diff --git a/mlir/include/mlir/IR/BuiltinAttributeInterfaces.td b/mlir/include/mlir/IR/BuiltinAttributeInterfaces.td
--- a/mlir/include/mlir/IR/BuiltinAttributeInterfaces.td
+++ b/mlir/include/mlir/IR/BuiltinAttributeInterfaces.td
@@ -432,6 +432,11 @@
       if (!values)
         return std::nullopt;
 
+      // Check that an element can be casted as 'T', which is a derived attribute.
+      if (!empty())
+        if (!(*(value_begin<Attribute>())).dyn_cast<T>())
+          return std::nullopt;
+
       auto castFn = [](Attribute attr) { return ::llvm::cast<T>(attr); };
       return DerivedAttrValueIteratorRange<T>(
         getShapedType(),
diff --git a/mlir/include/mlir/IR/BuiltinAttributes.h b/mlir/include/mlir/IR/BuiltinAttributes.h
--- a/mlir/include/mlir/IR/BuiltinAttributes.h
+++ b/mlir/include/mlir/IR/BuiltinAttributes.h
@@ -982,10 +982,11 @@
 template <typename T>
 auto SparseElementsAttr::try_value_begin_impl(OverloadToken<T>) const
     -> FailureOr<iterator<T>> {
-  auto zeroValue = getZeroValue<T>();
   auto valueIt = getValues().try_value_begin<T>();
   if (failed(valueIt))
     return failure();
+
+  auto zeroValue = getZeroValue<T>();
   const std::vector<ptrdiff_t> flatSparseIndices(getFlattenedSparseIndices());
   std::function<T(ptrdiff_t)> mapFn =
       [flatSparseIndices{flatSparseIndices}, valueIt{std::move(*valueIt)},
diff --git a/mlir/test/IR/elements-attr-interface.mlir b/mlir/test/IR/elements-attr-interface.mlir
--- a/mlir/test/IR/elements-attr-interface.mlir
+++ b/mlir/test/IR/elements-attr-interface.mlir
@@ -14,6 +14,12 @@
 // expected-error@below {{Test iterating `IntegerAttr`: 10 : i64, 11 : i64, 12 : i64, 13 : i64, 14 : i64}}
 arith.constant #test.i64_elements<[10, 11, 12, 13, 14]> : tensor<5xi64>
 
+// expected-error@below {{Test iterating `int64_t`: unable to iterate type}}
+// expected-error@below {{Test iterating `uint64_t`: unable to iterate type}}
+// expected-error@below {{Test iterating `APInt`: unable to iterate type}}
+// expected-error@below {{Test iterating `IntegerAttr`: unable to iterate type}}
+arith.constant dense<[1.0, 2.0, 3.0, 4.0, 5.0]> : tensor<5xf32>
+
 // expected-error@below {{Test iterating `int64_t`: 10, 11, 12, 13, 14}}
 // expected-error@below {{Test iterating `uint64_t`: 10, 11, 12, 13, 14}}
 // expected-error@below {{Test iterating `APInt`: 10, 11, 12, 13, 14}}