diff --git a/src/core/NEON/kernels/NEArithmeticAdditionKernel.cpp b/src/core/NEON/kernels/NEArithmeticAdditionKernel.cpp
index 59a454f..fc211f7 100644
--- a/src/core/NEON/kernels/NEArithmeticAdditionKernel.cpp
+++ b/src/core/NEON/kernels/NEArithmeticAdditionKernel.cpp
@@ -343,11 +343,7 @@
     const UniformQuantizationInfo iq2_info = in2->info()->quantization_info().uniform();
     const UniformQuantizationInfo oq_info  = out->info()->quantization_info().uniform();
 
-    const float32x4_t vscale1    = vdupq_n_f32(iq1_info.scale);
-    const float32x4_t vscale2    = vdupq_n_f32(iq2_info.scale);
     const float32x4_t invvscaleo = vdupq_n_f32(1.f / oq_info.scale);
-    const int32x4_t   voffset1   = vdupq_n_s32(iq1_info.offset);
-    const int32x4_t   voffset2   = vdupq_n_s32(iq2_info.offset);
     const float32x4_t voffseto   = vdupq_n_f32(oq_info.offset);
 
     if(is_broadcast_across_x)
@@ -360,6 +356,11 @@
         const UniformQuantizationInfo broadcast_qinfo      = broadcast_tensor->info()->quantization_info().uniform();
         const UniformQuantizationInfo non_broadcast_qinfo  = non_broadcast_tensor->info()->quantization_info().uniform();
 
+        const float32x4_t vscale1  = is_broadcast_input_2 ? vdupq_n_f32(iq1_info.scale) : vdupq_n_f32(iq2_info.scale);
+        const float32x4_t vscale2  = is_broadcast_input_2 ? vdupq_n_f32(iq2_info.scale) : vdupq_n_f32(iq1_info.scale);
+        const int32x4_t   voffset1 = is_broadcast_input_2 ? vdupq_n_s32(iq1_info.offset) : vdupq_n_s32(iq2_info.offset);
+        const int32x4_t   voffset2 = is_broadcast_input_2 ? vdupq_n_s32(iq2_info.offset) : vdupq_n_s32(iq1_info.offset);
+
         // Clear X Dimension on execution window as we handle manually
         non_broadcast_win.set(Window::DimX, Window::Dimension(0, 1, 1));
 
@@ -442,6 +443,10 @@
         Iterator input2(in2, input2_win);
         Iterator output(out, win);
 
+        const float32x4_t vscale1  = vdupq_n_f32(iq1_info.scale);
+        const float32x4_t vscale2  = vdupq_n_f32(iq2_info.scale);
+        const int32x4_t   voffset1 = vdupq_n_s32(iq1_info.offset);
+        const int32x4_t   voffset2 = vdupq_n_s32(iq2_info.offset);
         execute_window_loop(win, [&](const Coordinates &)
         {
             const auto input1_ptr = reinterpret_cast<const int8_t *>(input1.ptr());
diff --git a/tests/validation/NEON/ArithmeticAddition.cpp b/tests/validation/NEON/ArithmeticAddition.cpp
index 0d2ce6c..7b3d4f9 100644
--- a/tests/validation/NEON/ArithmeticAddition.cpp
+++ b/tests/validation/NEON/ArithmeticAddition.cpp
@@ -210,6 +210,9 @@
 template <typename T>
 using NEArithmeticAdditionQuantizedFixture = ArithmeticAdditionValidationQuantizedFixture<Tensor, Accessor, NEArithmeticAddition, T>;
 
+template <typename T>
+using NEArithmeticAdditionQuantizedBroadcastFixture = ArithmeticAdditionValidationQuantizedBroadcastFixture<Tensor, Accessor, NEArithmeticAddition, T>;
+
 TEST_SUITE(Quantized)
 TEST_SUITE(QASYMM8)
 FIXTURE_DATA_TEST_CASE(RunSmall,
@@ -247,6 +250,21 @@
     validate(Accessor(_target), _reference, tolerance_quant);
 #endif //__aarch64__
 }
+
+FIXTURE_DATA_TEST_CASE(RunSmallBroadcast, NEArithmeticAdditionQuantizedBroadcastFixture<int8_t>, framework::DatasetMode::ALL, combine(combine(combine(combine(combine(
+                           datasets::SmallShapesBroadcast(), ArithmeticAdditionQASYMM8SIGNEDDataset),
+                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE })),
+                       framework::dataset::make("Src0QInfo", { QuantizationInfo(0.5f, 20) })),
+                       framework::dataset::make("Src1QInfo", { QuantizationInfo(0.5f, 10) })),
+                       framework::dataset::make("OutQInfo", { QuantizationInfo(0.5f, 5) })))
+{
+    // Validate output
+#ifdef __aarch64__
+    validate(Accessor(_target), _reference);
+#else  //__aarch64__
+    validate(Accessor(_target), _reference, tolerance_quant);
+#endif //__aarch64__
+}
 TEST_SUITE_END() // QASYMM8_SIGNED
 
 TEST_SUITE(QSYMM16)
diff --git a/tests/validation/fixtures/ArithmeticOperationsFixture.h b/tests/validation/fixtures/ArithmeticOperationsFixture.h
index 6e00c46..9ba7bd3 100644
--- a/tests/validation/fixtures/ArithmeticOperationsFixture.h
+++ b/tests/validation/fixtures/ArithmeticOperationsFixture.h
@@ -195,6 +195,20 @@
 };
 
 template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
+class ArithmeticAdditionValidationQuantizedBroadcastFixture : public ArithmeticOperationGenericFixture<TensorType, AccessorType, FunctionType, T>
+{
+public:
+    template <typename...>
+    void setup(const TensorShape &shape0, const TensorShape &shape1, DataType data_type0, DataType data_type1, DataType output_data_type,
+               ConvertPolicy convert_policy, QuantizationInfo qinfo0, QuantizationInfo qinfo1, QuantizationInfo qinfo_out)
+    {
+        ArithmeticOperationGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(reference::ArithmeticOperation::ADD, shape0, shape1,
+                                                                                            data_type0, data_type1, output_data_type, convert_policy,
+                                                                                            qinfo0, qinfo1, qinfo_out, ActivationLayerInfo(), false);
+    }
+};
+
+template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
 class ArithmeticSubtractionBroadcastValidationFixture : public ArithmeticOperationGenericFixture<TensorType, AccessorType, FunctionType, T>
 {
 public: