diff --git a/slime/backends/megatron_utils/update_weight/hf_weight_iterator_bridge.py b/slime/backends/megatron_utils/update_weight/hf_weight_iterator_bridge.py
index edfd690482..8db37eb9b2 100644
--- a/slime/backends/megatron_utils/update_weight/hf_weight_iterator_bridge.py
+++ b/slime/backends/megatron_utils/update_weight/hf_weight_iterator_bridge.py
@@ -26,9 +26,9 @@ def _patch_bridge_expert_cache_to_cpu():
 
     _orig = GPTOSSBridge.maybe_modify_converted_hf_weight
 
-    def _patched(self, task, converted_weights_dict):
+    def _patched(self, task, converted_weights_dict, hf_state_dict=None):
         cpu_dict = {k: v.cpu() for k, v in converted_weights_dict.items()}
-        result = _orig(self, task, cpu_dict)
+        result = _orig(self, task, cpu_dict, hf_state_dict)
         # Move merged result back to GPU for CUDA IPC serialization
         return {k: v.cuda() for k, v in result.items()} if result else result