@@ -204,6 +204,7 @@ Status HudiHybridReader::_init_child_reader(format::TableReader* reader,
             .io_ctx = _io_ctx,
             .runtime_state = _runtime_state,
             .scanner_profile = _scanner_profile,
+            .parquet_profile_cache = _parquet_profile_cache,
             .push_down_agg_type = _push_down_agg_type,
             .push_down_count_columns = _push_down_count_columns,
             .condition_cache_digest = _condition_cache_digest,
 
@@ -157,6 +157,18 @@ class RefreshTrackingTableReader final : public TableReader {
     int refresh_count = 0;
 };
 
+class ProfileCacheTrackingTableReader final : public TableReader {
+public:
+    Status init(TableReadOptions&& options) override {
+        observed_cache = options.parquet_profile_cache;
+        return TableReader::init(std::move(options));
+    }
+
+    Status prepare_split(const SplitReadOptions&) override { return Status::OK(); }
+
+    parquet::ParquetProfileCache* observed_cache = nullptr;
+};
+
 // Scenario: FileScannerV2 Hudi native reader uses the split schema id to annotate the physical
 // file schema before TableColumnMapper runs. This keeps schema-evolved Hudi files on field-id
 // mapping, including renamed nested children.
@@ -308,6 +320,37 @@ TEST(HudiHybridReaderTest, AdaptiveBatchSizeReachesBothChildReaders) {
     EXPECT_EQ(child_batch_sizes.second, 123);
 }
 
+TEST(HudiHybridReaderTest, ForwardsParquetProfileCacheToNativeChild) {
+    RuntimeProfile profile("hudi_hybrid");
+    parquet::ParquetProfileCache cache(&profile);
+    TFileScanRangeParams scan_params;
+    scan_params.__set_format_type(TFileFormatType::FORMAT_PARQUET);
+    hudi::HudiHybridReader reader;
+    ProfileCacheTrackingTableReader* native = nullptr;
+    reader.TEST_set_child_reader_factories(
+            [&] {
+                auto child = std::make_unique<ProfileCacheTrackingTableReader>();
+                native = child.get();
+                return child;
+            },
+            [] { return std::make_unique<TableReader>(); });
+    ASSERT_TRUE(reader.init({.projected_columns = {},
+                             .conjuncts = {},
+                             .format = FileFormat::PARQUET,
+                             .scan_params = &scan_params,
+                             .io_ctx = nullptr,
+                             .runtime_state = nullptr,
+                             .scanner_profile = &profile,
+                             .parquet_profile_cache = &cache})
+                        .ok());
+    SplitReadOptions split;
+    split.current_split_format = FileFormat::PARQUET;
+    split.current_range.__set_format_type(TFileFormatType::FORMAT_PARQUET);
+    ASSERT_TRUE(reader.prepare_split(split).ok());
+    ASSERT_NE(native, nullptr);
+    EXPECT_EQ(native->observed_cache, &cache);
+}
+
 TEST(HudiHybridReaderTest, AggregatesConditionCacheHitsFromBothChildren) {
     hudi::HudiHybridReader reader;
     reader.TEST_install_batch_size_children();