diff --git a/muh/include/muh/tuning/tuning_select_if.cuh b/muh/include/muh/tuning/tuning_select_if.cuh
new file mode 100644
index 00000000..a75c7616
--- /dev/null
+++ b/muh/include/muh/tuning/tuning_select_if.cuh
@@ -0,0 +1,69 @@
+// muh/include/muh/tuning/tuning_select_if.cuh — BI-V100
+//
+// Mirrors: cccl_upstream/cub/cub/device/dispatch/tuning/tuning_select_if.cuh
+// CCCL SM100: 77 specializations, 38 SMEM OVERFLOW (max tile=163840)
+//   Most overflows from threads=896-1024 with items=20 at type_size=4-8
+//
+// vllm relevance: token filtering (select tokens matching criteria)
+// SMEM risk: CRITICAL. Must clamp threads and items aggressively.
+//   Strategy: cap tile = threads * items * max(key_size, value_size) ≤ 48KB
+
+#pragma once
+
+#include "muh/hardware.cuh"
+#include "muh/tuning/common.cuh"
+
+namespace muh::tuning::select_if {
+
+struct SelectIfPolicy {
+  int threads_per_block;
+  int items_per_thread;
+  BlockLoadAlgorithm load_algorithm;
+  CacheLoadModifier load_modifier;
+  BlockScanAlgorithm scan_algorithm;
+  LookbackDelayPolicy delay;
+  bool may_alias;
+};
+
+struct policy_selector {
+  int input_size;
+  int flag_size;
+  int output_size;
+  int offset_size;
+  bool may_alias;
+
+  constexpr SelectIfPolicy operator()(const hardware_capability& hw) const {
+    // Effective element size: max of input/output for SMEM tile
+    int elem_size = input_size > output_size ? input_size : output_size;
+    
+    // Start from conservative values and validate
+    int threads = 256;
+    int items = 14;
+    
+    if (elem_size <= 2) {
+      threads = 384; items = 20;
+    } else if (elem_size <= 4) {
+      threads = 320; items = 16;
+    } else if (elem_size <= 8) {
+      threads = 256; items = 12;
+    } else {
+      threads = 192; items = 8;
+    }
+    
+    // SMEM: select_if needs input tile + output tile + flags
+    // Conservative: 2 * threads * items * elem_size + threads * items * flag_size
+    int smem_needed = threads * items * (2 * elem_size + flag_size);
+    while (smem_needed > hw.max_shared_memory_per_block && items > 1) {
+      items--;
+      smem_needed = threads * items * (2 * elem_size + flag_size);
+    }
+
+    return {threads, items, BLOCK_LOAD_WARP_TRANSPOSE,
+            may_alias ? LOAD_CA : LOAD_DEFAULT,
+            BLOCK_SCAN_WARP_SCANS,
+            {LookbackDelayAlgorithm::exponential_backon, 350, 450},
+            may_alias};
+  }
+};
+
+} // namespace muh::tuning::select_if