diff --git a/python/sglang/srt/model_executor/model_runner.py b/python/sglang/srt/model_executor/model_runner.py
index abc308302b7a..1022d471e675 100644
--- a/python/sglang/srt/model_executor/model_runner.py
+++ b/python/sglang/srt/model_executor/model_runner.py
@@ -1357,9 +1357,13 @@ def load_model(self):
             f"mem usage={self.weight_load_mem_usage:.2f} GB."
         )
         if self.server_args.debug_tensor_dump_output_folder is not None:
+            dump_folder = self.server_args.debug_tensor_dump_output_folder
+            if self.spec_algorithm.is_eagle():
+                role = "draft" if self.is_draft_worker else "target"
+                dump_folder = os.path.join(dump_folder, role)
             register_forward_hook_for_model(
                 self.model,
-                self.server_args.debug_tensor_dump_output_folder,
+                dump_folder,
                 self.server_args.debug_tensor_dump_layers,
                 self.tp_size,
                 self.tp_rank,
diff --git a/python/sglang/srt/server_args.py b/python/sglang/srt/server_args.py
index 22c676ea5192..dffe0532b26f 100644
--- a/python/sglang/srt/server_args.py
+++ b/python/sglang/srt/server_args.py
@@ -6122,7 +6122,11 @@ def add_cli_args(parser: argparse.ArgumentParser):
             "--debug-tensor-dump-output-folder",
             type=str,
             default=ServerArgs.debug_tensor_dump_output_folder,
-            help="The output folder for dumping tensors.",
+            help=(
+                "The output folder for dumping tensors. "
+                "In Eagle mode, tensor outputs from draft and target models "
+                "are stored in separate subdirectories ('draft' and 'target')."
+            ),
         )
         parser.add_argument(
             "--debug-tensor-dump-layers",