diff --git a/experimental/builder/include/ck_tile/builder/conv_algorithm_concepts.hpp b/experimental/builder/include/ck_tile/builder/conv_algorithm_concepts.hpp
index 4d81becfb5..4feb09cfe9 100644
--- a/experimental/builder/include/ck_tile/builder/conv_algorithm_concepts.hpp
+++ b/experimental/builder/include/ck_tile/builder/conv_algorithm_concepts.hpp
@@ -52,23 +52,20 @@ concept GridwiseWmmaGemmDescriptor = requires(T t) {
 };
 
 
-template <typename T>
-concept HasGemmKBatch = requires(T t) {
-    { t.k_batch_size};
-};
-
-// Concept to check if GEMM k batch size is specified.
-template <typename T>
-concept GemmKBatchSizeWellDefinedIfProvided = 
-    !HasGemmKBatch<T> || requires(T t) { {t.k_batch_size} -> std::convertible_to<size_t>; };
-
 // Concept for vectorized data transfer for convolution input tensors.
 template <typename T>
 concept BlockTransferDescriptor = requires(T t) {
     { t.k0 } -> std::convertible_to<size_t>;
     { t.m_n } -> std::convertible_to<size_t>;
     { t.k1 } -> std::convertible_to<size_t>;
-    GemmKBatchSizeWellDefinedIfProvided<T>;
+};
+
+template <typename T>
+concept BlockTransferDescriptorBwd = requires(T t) {
+    { t.k0 } -> std::convertible_to<size_t>;
+    { t.m_n } -> std::convertible_to<size_t>;
+    { t.k1 } -> std::convertible_to<size_t>;
+    { t.k_batch_size } -> std::convertible_to<size_t>;;
 };
 
 // Concept for thread cluster dimensions for GEMM output tensor.
@@ -210,6 +207,14 @@ concept SpecifiesBlockTransfer = requires(T t) {
     { T::transfer.c.thread_cluster_dims } -> ThreadClusterDescriptor;
 };
 
+// Concept to check if a struct specifies convolution input and output block transfer info (Bwd direction).
+template <typename T>
+concept SpecifiesBlockTransferBwd = requires(T t) {
+    { T::transfer.a.block_transfer } -> BlockTransferDescriptorBwd;
+    { T::transfer.b.block_transfer } -> BlockTransferDescriptorBwd;
+    { T::transfer.c.thread_cluster_dims } -> ThreadClusterDescriptor;
+};
+
 // Concept to check if a struct specifies convolution scalar per vector infor for A, B and C.
 template <typename T>
 concept SpecifiesTileTransfer = requires(T t) {
diff --git a/experimental/builder/include/ck_tile/builder/factory/conv_algorithms.hpp b/experimental/builder/include/ck_tile/builder/factory/conv_algorithms.hpp
index 4f268b05e9..71e3d03da3 100644
--- a/experimental/builder/include/ck_tile/builder/factory/conv_algorithms.hpp
+++ b/experimental/builder/include/ck_tile/builder/factory/conv_algorithms.hpp
@@ -242,7 +242,7 @@ template <typename T>
 struct BwdXdlAlgorithm {
     CHECK_CONCEPT(T, ConvAlgorithmDescriptor)
     CHECK_CONCEPT(T, SpecifiesThreadBlock)
-    CHECK_CONCEPT(T, SpecifiesBlockTransfer)
+    CHECK_CONCEPT(T, SpecifiesBlockTransferBwd)
     CHECK_CONCEPT(T, SpecifiesLdsTransfer)
     CHECK_CONCEPT(T, SpecifiesThreadClusterAccessOrder)
     CHECK_CONCEPT(T, SpecifiesSourceAccessOrder)
@@ -252,7 +252,7 @@ struct BwdXdlAlgorithm {
 
     static constexpr bool c1 = c_ConvAlgorithmDescriptor;
     static constexpr bool c2 = c_SpecifiesThreadBlock;
-    static constexpr bool c3 = c_SpecifiesBlockTransfer;
+    static constexpr bool c3 = c_SpecifiesBlockTransferBwd;
     static constexpr bool c4 = c_SpecifiesLdsTransfer;
     static constexpr bool c5 = c_SpecifiesThreadClusterAccessOrder;
     static constexpr bool c6 = c_SpecifiesSourceAccessOrder;
@@ -269,7 +269,7 @@ struct BwdXdlAlgorithm {
                "Concepts for BwdXdl Algorithm:\n") +
                DIAGNOSTIC_LINE(ConvAlgorithmDescriptor) +
                DIAGNOSTIC_LINE(SpecifiesThreadBlock) +
-               DIAGNOSTIC_LINE(SpecifiesBlockTransfer) +
+               DIAGNOSTIC_LINE(SpecifiesBlockTransferBwd) +
                DIAGNOSTIC_LINE(SpecifiesLdsTransfer) +
                DIAGNOSTIC_LINE(SpecifiesThreadClusterAccessOrder) +
                DIAGNOSTIC_LINE(SpecifiesSourceAccessOrder) +