diff --git a/lm_eval/models/utils.py b/lm_eval/models/utils.py
index 89281708c8..38c74359ff 100644
--- a/lm_eval/models/utils.py
+++ b/lm_eval/models/utils.py
@@ -750,6 +750,9 @@ def segmented_tok_encode(string: SegmentedString, tokenizer: PreTrainedTokenizer
         is a list of segment labels.
 
     """
+    if type(string)!=SegmentedString:
+        raise ValueError(f"Input must be of type SegmentedString (found type {str(type(string))}).\n"
+                         f"Do not use smart truncation strategy for language modeling tasks.")
     assert type(string) == SegmentedString, "string must be a SegmentedString"
     encoding = tokenizer(
         string,