diff --git a/examples/scripts/dpo_online.py b/examples/scripts/dpo_online.py
index 58d7c4a2c0..b38e0bc95b 100644
--- a/examples/scripts/dpo_online.py
+++ b/examples/scripts/dpo_online.py
@@ -65,7 +65,7 @@
 if __name__ == "__main__":
     parser = TrlParser((ScriptArguments, OnlineDPOConfig, ModelConfig))
     script_args, training_args, model_config = parser.parse_args_and_config()
-    script_args.gradient_checkpointing_kwargs = {"use_reentrant": True}
+    training_args.gradient_checkpointing_kwargs = {"use_reentrant": True}
 
     torch_dtype = (
         model_config.torch_dtype
diff --git a/examples/scripts/nash_md.py b/examples/scripts/nash_md.py
index b9492e6042..d4915e0347 100644
--- a/examples/scripts/nash_md.py
+++ b/examples/scripts/nash_md.py
@@ -70,7 +70,7 @@
 if __name__ == "__main__":
     parser = TrlParser((ScriptArguments, NashMDConfig, ModelConfig))
     script_args, training_args, model_config = parser.parse_args_and_config()
-    script_args.gradient_checkpointing_kwargs = {"use_reentrant": True}
+    training_args.gradient_checkpointing_kwargs = {"use_reentrant": True}
 
     torch_dtype = (
         model_config.torch_dtype
diff --git a/examples/scripts/xpo.py b/examples/scripts/xpo.py
index 1c465d90e6..d2d3cb05f8 100644
--- a/examples/scripts/xpo.py
+++ b/examples/scripts/xpo.py
@@ -55,7 +55,7 @@
 if __name__ == "__main__":
     parser = TrlParser((ScriptArguments, XPOConfig, ModelConfig))
     script_args, training_args, model_config = parser.parse_args_and_config()
-    script_args.gradient_checkpointing_kwargs = {"use_reentrant": True}
+    training_args.gradient_checkpointing_kwargs = {"use_reentrant": True}
 
     torch_dtype = (
         model_config.torch_dtype