graders result clampped between 0 and 1

environment.py

@@ -279,17 +279,18 @@ class DataCleaningEnv:
             component_scores[k] * w
             for k, w in weights.items()
         ) - penalty
-        # Clamp to strictly (0, 1) excluding endpoints for grader compliance
-        total = max(0.0001, min(0.9999, total))
+        # Clamp ALL scores strictly to (0, 1) — grader rejects 0.0 and 1.0
+        def _sc(v):
+            return round(max(0.0001, min(0.9999, float(v))), 4)
 
         return Reward(
-            total=round(total, 4),
-            duplicate_score=round(dup_score, 4),
-            missing_score=round(missing_score, 4),
-            dtype_score=round(dtype_score, 4),
-            outlier_score=round(outlier_score, 4),
-            schema_score=round(schema_score, 4),
-            penalty=round(penalty, 4)
+            total=_sc(total),
+            duplicate_score=_sc(dup_score),
+            missing_score=_sc(missing_score),
+            dtype_score=_sc(dtype_score),
+            outlier_score=_sc(outlier_score),
+            schema_score=_sc(schema_score),
+            penalty=round(max(0.0, min(0.9999, float(penalty))), 4)
         )
 
     # ─────────────────────────────────────────

main.py

@@ -1,4 +1,3 @@
-
 import sys
 import os
 sys.path.insert(0, os.path.dirname(__file__))
@@ -280,7 +279,7 @@ def validate():
             )
             step_result = env.step(action)
             assert step_result.observation is not None
-            assert 0.0 <= step_result.reward.total <= 1.0
+            assert 0.0 < step_result.reward.total < 1.0  # strict bounds required by grader
 
             # Test state
             state_result = env.state()
@@ -327,7 +326,7 @@ def submit_score(entry: Dict[str, Any]):
     leaderboard_data.append({
         "model_name": entry["model_name"],
         "task_id":    entry["task_id"],
-        "score":      round(float(entry["score"]), 4),
+        "score":      round(max(0.0001, min(0.9999, float(entry["score"]))), 4),
         "steps":      entry.get("steps", 0),
         "timestamp":  __import__("datetime").datetime.utcnow().isoformat()
     })
@@ -365,7 +364,7 @@ def get_leaderboard():
         ranked.append({
             "model_name":   model,
             "scores":       scores,
-            "avg_score":    round(avg, 4)
+            "avg_score":    round(max(0.0001, min(0.9999, avg)), 4)
         })
 
     ranked.sort(key=lambda x: x["avg_score"], reverse=True)
@@ -375,4 +374,4 @@ def get_leaderboard():
     return {
         "leaderboard":       ranked,
         "total_submissions": len(leaderboard_data)
-    }
+    }