diff --git a/optimum/intel/ipex/modeling_base.py b/optimum/intel/ipex/modeling_base.py
index 00fe3de115..a125194250 100644
--- a/optimum/intel/ipex/modeling_base.py
+++ b/optimum/intel/ipex/modeling_base.py
@@ -506,12 +506,6 @@ def forward(
             "attention_mask": attention_mask,
         }
 
-        if "position_ids" in self.input_names and position_ids is None:
-            position_ids = attention_mask.long().cumsum(-1) - 1
-            position_ids.masked_fill_(attention_mask == 0, 1)
-            if past_key_values:
-                position_ids = position_ids[:, -1].unsqueeze(-1)
-
         if "position_ids" in self.input_names or not self.input_names:
             inputs["position_ids"] = position_ids
 
diff --git a/tests/ipex/test_modeling.py b/tests/ipex/test_modeling.py
index 68119287d8..c46ce1cdc2 100644
--- a/tests/ipex/test_modeling.py
+++ b/tests/ipex/test_modeling.py
@@ -32,7 +32,6 @@
     set_seed,
 )
 
-from optimum.exporters.onnx import MODEL_TYPES_REQUIRING_POSITION_IDS
 from optimum.intel import (
     IPEXModel,
     IPEXModelForAudioClassification,
@@ -236,11 +235,8 @@ def test_compare_to_transformers(self, model_arch):
             return_tensors="pt",
             return_token_type_ids=False if model_arch in ("llama", "llama2") else None,
         )
-        position_ids = None
-        if model_arch.replace("_", "-") in MODEL_TYPES_REQUIRING_POSITION_IDS:
-            input_shape = tokens["input_ids"].shape
-            position_ids = torch.arange(0, input_shape[-1], dtype=torch.long).unsqueeze(0).view(-1, input_shape[-1])
-        outputs = ipex_model(**tokens, position_ids=position_ids)
+        inputs = ipex_model.prepare_inputs_for_generation(**tokens)
+        outputs = ipex_model(**inputs)
 
         self.assertIsInstance(outputs.logits, torch.Tensor)
         self.assertIsInstance(outputs.past_key_values, (tuple, list))
@@ -263,6 +259,22 @@ def test_pipeline(self, model_arch):
         self.assertEqual(pipe.device, model.device)
         self.assertTrue(all("This is a sample" in item["generated_text"] for item in outputs))
 
+    @parameterized.expand(SUPPORTED_ARCHITECTURES)
+    def test_assisted_decoding(self, model_arch):
+        model_id = MODEL_NAMES[model_arch]
+        tokenizer = AutoTokenizer.from_pretrained(model_id)
+        ipex_model = IPEXModelForCausalLM.from_pretrained(model_id, export=True)
+        transformers_model = AutoModelForCausalLM.from_pretrained(model_id)
+        tokens = tokenizer("This is a sample input", return_tensors="pt")
+        ipex_output = ipex_model.generate(**tokens, do_sample=False)
+        ipex_output_assisted = ipex_model.generate(**tokens, do_sample=False, assistant_model=transformers_model)
+        transformers_output = transformers_model.generate(**tokens, do_sample=False)
+        transformers_output_assisted = transformers_model.generate(
+            **tokens, do_sample=False, assistant_model=ipex_model
+        )
+        self.assertTrue(torch.equal(ipex_output, ipex_output_assisted))
+        self.assertTrue(torch.equal(transformers_output, transformers_output_assisted))
+
     @parameterized.expand(
         grid_parameters(
             {