diff --git a/mindspeed_rl/utils/utils.py b/mindspeed_rl/utils/utils.py
index 857ed67b8410af0eb32a3ad9682838ce3abf3055..86197079e5e48e001f1687a328b15592f5628d19 100644
--- a/mindspeed_rl/utils/utils.py
+++ b/mindspeed_rl/utils/utils.py
@@ -306,7 +306,7 @@ class MsProbe:
     hooked_model = []
 
     @classmethod
-    def config_init(cls, msprobe_config):
+    def config_init(cls, msprobe_config, init_step=0):
         if not msprobe_config.msprobe:
             return
         cls.config = msprobe_config
@@ -322,6 +322,7 @@ class MsProbe:
         if cls.need_debugger():
             step = [f"{cls.config.step_start}-{cls.config.step_end}"]
             cls.debugger = PrecisionDebugger(task="statistics", level="L0", step=step, dump_path=cls.config.dump_path)
+            cls.debugger.set_init_step(init_step)
 
         cls.enabled = True
         print("msprobe enabled")
diff --git a/mindspeed_rl/workers/actor_hybrid_worker.py b/mindspeed_rl/workers/actor_hybrid_worker.py
index e4f5edb60f111478929a09a4add5cc98106dbdf1..3083de549d3c5d1561f740d0ea0d02d4c265b59a 100644
--- a/mindspeed_rl/workers/actor_hybrid_worker.py
+++ b/mindspeed_rl/workers/actor_hybrid_worker.py
@@ -121,7 +121,7 @@ class ActorHybridWorkerBase(BaseWorker):
         )
         self.empty_cache()
         self.actor_profiler = profiler_start(self.profiler_config, self.profiler_config.role)
-        MsProbe.config_init(self.msprobe_config)
+        MsProbe.config_init(self.msprobe_config, self.iteration + 1)
 
     def init_transfer_dock(self, td, mm_td):
         self.td = td
diff --git a/mindspeed_rl/workers/integrated_worker.py b/mindspeed_rl/workers/integrated_worker.py
index 175ee11ebb4753112b3b4096d8b4a72f6fa73ece..263a445772f7eddd2002b62045c78b56c280f230 100644
--- a/mindspeed_rl/workers/integrated_worker.py
+++ b/mindspeed_rl/workers/integrated_worker.py
@@ -111,7 +111,6 @@ class IntegratedWorker(ActorHybridWorkerBase, ReferenceWorkerBase, RewardWorkerB
             context_parallel_size=self.megatron_config.context_parallel_size,
             temperature=self.generate_config.sampling_config["temperature"]
         )
-        MsProbe.config_init(self.msprobe_config)
 
     @mstx_timer_decorator
     def compute_ref_log_prob(self):
diff --git a/tests/st/configs/test_grpo_trainer_qwen25_7b_integrated.yaml b/tests/st/configs/test_grpo_trainer_qwen25_7b_integrated.yaml
index 196eb9af9c609671c8be435b860efe27b5f95436..2776a7021e887d7ba80bc1632e724463df0401ab 100644
--- a/tests/st/configs/test_grpo_trainer_qwen25_7b_integrated.yaml
+++ b/tests/st/configs/test_grpo_trainer_qwen25_7b_integrated.yaml
@@ -146,5 +146,5 @@ msprobe_config:
   configurations_dump: true
   actor_train_dump: true
   reference_dump: true
-  step_start: 0
-  step_end: 0
\ No newline at end of file
+  step_start: 1
+  step_end: 1
\ No newline at end of file
diff --git a/tests/st/mindstudio/check_and_clean_mindstudio_output.py b/tests/st/mindstudio/check_and_clean_mindstudio_output.py
index 676dead62b8cd2a63cb8ac76724f74d1d1ff61af..8be04bca92dcb9cea25e6f54b95dbed1770d8627 100644
--- a/tests/st/mindstudio/check_and_clean_mindstudio_output.py
+++ b/tests/st/mindstudio/check_and_clean_mindstudio_output.py
@@ -61,9 +61,9 @@ def check_msprobe_output(msprobe_dir: str) -> bool:
         logger.error(f'Configurations directory not found: {os.path.join(msprobe_dir, "configurations.json")}')
         return False
     
-    if not os.path.isfile(os.path.join(msprobe_dir, "data", "responses", "step0", "rank0", "responses.json")):
+    if not os.path.isfile(os.path.join(msprobe_dir, "data", "responses", "step1", "rank0", "responses.json")):
         logger.error(f'Msprobe key data response not found: '
-                     f'{os.path.join(msprobe_dir, "data", "responses", "step0", "rank0", "responses.json")}')
+                     f'{os.path.join(msprobe_dir, "data", "responses", "step1", "rank0", "responses.json")}')
         return False
     
     if not os.path.isdir(os.path.join(msprobe_dir, "actor_update")):