Improved exception error logging

restrictedpython_code_eval.py

@@ -349,15 +349,22 @@ def _check_correctness(check_program, timeout, task_id, completion_id,
 
 
     if not result:
-        result.append("Result evaluates to False (empty?)")
+        result.append("Result evaluates to False (probably timed out)")
 
-    return dict(
+    out_dict = dict(
         task_id=task_id,
         passed=result[0] == "passed",
         result=result[0],
         completion_id=completion_id,
     )
 
+    if 'failed' in result[0] and len(result) > 1:
+        exc = result[1]
+        out_dict["exception_type"] = type(exc).__name__
+        out_dict["exception_description"] = str(exc)
+
+    return out_dict
+
 
 class AllowListImporter:
     def __init__(self, allowed_imports: List[str]):
@@ -500,6 +507,7 @@ def _unsafe_execute(check_program, result, timeout,
                 with time_limit(timeout):
                     byte_code = compile_restricted(check_program, filename="<model output>", mode="exec", policy=policy_class)
                     exec(byte_code, exec_globals, additional_locals)
+            
             result.append("passed")
         except EOFError:
             result.append("EOF error")
@@ -507,6 +515,7 @@ def _unsafe_execute(check_program, result, timeout,
             result.append("timed out")
         except BaseException as e:
             result.append(f"failed ({type(e)}): {str(e)}")
+            result.append(e)
 
         # Needed for cleaning up.
         shutil.rmtree = rmtree