diff --git a/vllm/model_executor/layers/pooler/tokwise/methods.py b/vllm/model_executor/layers/pooler/tokwise/methods.py
index baa9d4075..f242d215d 100644
--- a/vllm/model_executor/layers/pooler/tokwise/methods.py
+++ b/vllm/model_executor/layers/pooler/tokwise/methods.py
@@ -47,10 +47,13 @@ class AllPool(TokenPoolingMethod):
         pooling_metadata: PoolingMetadata,
     ) -> list[TokenPoolingMethodOutputItem]:
         pooling_cursor = pooling_metadata.get_pooling_cursor()
-        hidden_states_all = hidden_states.split(
-            pooling_cursor.num_scheduled_tokens_cpu.tolist()
-        )
-        hidden_states_lst = [hidden_states_all[i] for i in pooling_cursor.index]
+        hidden_states_lst = [
+            hidden_states[first : last + 1]
+            for first, last in zip(
+                pooling_cursor.first_token_indices_gpu.tolist(),
+                pooling_cursor.last_token_indices_gpu.tolist(),
+            )
+        ]
 
         if not self.enable_chunked_prefill:
             return hidden_states_lst