diff --git a/flaml/tune/searcher/suggestion.py b/flaml/tune/searcher/suggestion.py
index 8380fc0fda..b9bca38127 100644
--- a/flaml/tune/searcher/suggestion.py
+++ b/flaml/tune/searcher/suggestion.py
@@ -561,7 +561,15 @@ def __init__(
         self._space = space
 
         self._points_to_evaluate = points_to_evaluate or []
-        self._evaluated_rewards = evaluated_rewards
+        # rewards should be a list of floats, not a dict
+        # After Optuna > 3.5.0, there is a check for NaN in the list "any(math.isnan(x) for x in self._values)"
+        # which will raise an error when encountering a dict
+        if evaluated_rewards is not None:
+            self._evaluated_rewards = [
+                list(item.values())[0] if isinstance(item, dict) else item for item in evaluated_rewards
+            ]
+        else:
+            self._evaluated_rewards = evaluated_rewards
 
         self._study_name = "optuna"  # Fixed study name for in-memory storage
 
diff --git a/setup.py b/setup.py
index 1c6d045bfe..7bb35716bb 100644
--- a/setup.py
+++ b/setup.py
@@ -65,7 +65,7 @@
             "catboost>=0.26,<1.2; python_version<'3.11'",
             "catboost>=0.26; python_version>='3.11'",
             "rgf-python",
-            "optuna==2.8.0",
+            "optuna>=2.8.0,<=3.6.1",
             "openml",
             "statsmodels>=0.12.2",
             "psutil==5.8.0",
@@ -94,7 +94,7 @@
         ],
         "catboost": ["catboost>=0.26"],
         "blendsearch": [
-            "optuna==2.8.0",
+            "optuna>=2.8.0,<=3.6.1",
             "packaging",
         ],
         "ray": [
@@ -153,7 +153,7 @@
         ],
         "synapse": [
             "joblibspark>=0.5.0",
-            "optuna==2.8.0",
+            "optuna>=2.8.0,<=3.6.1",
             "pyspark>=3.2.0",
         ],
         "autozero": ["scikit-learn", "pandas", "packaging"],