diff --git a/example/ck_tile/38_block_scale_gemm/run_gemm_quant_example.inc b/example/ck_tile/38_block_scale_gemm/run_gemm_quant_example.inc
index 2ffe146a75..bce716079e 100644
--- a/example/ck_tile/38_block_scale_gemm/run_gemm_quant_example.inc
+++ b/example/ck_tile/38_block_scale_gemm/run_gemm_quant_example.inc
@@ -505,7 +505,8 @@ int run_gemm_example_with_layouts(const ck_tile::ArgParser& arg_parser,
             {
                 ck_tile::FillUniformDistribution<ADataType>{-2.0f, 3.0f, fill_seed(gen)}(a_m_k);
             }
-            ck_tile::FillUniformDistribution<AQDataType>{0, 0.5f, fill_seed(gen)}(*aq_tensor_ptr);
+            ck_tile::FillUniformDistribution<AQDataType>{-2.0f, 2.0f, fill_seed(gen)}(
+                *aq_tensor_ptr);
             ck_tile::FillUniformDistribution<BDataType>{-5.0f, 5.0f, fill_seed(gen)}(b_k_n);
         }
         else
@@ -528,7 +529,7 @@ int run_gemm_example_with_layouts(const ck_tile::ArgParser& arg_parser,
         if constexpr(QuantMode == ck_tile::QuantType::BQuantGrouped)
         {
             ck_tile::FillConstant<ADataType>{static_cast<ADataType>(0x38)}(a_m_k);
-            ck_tile::FillConstant<BDataType>{static_cast<BDataType>(0x38)}(b_k_n);
+            ck_tile::FillConstant<BDataType>{static_cast<BDataType>(0x22)}(b_k_n);
             ck_tile::FillConstant<BQDataType>{static_cast<BQDataType>(0.5f)}(*bq_tensor_ptr);
         }
         else
diff --git a/include/ck_tile/ops/gemm_quant/block/block_universal_gemm_ar_flatbr_bquant_cr.hpp b/include/ck_tile/ops/gemm_quant/block/block_universal_gemm_ar_flatbr_bquant_cr.hpp
index 2e4f60d0b2..a80583f999 100644
--- a/include/ck_tile/ops/gemm_quant/block/block_universal_gemm_ar_flatbr_bquant_cr.hpp
+++ b/include/ck_tile/ops/gemm_quant/block/block_universal_gemm_ar_flatbr_bquant_cr.hpp
@@ -190,6 +190,7 @@ struct BlockGemmWeightPreshuffleBQuantARegBRegCReg
                             scale_reg_dword = static_cast<uint32_t>(scale_reg);
                         }
 
+                        // cross lane ops to get the value of scale_reg.
                         int gathered_scale_reg = __builtin_amdgcn_ds_bpermute(
                             pull_from_lane << 2, __builtin_bit_cast(int, scale_reg_dword));
 
diff --git a/include/ck_tile/ops/gemm_quant/kernel/grouped_gemm_quant_kernel.hpp b/include/ck_tile/ops/gemm_quant/kernel/grouped_gemm_quant_kernel.hpp
index 8eb1ee4496..32f1279e93 100644
--- a/include/ck_tile/ops/gemm_quant/kernel/grouped_gemm_quant_kernel.hpp
+++ b/include/ck_tile/ops/gemm_quant/kernel/grouped_gemm_quant_kernel.hpp
@@ -390,7 +390,6 @@ struct QuantGroupedGemmKernel
         const auto& c_block_tile    = GemmPipeline{}.template operator()(a_block_window,
                                                                       b_block_window,
                                                                       bq_block_window,
-                                                                      kargs.N,
                                                                       num_loop,
                                                                       tail_num,
                                                                       smem_ptr_0,
@@ -458,7 +457,6 @@ struct QuantGroupedGemmKernel
             const auto& c_block_tile = GemmPipeline{}.template operator()(a_block_window,
                                                                           b_block_window,
                                                                           bq_block_window,
-                                                                          kargs.N,
                                                                           num_loop,
                                                                           has_hot_loop,
                                                                           tail_num,
diff --git a/include/ck_tile/ops/gemm_quant/pipeline/gemm_wp_bquant_pipeline_ag_bg_cr_v2.hpp b/include/ck_tile/ops/gemm_quant/pipeline/gemm_wp_bquant_pipeline_ag_bg_cr_v2.hpp
index 611ba21760..b92c9ee1fd 100644
--- a/include/ck_tile/ops/gemm_quant/pipeline/gemm_wp_bquant_pipeline_ag_bg_cr_v2.hpp
+++ b/include/ck_tile/ops/gemm_quant/pipeline/gemm_wp_bquant_pipeline_ag_bg_cr_v2.hpp
@@ -70,7 +70,7 @@ struct WPQuantBPipelineAgBgCrV2 : public WeightPreshufflePipelineAGmemBGmemCRegV
 
     static constexpr bool PreshuffleQuant = Problem::Traits::PreshuffleQuant;
     static constexpr index_t KPerBlockBQ =
-        integer_divide_ceil(BlockGemmShape::kK, QuantGroupSize::kK); // 256/128
+        integer_divide_ceil(BlockGemmShape::kK, QuantGroupSize::kK);
     static constexpr index_t QScalesPerBlockRow =
         integer_divide_ceil(kKPerBlock, QuantGroupSize::kK);