diff --git a/ethosu/regor/common/shape.hpp b/ethosu/regor/common/shape.hpp
index 1a2fa0004d51ffbb24a0cf06c8288fe0865c1988..71a248f8e165fe5e64c13b709281e7190319ab91 100644
--- a/ethosu/regor/common/shape.hpp
+++ b/ethosu/regor/common/shape.hpp
@@ -556,6 +556,12 @@ public:
         return true;
     }
 
+    // Returns true if two shapes are equal, ignoring leading dimensions that are 1
+    static bool IsReducedEqual(const Shape &a, const Shape &b)
+    {
+        return MaxAxisFunc<std::not_equal_to<int32_t>, 1>(a, b) == 0;
+    }
+
     template<typename TYPE>
     int ToNHWC(TYPE *buffer, size_t length) const
     {
@@ -663,6 +669,32 @@ private:
         return tmp;
     }
 
+    template<typename FUNC, int MISSING_VALUE = 0>
+    static unsigned MaxAxisFunc(const Shape &a, const Shape &b)
+    {
+        bool a_longer = a.Size() >= b.Size();
+        int length = a_longer ? a.Size() : b.Size();
+        assert(length < 32);
+        int shortest = a_longer ? b.Size() : a.Size();
+        assert(shortest < 32);
+
+        auto *pa = a.Storage();
+        auto *pb = b.Storage();
+        unsigned axisMask = 0;
+
+        int i = 0;
+        for ( ; i < shortest; i++ )
+        {
+            if ( FUNC()(pa[i], pb[i]) ) axisMask |= 1 << i;
+        }
+        for ( ; i < length; i++ )
+        {
+            if ( a_longer && FUNC()(pa[i], MISSING_VALUE) ) axisMask |= 1 << i;
+            else if ( !a_longer && FUNC()(MISSING_VALUE, pb[i]) ) axisMask |= 1 << i;
+        }
+        return axisMask;
+    }
+
     // Apply a function to the maximum number of axes between two shapes. For missing
     // axes either take from the longest shape, or substitute a constant value.
     template<typename FUNC, bool TAKE_LONGEST, int MISSING_VALUE = 0>
diff --git a/ethosu/regor/compiler/graphir_optimiser.cpp b/ethosu/regor/compiler/graphir_optimiser.cpp
index eb103f68a9195a7aac133e0ea12f441a08889742..30add616e9b30d03618df6d5531fce80edef7e9d 100644
--- a/ethosu/regor/compiler/graphir_optimiser.cpp
+++ b/ethosu/regor/compiler/graphir_optimiser.cpp
@@ -2035,7 +2035,6 @@ Operation *GraphIrOptimiser::MoveSplitSliceToConsumer(Graph *const, Operation *c
         if ( ofm->Readers().size() == 1 )
         {
             auto cons = ofm->Readers().front();
-            auto consOfmConn = cons->Output(TensorUsage::OFM);
             auto *consIfm0 = cons->IFM(0);
             auto *consIfm1 = cons->IFM(1);
 
@@ -2044,15 +2043,16 @@ Operation *GraphIrOptimiser::MoveSplitSliceToConsumer(Graph *const, Operation *c
             {
                 // Check if ifm0 consumer has correct shape
                 auto *consIfm0Conn = cons->Input(TensorUsage::IFM0);
-                ifmShapeEqual = consIfm0Conn->shape == ofmConn->shape;
+                ifmShapeEqual = Shape::IsReducedEqual(consIfm0Conn->shape, ofmConn->shape);
             }
             else if ( consIfm1 != nullptr && consIfm1 == ofm )
             {
                 // Check if ifm1 consumer has correct shape
                 auto *consIfm1Conn = cons->Input(TensorUsage::IFM1);
-                ifmShapeEqual = consIfm1Conn->shape == ofmConn->shape;
+                ifmShapeEqual = Shape::IsReducedEqual(consIfm1Conn->shape, ofmConn->shape);
             }
 
+            // Calculate the consumer transpose type
             TransposeType consumerTranspose = TransposeType::None;
             if ( cons->Type() == OpType::Transpose )
             {
@@ -2061,7 +2061,7 @@ Operation *GraphIrOptimiser::MoveSplitSliceToConsumer(Graph *const, Operation *c
 
             // We can only move to consumer if there is no transpose on the op that we move to,
             // otherwise the IFM shape may change and transposition will be wrong.
-            if ( !IsReshape(cons->Type()) && ofmConn->shape == Shape::PadAxes(ofm->StorageShape(), 4, 1) && IsNone(consumerTranspose) && ifmShapeEqual )
+            if ( !IsReshape(cons->Type()) && Shape::IsReducedEqual(ofmConn->shape, ofm->StorageShape()) && IsNone(consumerTranspose) && ifmShapeEqual )
             {
                 // Split/Slice can be performed by tensor consumer
                 MoveToConsumer(operation, cons.get());
diff --git a/ethosu/regor/test/test_shape.cpp b/ethosu/regor/test/test_shape.cpp
index 055413e8ddd946d2399d60ddc8c03cb35f9b7539..735b950d2cf052c75ee9cfac683e207afe4630a5 100644
--- a/ethosu/regor/test/test_shape.cpp
+++ b/ethosu/regor/test/test_shape.cpp
@@ -1,5 +1,5 @@
 //
-// SPDX-FileCopyrightText: Copyright 2021, 2023-2024 Arm Limited and/or its affiliates <open-source-office@arm.com>
+// SPDX-FileCopyrightText: Copyright 2021, 2023-2025 Arm Limited and/or its affiliates <open-source-office@arm.com>
 //
 // SPDX-License-Identifier: Apache-2.0
 //
@@ -458,3 +458,21 @@ TEST_CASE("To Mask")
     uint32_t mask2 = shape2.ToMask();
     REQUIRE((mask2 == 0x2301));
 }
+
+TEST_CASE("Is reduced equal")
+{
+    Shape shape1a(3, 3);
+    Shape shape1b(1, 1, 3, 3);
+    REQUIRE(Shape::IsReducedEqual(shape1a, shape1b));
+    REQUIRE(Shape::IsReducedEqual(shape1b, shape1a));
+
+    Shape shape3a(3, 3);
+    Shape shape3b(3, 3, 1);
+    REQUIRE_FALSE(Shape::IsReducedEqual(shape3a, shape3b));
+    REQUIRE_FALSE(Shape::IsReducedEqual(shape3b, shape3a));
+
+    Shape shape2a(1);
+    Shape shape2b(1, 1, 1);
+    REQUIRE(Shape::IsReducedEqual(shape2a, shape2b));
+    REQUIRE(Shape::IsReducedEqual(shape2b, shape2a));
+}