 
@@ -228,10 +225,8 @@ def run_task(
     rewards_str = ",".join([f"{r:.2f}" for r in rewards])
     print(f"[END] success={success_str} steps={step} rewards={rewards_str}")
 
-    final_score = 0.0001
-    if rewards:
-        final_score = max(0.0001, min(0.9999, round(rewards[-1], 4)))
-
+    raw_score = rewards[-1] if rewards else 0.0001
+    final_score = max(0.0001, min(0.9999, round(raw_score, 4)))
     return {
         "task_id":      task_id,
         "final_score":  final_score,
@@ -284,4 +279,4 @@ def main():
 
 
 if __name__ == "__main__":
-    main()
+    main()