diff --git a/sql/core/src/main/scala/org/apache/spark/sql/Dataset.scala b/sql/core/src/main/scala/org/apache/spark/sql/Dataset.scala
index ae8d33d8558b..524e231eb7eb 100644
--- a/sql/core/src/main/scala/org/apache/spark/sql/Dataset.scala
+++ b/sql/core/src/main/scala/org/apache/spark/sql/Dataset.scala
@@ -2541,7 +2541,9 @@ class Dataset[T] private[sql](
   def dropDuplicates(colNames: Seq[String]): Dataset[T] = withTypedPlan {
     val resolver = sparkSession.sessionState.analyzer.resolver
     val allColumns = queryExecution.analyzed.output
-    val groupCols = colNames.distinct.flatMap { (colName: String) =>
+    // SPARK-31990: We must keep `toSet.toSeq` here because of the backward compatibility issue
+    // (the Streaming's state store depends on the `groupCols` order).
+    val groupCols = colNames.toSet.toSeq.flatMap { (colName: String) =>
       // It is possibly there are more than one columns with the same name,
       // so we call filter instead of find.
       val cols = allColumns.filter(col => resolver(col.name, colName))