diff --git a/sagemaker-train/tests/integ/train/test_mtrl_evaluator.py b/sagemaker-train/tests/integ/train/test_mtrl_evaluator.py
index b79a2763dd..9d493a0508 100644
--- a/sagemaker-train/tests/integ/train/test_mtrl_evaluator.py
+++ b/sagemaker-train/tests/integ/train/test_mtrl_evaluator.py
@@ -40,6 +40,10 @@ def _get_test_config():
     """Build test configuration lazily (only when tests actually run)."""
     boto_session = boto3.Session(region_name=_REGION)
     account_id = boto_session.client("sts").get_caller_identity()["Account"]
+    from sagemaker.core.helper.session_helper import Session
+    from sagemaker.train.defaults import TrainDefaults
+    sagemaker_session = Session(boto_session=boto_session)
+    role_arn = TrainDefaults.get_role(role=None, sagemaker_session=sagemaker_session)
     return {
         "base_model": "mock-oss-test",
         "agent_arn": f"arn:aws:bedrock-agentcore:{_REGION}:{account_id}:runtime/sagemaker_rft_prod_gsm8k_streaming-Yk6O377mUS",
@@ -47,7 +51,7 @@ def _get_test_config():
         "s3_output_path": f"s3://sagemaker-{_REGION}-{account_id}/model-evaluation/output-artifacts/",
         "mlflow_resource_arn": f"arn:aws:sagemaker:{_REGION}:{account_id}:mlflow-app/app-TTAUWUNMUHH6",
         "model_package_group": f"arn:aws:sagemaker:{_REGION}:{account_id}:model-package-group/openai-reasoning-gpt-oss-20b-mtrl-mpg",
-        "role": f"arn:aws:iam::{account_id}:role/Admin",
+        "role": role_arn,
         "region": _REGION,
         "account_id": account_id,
     }
@@ -172,7 +176,6 @@ def test_bedrock_agent_config_fields(self, mtrl_trainer, test_config):
             dataset=test_config["dataset"],
             s3_output_path=f'{test_config["s3_output_path"]}integ-fields-bedrock/',
             mlflow_resource_arn=test_config["mlflow_resource_arn"],
-            role=test_config["role"],
             region=test_config["region"],
             agent_config=test_config["agent_arn"],
             agent_qualifier="PROD",
@@ -182,8 +185,7 @@ def test_bedrock_agent_config_fields(self, mtrl_trainer, test_config):
         evaluator._resolve_agent_arn()
 
         ctx = evaluator._build_template_context(
-            aws_context={"region": test_config["region"], "account_id": test_config["account_id"],
-                         "role_arn": test_config["role"]},
+            aws_context={"region": test_config["region"], "account_id": test_config["account_id"], "role_arn": test_config["role"]},
             artifacts={},
             model_package_group_arn=test_config["model_package_group"],
         )
@@ -208,7 +210,6 @@ def test_lambda_agent_config_fields(self, mtrl_trainer, test_config):
             dataset=test_config["dataset"],
             s3_output_path=f'{test_config["s3_output_path"]}integ-fields-lambda/',
             mlflow_resource_arn=test_config["mlflow_resource_arn"],
-            role=test_config["role"],
             region=test_config["region"],
             agent_config=lambda_arn,
         )
@@ -217,8 +218,7 @@ def test_lambda_agent_config_fields(self, mtrl_trainer, test_config):
         evaluator._resolve_agent_arn()
 
         ctx = evaluator._build_template_context(
-            aws_context={"region": test_config["region"], "account_id": test_config["account_id"],
-                         "role_arn": test_config["role"]},
+            aws_context={"region": test_config["region"], "account_id": test_config["account_id"], "role_arn": test_config["role"]},
             artifacts={},
             model_package_group_arn=test_config["model_package_group"],
         )
@@ -239,7 +239,6 @@ def test_model_package_config_fields(self, mtrl_trainer, test_config):
             dataset=test_config["dataset"],
             s3_output_path=f'{test_config["s3_output_path"]}integ-fields-mpc/',
             mlflow_resource_arn=test_config["mlflow_resource_arn"],
-            role=test_config["role"],
             region=test_config["region"],
             agent_config=test_config["agent_arn"],
         )
@@ -248,8 +247,7 @@ def test_model_package_config_fields(self, mtrl_trainer, test_config):
         evaluator._resolve_agent_arn()
 
