Upload folder using huggingface_hub

README.md

@@ -39,7 +39,7 @@ step(violation_report) → Correct? +1.0 | False positive? -0.3 | Duplicate? -0.
 step(DONE) → Episode ends with completeness bonus
 ```
 
-## Tasks (4 difficulty levels)
+## Tasks (5 difficulty levels)
 
 | Task | Difficulty | Violations | Max Steps | What the Agent Must Find |
 |------|-----------|------------|-----------|--------------------------|
@@ -47,6 +47,7 @@ step(DONE) → Episode ends with completeness bonus
 | `validate_nested_objects` | Medium | 7 | 15 | Violations inside nested objects and arrays — requires traversing deep structures |
 | `detect_breaking_changes` | Hard | 9 | 20 | Breaking changes between two API spec versions — type changes, removed fields, narrowed enums, new requirements |
 | `validate_response_schema` | Expert | 10 | 25 | Subtle format errors in an API response: invalid date formats, pattern mismatches, out-of-range numerics, and bad enum values scattered across nested objects and arrays |
+| `validate_cross_field_constraints` | Expert | 7 | 18 | Cross-field arithmetic and date ordering constraints on Invoice API — due_date must be after invoice_date, line totals, subtotal sum, tax calculation, discount rules for trial accounts |
 
 ### Randomised Episode Generation
 
@@ -149,6 +150,8 @@ openenv validate
 | `find_type_mismatches` | Qwen2.5-72B-Instruct | ~0.75 | 5–7 |
 | `validate_nested_objects` | Qwen2.5-72B-Instruct | ~0.57 | 8–12 |
 | `detect_breaking_changes` | Qwen2.5-72B-Instruct | ~0.44 | 12–18 |
+| `validate_response_schema` | Qwen2.5-72B-Instruct | ~0.40 | 15–22 |
+| `validate_cross_field_constraints` | Qwen2.5-72B-Instruct | ~0.43 | 10–16 |
 
 *Scores are approximate and may vary with temperature/sampling.*
 

server/rewards.py

@@ -161,8 +161,10 @@ def compute_step_reward(
 def compute_episode_score(correct_count: int, total_violations: int) -> float:
     """Compute the final normalised score for the episode.
 
-    Returns a float in ``[0.0, 1.0]``.
+    Returns a float strictly in ``(0.0, 1.0)`` — endpoints excluded — as
+    required by the OpenEnv evaluation pipeline.
     """
     if total_violations == 0:
-        return 1.0
-    return round(correct_count / total_violations, 4)
+        return 0.5
+    raw = correct_count / total_violations
+    return round(max(0.0001, min(0.9999, raw)), 4)