diff --git a/langfuse/_client/datasets.py b/langfuse/_client/datasets.py
index 97b2e497d..ba58af7b8 100644
--- a/langfuse/_client/datasets.py
+++ b/langfuse/_client/datasets.py
@@ -228,7 +228,7 @@ def quality_evaluator(*, input, output, expected_output=None, metadata=None, **k
 
             def content_diversity(*, item_results, **kwargs):
                 # Analyze diversity across all generated content
-                all_outputs = [result["output"] for result in item_results]
+                all_outputs = [result.output for result in item_results]
                 unique_words = set()
                 total_words = 0