diff --git a/iit/model_pairs/base_model_pair.py b/iit/model_pairs/base_model_pair.py
index 78434b5..f61f3eb 100644
--- a/iit/model_pairs/base_model_pair.py
+++ b/iit/model_pairs/base_model_pair.py
@@ -374,12 +374,17 @@ def _check_early_stop_condition(self, test_metrics: MetricStoreCollection) -> bo
         """
         Returns True if all types of accuracy metrics reach 100%
         """
+        assert "early_stop_accuracy_threshold" in self.training_args, ValueError(
+            "early_stop_accuracy_threshold not found in training_args"
+        )
+        early_stop_accuracy_threshold = float(self.training_args["early_stop_accuracy_threshold"])
+
         got_accuracy_metric = False
         for metric in test_metrics:
             if metric.type == MetricType.ACCURACY:
                 got_accuracy_metric = True
                 val = metric.get_value()
-                if isinstance(val, float) and val < 99.5:
+                if isinstance(val, float) and val < early_stop_accuracy_threshold:
                     return False
         if not got_accuracy_metric:
             raise ValueError("No accuracy metric found in test_metrics!")
diff --git a/iit/model_pairs/iit_model_pair.py b/iit/model_pairs/iit_model_pair.py
index 0e0af62..a97fd4e 100644
--- a/iit/model_pairs/iit_model_pair.py
+++ b/iit/model_pairs/iit_model_pair.py
@@ -28,6 +28,7 @@ def __init__(
             "batch_size": 256,
             "num_workers": 0,
             "early_stop": True,
+            "early_stop_accuracy_threshold": 99.5,
             "lr_scheduler": None,
             "scheduler_val_metric": ["val/accuracy", "val/IIA"],
             "scheduler_mode": "max",