diff --git a/example/ck_tile/18_hstu_attention/scripts/bench_batched_causal.sh b/example/ck_tile/18_hstu_attention/scripts/bench_batched_causal.sh
index 168b3196fc..3df6072c03 100644
--- a/example/ck_tile/18_hstu_attention/scripts/bench_batched_causal.sh
+++ b/example/ck_tile/18_hstu_attention/scripts/bench_batched_causal.sh
@@ -1,7 +1,19 @@
 #!/bin/bash
 
+set +x
+
 BUILD=build
-EXE=$BUILD/bin/tile_example_hstu_attention
+
+USE_SOFTMAX=0
+if [ $# -ge 1 ]; then
+    USE_SOFTMAX=$1
+fi
+
+if [ $USE_SOFTMAX -eq 1 ]; then
+    EXE="$BUILD/bin/tile_example_hstu_attention -softmax=1"
+else
+    EXE="$BUILD/bin/tile_example_hstu_attention"
+fi
 
 dtype="bf16"
 hdim=128
diff --git a/example/ck_tile/18_hstu_attention/scripts/bench_jagged_causal.sh b/example/ck_tile/18_hstu_attention/scripts/bench_jagged_causal.sh
index 165a99d1c2..8069106d07 100644
--- a/example/ck_tile/18_hstu_attention/scripts/bench_jagged_causal.sh
+++ b/example/ck_tile/18_hstu_attention/scripts/bench_jagged_causal.sh
@@ -1,8 +1,19 @@
 #!/bin/bash
 
 set +x
+
 BUILD=build
-EXE=$BUILD/bin/tile_example_hstu_attention
+
+USE_SOFTMAX=0
+if [ $# -ge 1 ]; then
+    USE_SOFTMAX=$1
+fi
+
+if [ $USE_SOFTMAX -eq 1 ]; then
+    EXE="$BUILD/bin/tile_example_hstu_attention -softmax=1"
+else
+    EXE="$BUILD/bin/tile_example_hstu_attention"
+fi
 
 dtype="bf16"
 hdim=128
diff --git a/example/ck_tile/18_hstu_attention/scripts/bench_jagged_causal_local.sh b/example/ck_tile/18_hstu_attention/scripts/bench_jagged_causal_local.sh
index fd6c8acae6..f5fee117d6 100644
--- a/example/ck_tile/18_hstu_attention/scripts/bench_jagged_causal_local.sh
+++ b/example/ck_tile/18_hstu_attention/scripts/bench_jagged_causal_local.sh
@@ -1,8 +1,19 @@
 #!/bin/bash
 
 set +x
+
 BUILD=build
-EXE=$BUILD/bin/tile_example_hstu_attention
+
+USE_SOFTMAX=0
+if [ $# -ge 1 ]; then
+    USE_SOFTMAX=$1
+fi
+
+if [ $USE_SOFTMAX -eq 1 ]; then
+    EXE="$BUILD/bin/tile_example_hstu_attention -softmax=1"
+else
+    EXE="$BUILD/bin/tile_example_hstu_attention"
+fi
 
 dtype="bf16"
 hdim=128
diff --git a/example/ck_tile/18_hstu_attention/scripts/bench_jagged_causal_mattn0_full0.sh b/example/ck_tile/18_hstu_attention/scripts/bench_jagged_causal_mattn0_full0.sh
index 51972213cd..9cd371aaa6 100644
--- a/example/ck_tile/18_hstu_attention/scripts/bench_jagged_causal_mattn0_full0.sh
+++ b/example/ck_tile/18_hstu_attention/scripts/bench_jagged_causal_mattn0_full0.sh
@@ -1,8 +1,19 @@
 #!/bin/bash
 
 set +x
+
 BUILD=build
-EXE=$BUILD/bin/tile_example_hstu_attention
+
+USE_SOFTMAX=0
+if [ $# -ge 1 ]; then
+    USE_SOFTMAX=$1
+fi
+
+if [ $USE_SOFTMAX -eq 1 ]; then
+    EXE="$BUILD/bin/tile_example_hstu_attention -softmax=1"
+else
+    EXE="$BUILD/bin/tile_example_hstu_attention"
+fi
 
 dtype="bf16"
 
diff --git a/example/ck_tile/18_hstu_attention/scripts/bench_jagged_causal_mattn256_full256.sh b/example/ck_tile/18_hstu_attention/scripts/bench_jagged_causal_mattn256_full256.sh
index 92ec23deab..e992d61849 100644
--- a/example/ck_tile/18_hstu_attention/scripts/bench_jagged_causal_mattn256_full256.sh
+++ b/example/ck_tile/18_hstu_attention/scripts/bench_jagged_causal_mattn256_full256.sh
@@ -1,8 +1,19 @@
 #!/bin/bash
 
 set +x
+
 BUILD=build
-EXE=$BUILD/bin/tile_example_hstu_attention
+
+USE_SOFTMAX=0
+if [ $# -ge 1 ]; then
+    USE_SOFTMAX=$1
+fi
+
+if [ $USE_SOFTMAX -eq 1 ]; then
+    EXE="$BUILD/bin/tile_example_hstu_attention -softmax=1"
+else
+    EXE="$BUILD/bin/tile_example_hstu_attention"
+fi
 
 dtype="bf16"
 
diff --git a/example/ck_tile/18_hstu_attention/scripts/benchmark_hstu_attention.sh b/example/ck_tile/18_hstu_attention/scripts/benchmark_hstu_attention.sh
index 4bc04f6fc9..9c3cbf2fd6 100644
--- a/example/ck_tile/18_hstu_attention/scripts/benchmark_hstu_attention.sh
+++ b/example/ck_tile/18_hstu_attention/scripts/benchmark_hstu_attention.sh
@@ -1,7 +1,19 @@
 #!/bin/bash
 
+set +x
+
 BUILD=build
-EXE=$BUILD/bin/tile_example_hstu_attention
+
+USE_SOFTMAX=0
+if [ $# -ge 1 ]; then
+    USE_SOFTMAX=$1
+fi
+
+if [ $USE_SOFTMAX -eq 1 ]; then
+    EXE="$BUILD/bin/tile_example_hstu_attention -softmax=1"
+else
+    EXE="$BUILD/bin/tile_example_hstu_attention"
+fi
 
 for dtype in "fp16" "bf16"; do
     for seqlen in 512 1024 3072; do