diff --git a/kleidicv/src/arithmetics/exp_neon.cpp b/kleidicv/src/arithmetics/exp_neon.cpp
index 274dc5ee3344b83cd6c2ffc47997730fc1e9343d..9a4b19ccc0e48a1e609b0aeab43b44abf1cad255 100644
--- a/kleidicv/src/arithmetics/exp_neon.cpp
+++ b/kleidicv/src/arithmetics/exp_neon.cpp
@@ -60,7 +60,7 @@ class Exp<float> final : public UnrollOnce {
     float32x4_t s2 = vreinterpretq_f32_u32(e - b);
     uint32x4_t cmp = vcagtq_f32(n, vdupq_n(192.0F));
     float32x4_t r1 = s1 * s1;
-    float32x4_t r0 = poly * s1 * s2;
+    float32x4_t r0 = (poly * s1) * s2;
     return vreinterpretq_f32_u32((cmp & vreinterpretq_u32_f32(r1)) |
                                  (~cmp & vreinterpretq_u32_f32(r0)));
   }
diff --git a/kleidicv/src/arithmetics/exp_sc.h b/kleidicv/src/arithmetics/exp_sc.h
index fea6b6ad9921bacb27ed10034e7ff8c30f51f976..41f9f813ec0048df8b480f3488f82cd8b4db068f 100644
--- a/kleidicv/src/arithmetics/exp_sc.h
+++ b/kleidicv/src/arithmetics/exp_sc.h
@@ -64,7 +64,7 @@ class Exp<float, TryShortPath> final : public UnrollOnce {
     svfloat32_t s2 = svreinterpret_f32(svsub_x(pg, e, b));
     svbool_t cmp = svacgt(pg, n, 192.0F);
     svfloat32_t r1 = svmul_x(pg, s1, s1);
-    svfloat32_t r0 = svmul_x(pg, poly, svmul_x(pg, s1, s2));
+    svfloat32_t r0 = svmul_x(pg, s2, svmul_x(pg, poly, s1));
 
     return svsel(cmp, r1, r0);
   }
diff --git a/test/api/test_exp.cpp b/test/api/test_exp.cpp
index f4f7b2af139884d205c9114a83b56a667a60e28b..0b87c7aba4ce04024ee1369d2ee0421524154229 100644
--- a/test/api/test_exp.cpp
+++ b/test/api/test_exp.cpp
@@ -61,7 +61,7 @@ class ExpTestSpecial<float> final : public UnaryOperationTest<float> {
   static const std::vector<ElementType>& input_values() {
     static const std::vector<ElementType> kInputValues = {
         -105.31, -100.07, -81.012, -47.66, -3.1088, -0.21,
-        0.7,     6.2,     39.7201, 86.11,  88.947};
+        0.7,     6.2,     39.7201, 86.11,  88.7,    88.947};
 
     return kInputValues;
   }