         ctx = evaluator._build_template_context(
-            aws_context={"region": test_config["region"], "account_id": test_config["account_id"],
-                         "role_arn": test_config["role"]},
+            aws_context={"region": test_config["region"], "account_id": test_config["account_id"], "role_arn": test_config["role"]},
             artifacts={},
             model_package_group_arn=test_config["model_package_group"],
         )
@@ -280,7 +278,6 @@ def test_evaluator_construction_with_trainer(self, mtrl_trainer, test_config):
             dataset=test_config["dataset"],
             s3_output_path=f'{test_config["s3_output_path"]}integ-construct/',
             mlflow_resource_arn=test_config["mlflow_resource_arn"],
-            role=test_config["role"],
             region=test_config["region"],
             agent_config=test_config["agent_arn"],
         )
@@ -298,7 +295,6 @@ def test_evaluator_construction_with_base_model(self, test_config):
             s3_output_path=f'{test_config["s3_output_path"]}integ-base/',
             agent_config=test_config["agent_arn"],
             mlflow_resource_arn=test_config["mlflow_resource_arn"],
-            role=test_config["role"],
             region=test_config["region"],
         )
 
diff --git a/sagemaker-train/tests/integ/train/test_mtrl_evaluator_3p_agent.py b/sagemaker-train/tests/integ/train/test_mtrl_evaluator_3p_agent.py
index 2d16d4318f..1a9d1402c3 100644
--- a/sagemaker-train/tests/integ/train/test_mtrl_evaluator_3p_agent.py
+++ b/sagemaker-train/tests/integ/train/test_mtrl_evaluator_3p_agent.py
@@ -161,10 +161,6 @@ def _get_3p_test_config():
             "MTRL_3P_MLFLOW_ARN",
             f"arn:aws:sagemaker:{_REGION}:{account_id}:mlflow-app/app-TTAUWUNMUHH6",
         ),
-        "role": os.environ.get(
-            "MTRL_3P_ROLE",
-            f"arn:aws:iam::{account_id}:role/Admin",
-        ),
         "region": os.environ.get("MTRL_3P_REGION", _REGION),
         "account_id": account_id,
     }
@@ -258,7 +254,6 @@ def test_evaluate_with_lambda_agent_wait_for_completion(self, lambda_agent_arn,
             agent_config=lambda_agent_arn,
             s3_output_path=f'{test_config["s3_output_path"]}lambda-e2e/',
             mlflow_resource_arn=test_config["mlflow_resource_arn"],
-            role=test_config["role"],
             region=test_config["region"],
             accept_eula=True,
         )
@@ -304,7 +299,6 @@ def test_evaluate_base_model_with_agent_lambda_object(self, lambda_agent_arn, te
             agent_config=agent,
             s3_output_path=f'{test_config["s3_output_path"]}lambda-object/',
             mlflow_resource_arn=test_config["mlflow_resource_arn"],
-            role=test_config["role"],
             region=test_config["region"],
             accept_eula=True,
         )
@@ -332,7 +326,6 @@ def test_evaluate_with_attached_trainer(self, lambda_agent_arn, test_config):
             agent_config=lambda_agent_arn,
             s3_output_path=f'{test_config["s3_output_path"]}attached-trainer/',
             mlflow_resource_arn=test_config["mlflow_resource_arn"],
-            role=test_config["role"],
             region=test_config["region"],
             accept_eula=True,
         )
diff --git a/sagemaker-train/tests/integ/train/test_mtrl_trainer_integration.py b/sagemaker-train/tests/integ/train/test_mtrl_trainer_integration.py
index 880900f4b8..9f30c3c86b 100644
--- a/sagemaker-train/tests/integ/train/test_mtrl_trainer_integration.py
+++ b/sagemaker-train/tests/integ/train/test_mtrl_trainer_integration.py
@@ -44,6 +44,7 @@ def _get_account_id():
     boto_session = boto3.Session(region_name=_REGION)
     return boto_session.client("sts").get_caller_identity()["Account"]
 
+
 # ============================================================
 # Per-account resource configuration
 # ============================================================
@@ -60,7 +61,6 @@ def _get_account_id():
         "s3_output_path": "s3://sagemaker-us-west-2-729646638167/mtrl-integ/eval-output/",
         "mlflow_resource_arn": "arn:aws:sagemaker:us-west-2:729646638167:mlflow-app/app-TTAUWUNMUHH6",
         "model_package_group": "arn:aws:sagemaker:us-west-2:729646638167:model-package-group/openai-reasoning-gpt-oss-20b-mtrl-mpg",
-        "role": "arn:aws:iam::729646638167:role/Admin",
     },
     # PREPROD — Staging account (391266019386)
     "391266019386": {
@@ -72,7 +72,6 @@ def _get_account_id():
         "s3_output_path": "s3://sagemaker-us-west-2-391266019386/mtrl-integ/eval-output/",
         "mlflow_resource_arn": "arn:aws:sagemaker:us-west-2:391266019386:mlflow-app/app-P3FRQFRQTNGI",
         "model_package_group": "arn:aws:sagemaker:us-west-2:391266019386:model-package-group/mtrl-integ-gpt-oss-agentcore",
-        "role": "arn:aws:iam::391266019386:role/Admin",
     },
     # BETA — Dev/test account (742774200982)
     "742774200982": {
@@ -84,7 +83,6 @@ def _get_account_id():
         "s3_output_path": "s3://sagemaker-us-west-2-742774200982/mtrl-integ/eval-output/",
         "mlflow_resource_arn": "arn:aws:sagemaker:us-west-2:742774200982:mlflow-app/app-6ZU5TXXH2GUX",
         "model_package_group": "arn:aws:sagemaker:us-west-2:742774200982:model-package-group/openai-reasoning-gpt-oss-20b-mtrl-mpg",
-        "role": "arn:aws:iam::742774200982:role/Admin",
     },
 }
 
@@ -133,7 +131,6 @@ def attached_trainer(config):
         output_model_package_group=config["model_package_group"],
         mlflow_app_arn=config["mlflow_resource_arn"],
         s3_output_path=config["s3_output_path"],
-        role=config["role"],
         accept_eula=True,
     )
     trainer._latest_job = job
@@ -165,7 +162,6 @@ def test_evaluate_finetuned_model(self, attached_trainer, config):
             dataset=config["dataset"],
             s3_output_path=f'{config["s3_output_path"]}finetuned/',
             mlflow_resource_arn=config["mlflow_resource_arn"],
-            role=config["role"],
             region=_REGION,
         )
 
@@ -195,7 +191,6 @@ def test_evaluate_base_model(self, config):
             agent_config=config["agent_core_arn"],
             s3_output_path=f'{config["s3_output_path"]}basemodel/',
             mlflow_resource_arn=config["mlflow_resource_arn"],
-            role=config["role"],
             region=_REGION,
         )
 
@@ -225,7 +220,6 @@ def test_evaluate_comparison(self, attached_trainer, config):
             dataset=config["dataset"],
             s3_output_path=f'{config["s3_output_path"]}comparison/',
             mlflow_resource_arn=config["mlflow_resource_arn"],
-            role=config["role"],
             region=_REGION,
             evaluate_base_model=True,
         )
diff --git a/sagemaker-train/tests/integ/train/test_multi_turn_rl_trainer_integration.py b/sagemaker-train/tests/integ/train/test_multi_turn_rl_trainer_integration.py
index 71d41128e7..be47203dcb 100644
--- a/sagemaker-train/tests/integ/train/test_multi_turn_rl_trainer_integration.py
+++ b/sagemaker-train/tests/integ/train/test_multi_turn_rl_trainer_integration.py
@@ -56,7 +56,6 @@ def test_resources():
     """Resolve account-specific resource ARNs lazily."""
     account_id = _get_account_id()
     return {
-        "role_arn": f"arn:aws:iam::{account_id}:role/Admin",
         "mlflow_arn": f"arn:aws:sagemaker:{_REGION}:{account_id}:mlflow-app/app-TTAUWUNMUHH6",
         "s3_input_path": f"s3://sagemaker-rft-{account_id}/prompts/gsm8k_small/prompts.parquet",
         "s3_output_path": f"s3://sagemaker-{_REGION}-{account_id}/model-evaluation/mtrl-trainer-integ/",
@@ -77,7 +76,6 @@ def test_train_and_wait(self, sagemaker_session, test_resources):
             training_dataset=test_resources["s3_input_path"],
             mlflow_app_arn=test_resources["mlflow_arn"],
             s3_output_path=test_resources["s3_output_path"],
-            role=test_resources["role_arn"],
             accept_eula=True,
             sagemaker_session=sagemaker_session,
         )
@@ -101,7 +99,6 @@ def test_train_and_stop(self, sagemaker_session, test_resources):
             agent_env=AGENT_RUNTIME_ID,
             training_dataset=test_resources["s3_input_path"],
             mlflow_app_arn=test_resources["mlflow_arn"],
-            role=test_resources["role_arn"],
             accept_eula=True,
             sagemaker_session=sagemaker_session,
         )
@@ -132,7 +129,6 @@ def test_train_with_lambda_arn(self, sagemaker_session, test_resources):
             mlflow_app_arn=test_resources["mlflow_arn"],
             s3_output_path=test_resources["s3_output_path"],
             accept_eula=True,
-            role=test_resources["role_arn"],
             sagemaker_session=sagemaker_session,
         )
         trainer.hyperparameters.global_batch_size = 32