[NPUW] Added first NPUW LLM test (openvinotoolkit#32992)

### Details:
 - *Added first NPUW LLM integration test to check EISW-164271*

### Tickets:
 - *EISW-164271*

Author: AsyaPronina

src/plugins/intel_npu/tests/functional/CMakeLists.txt

@@ -52,6 +52,11 @@ ov_add_test_target(
   openvino::runtime
   openvino::npu_al)
 
+if(ENABLE_INTEL_CPU)
+    add_dependencies(${TARGET_NAME} openvino_intel_cpu_plugin)
+    target_compile_definitions(${TARGET_NAME} PRIVATE WITH_CPU_PLUGIN)
+endif()
+
 if(WIN32)
     if(ENABLE_DX12)
         target_compile_definitions(${TARGET_NAME} PRIVATE ENABLE_DX12)

src/plugins/intel_npu/tests/functional/behavior/npuw/behavior_tests.cpp

@@ -1,9 +1,9 @@
-// Copyright (C) 2024 Intel Corporation
+// Copyright (C) 2024-2025 Intel Corporation
 // SPDX-License-Identifier: Apache-2.0
 //
 
 #include "behavior_tests.hpp"
-#include "comparators/nrmse.hpp"
+#include "test_engine/comparators/nrmse.hpp"
 #include "openvino/util/shared_object.hpp"
 #include "openvino/util/file_util.hpp"
 

src/plugins/intel_npu/tests/functional/behavior/npuw/behavior_tests.hpp

@@ -1,4 +1,4 @@
-// Copyright (C) 2024 Intel Corporation
+// Copyright (C) 2024-2025 Intel Corporation
 // SPDX-License-Identifier: Apache-2.0
 //
 #pragma once
@@ -8,10 +8,10 @@
 #include <memory>
 #include <random>
 
-#include "generators/model_generator.hpp"
+#include "test_engine/models/model_generator.hpp"
+#include "test_engine/mocks/mock_plugins.hpp"
+#include "test_engine/mocks/register_in_ov.hpp"
 #include "intel_npu/npuw_private_properties.hpp"
-#include "mocks/mock_plugins.hpp"
-#include "mocks/register_in_ov.hpp"
 #include "openvino/runtime/core.hpp"
 #include "openvino/runtime/iplugin.hpp"
 #include "openvino/runtime/make_tensor.hpp"

src/plugins/intel_npu/tests/functional/behavior/npuw/llm_smoke_accuracy_tests.cpp

@@ -0,0 +1,107 @@
+// Copyright (C) 2025 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "test_engine/models/find_model.hpp"
+#include "test_engine/comparators/nrmse.hpp"
+#include "test_engine/simple_llm_pipeline.hpp"
+#include "intel_npu/npuw_private_properties.hpp"
+#include "openvino/util/shared_object.hpp"
+#include "openvino/util/file_util.hpp"
+
+#include <filesystem>
+#include <algorithm>
+
+#include <gtest/gtest.h>
+
+using namespace testing;
+using namespace ov::npuw::tests;
+using namespace ov::intel_npu::npuw;
+
+namespace {
+const std::string minicpm_05_b_name = "MiniCPM4-0.5B_int4_sym_group128_dyn_stateful";
+
+struct GroundTruth {
+    std::vector<int64_t> prompt;
+    std::vector<int64_t> answer;
+};
+
+const GroundTruth What_is_OpenVINO_templated = {
+    // Prompt:
+    // <|im_start|>user
+    // What is OpenVINO?<|im_end|>
+    // <|im_start|>assistant
+    {73441, 3060, 5, 5856, 1410, 6404, 59408, 2097, 59383, 74, 73440, 59320, 5, 73441, 16434, 5},
+    // Answer: OpenVINO
+    { 9254, 59408, 2097, 59383}
+};
+
+using LLMTestParams = std::tuple<std::string, ov::AnyMap, GroundTruth>;
+} // anonymous namespace
+
+class LLMSmokeAccuracyTestsNPUW : public ::testing::TestWithParam<LLMTestParams> {
+public:
+    void SetUp() override {      
+        // NOTE: TEMPLATE plugin in OpenVINO works for ~20 minute to generate
+        //       first token from prefill model and crashes on launch of
+        //       3rd subrequest in generate model.
+        //       There is no such issue with CPU plugin, so CPU plugin was choosen
+        //       for accuracy checks.
+        // Test only makes sense if CPU plugin is enabled.
+#if !defined(OPENVINO_STATIC_LIBRARY) && defined(WITH_CPU_PLUGIN) && \
+    (defined(OPENVINO_ARCH_X86) || defined(OPENVINO_ARCH_X86_64))
+        const char* cpu_plugin_file_name = "openvino_intel_cpu_plugin";
+        // Register CPU plugin in OpenVINO:
+        try {
+            core.register_plugin(std::string(cpu_plugin_file_name) + OV_BUILD_POSTFIX, "CPU");
+        } catch (ov::Exception& ex) {
+            if (std::string{ex.what()}.find("Device with \"CPU\"  is already registered in the OpenVINO Runtime")
+                == std::string::npos) {
+                throw ex;
+            }
+        }
+#else
+        GTEST_SKIP() << "CPU plugin is not enabled or platform requirements are not met, skipping the test!";
+#endif // not OPENVINO_STATIC_LIBRARY && WITH_CPU_PLUGIN &&
+       // (defined(OPENVINO_ARCH_X86) || defined(OPENVINO_ARCH_X86_64))
+
+        auto param = GetParam();
+        std::string model_name;
+        ov::AnyMap config;
+        GroundTruth input_and_reference_ids;
+        std::tie(model_name, config, input_and_reference_ids) = param;
+
+        model_path = find_model(model_name);
+        if (model_path == "") {
+            GTEST_SKIP() << "Test model is not found, skipping the test!";
+        }
+
+        input_ids = input_and_reference_ids.prompt;
+        reference_ids = input_and_reference_ids.answer;
+
+        config["NPUW_DEVICES"] = "CPU";
+        config["NPUW_LLM_MAX_PROMPT_LEN"] = 128;
+        config["NPUW_LLM_MIN_RESPONSE_LEN"] = 4;
+        simple_llm.initialize(model_path, core, config);
+    }
+
+protected:
+    ov::Core core;
+    SimpleLLMPipeline simple_llm;
+    std::string model_path;
+    std::vector<int64_t> input_ids;
+    std::vector<int64_t> actual_ids;
+    std::vector<int64_t> reference_ids;
+};
+
+TEST_P(LLMSmokeAccuracyTestsNPUW, ConfigIsAccurate) {
+    actual_ids = simple_llm.generate(input_ids);
+    for (std::size_t i = 0; i < actual_ids.size(); ++i) {
+        ASSERT_EQ(actual_ids[i], reference_ids[i]);
+    }
+}
+
+INSTANTIATE_TEST_SUITE_P(LLMSmokeAccuracyNPUW_FAST_COMPILE, LLMSmokeAccuracyTestsNPUW,
+    ::testing::Combine(testing::Values(minicpm_05_b_name),
+                       testing::Values(ov::AnyMap{}),
+                       testing::Values(What_is_OpenVINO_templated)));