diff --git a/exercises/00_hello_world/main.cpp b/exercises/00_hello_world/main.cpp
index 8866f3c15..eb1a725a7 100644
--- a/exercises/00_hello_world/main.cpp
+++ b/exercises/00_hello_world/main.cpp
@@ -6,6 +6,6 @@
 
 int main(int argc, char **argv) {
     // TODO: 在控制台输出 "Hello, InfiniTensor!" 并换行
-    std::cout : "Hello, InfiniTensor!" + std::endl;
+    std::cout<< "Hello, InfiniTensor!" << std::endl;
     return 0;
 }
diff --git a/exercises/01_variable&add/main.cpp b/exercises/01_variable&add/main.cpp
index 5014863fd..fb05c57f3 100644
--- a/exercises/01_variable&add/main.cpp
+++ b/exercises/01_variable&add/main.cpp
@@ -5,6 +5,7 @@
 int main(int argc, char **argv) {
     // TODO: 补全变量定义并打印加法运算
     // x ?
+    int x = 10;
     std::cout << x << " + " << x << " = " << x + x << std::endl;
     return 0;
 }
diff --git a/exercises/02_function/main.cpp b/exercises/02_function/main.cpp
index b5eef7f28..603a9ec04 100644
--- a/exercises/02_function/main.cpp
+++ b/exercises/02_function/main.cpp
@@ -5,6 +5,7 @@
 // NOTICE: 补充由内而外读法的机翻解释 <https://learn.microsoft.com/zh-cn/cpp/c-language/interpreting-more-complex-declarators?view=msvc-170>
 
 // TODO: 在这里声明函数
+int add(int a, int b);
 
 int main(int argc, char **argv) {
     ASSERT(add(123, 456) == 123 + 456, "add(123, 456) should be 123 + 456");
@@ -16,4 +17,5 @@ int main(int argc, char **argv) {
 
 int add(int a, int b) {
     // TODO: 补全函数定义，但不要移动代码行
+    return a+b;
 }
diff --git a/exercises/03_argument&parameter/main.cpp b/exercises/03_argument&parameter/main.cpp
index 7fb5d3c2f..76136eae6 100644
--- a/exercises/03_argument&parameter/main.cpp
+++ b/exercises/03_argument&parameter/main.cpp
@@ -8,19 +8,19 @@ void func(int);
 // TODO: 为下列 ASSERT 填写正确的值
 int main(int argc, char **argv) {
     auto arg = 99;
-    ASSERT(arg == ?, "arg should be ?");
+    ASSERT(arg == 99, "arg should be ?");
     std::cout << "befor func call: " << arg << std::endl;
     func(arg);
-    ASSERT(arg == ?, "arg should be ?");
+    ASSERT(arg == 99, "arg should be ?");
     std::cout << "after func call: " << arg << std::endl;
     return 0;
 }
 
 // TODO: 为下列 ASSERT 填写正确的值
 void func(int param) {
-    ASSERT(param == ?, "param should be ?");
+    ASSERT(param == 99, "param should be ?");
     std::cout << "befor add: " << param << std::endl;
     param += 1;
-    ASSERT(param == ?, "param should be ?");
+    ASSERT(param == 100, "param should be ?");
     std::cout << "after add: " << param << std::endl;
 }
diff --git a/exercises/04_static/main.cpp b/exercises/04_static/main.cpp
index f107762fa..64f417694 100644
--- a/exercises/04_static/main.cpp
+++ b/exercises/04_static/main.cpp
@@ -10,10 +10,10 @@ static int func(int param) {
 
 int main(int argc, char **argv) {
     // TODO: 将下列 `?` 替换为正确的数字
-    ASSERT(func(5) == ?, "static variable value incorrect");
-    ASSERT(func(4) == ?, "static variable value incorrect");
-    ASSERT(func(3) == ?, "static variable value incorrect");
-    ASSERT(func(2) == ?, "static variable value incorrect");
-    ASSERT(func(1) == ?, "static variable value incorrect");
+    ASSERT(func(5) == 5, "static variable value incorrect");
+    ASSERT(func(4) == 6, "static variable value incorrect");
+    ASSERT(func(3) == 7, "static variable value incorrect");
+    ASSERT(func(2) == 8, "static variable value incorrect");
+    ASSERT(func(1) == 9, "static variable value incorrect");
     return 0;
 }
diff --git a/exercises/05_constexpr/main.cpp b/exercises/05_constexpr/main.cpp
index d1db6c9d8..2f6a5451a 100644
--- a/exercises/05_constexpr/main.cpp
+++ b/exercises/05_constexpr/main.cpp
@@ -18,8 +18,10 @@ int main(int argc, char **argv) {
 
     // TODO: 观察错误信息，修改一处，使代码编译运行
     // PS: 编译运行，但是不一定能算出结果……
-    constexpr auto ANS_N = 90;
-    constexpr auto ANS = fibonacci(ANS_N);
+    constexpr auto ANS_N = 9;
+    //constexpr auto ANS = fibonacci(ANS_N);
+    auto ANS = fibonacci(ANS_N);
+
     std::cout << "fibonacci(" << ANS_N << ") = " << ANS << std::endl;
 
     return 0;
diff --git a/exercises/06_array/main.cpp b/exercises/06_array/main.cpp
index 61ed99ec0..bbce324cf 100644
--- a/exercises/06_array/main.cpp
+++ b/exercises/06_array/main.cpp
@@ -11,13 +11,13 @@ unsigned long long fibonacci(int i) {
             return 1;
         default:
             // TODO: 补全三目表达式缺失的部分
-            return <condition> ? <cache> : (arr[i] = fibonacci(i - 1) + fibonacci(i - 2));
+            return arr[i] ? arr[i] : (arr[i] = fibonacci(i - 1) + fibonacci(i - 2));
     }
 }
 
 int main(int argc, char **argv) {
     // TODO: 为此 ASSERT 填写正确的值
-    ASSERT(sizeof(arr) == ?, "sizeof array is size of all its elements");
+    ASSERT(sizeof(arr) == 720, "sizeof array is size of all its elements");
     // ---- 不要修改以下代码 ----
     ASSERT(fibonacci(2) == 1, "fibonacci(2) should be 1");
     ASSERT(fibonacci(20) == 6765, "fibonacci(20) should be 6765");
diff --git a/exercises/07_loop/main.cpp b/exercises/07_loop/main.cpp
index 44fd835cd..ed485308b 100644
--- a/exercises/07_loop/main.cpp
+++ b/exercises/07_loop/main.cpp
@@ -5,9 +5,11 @@
 // READ: 纯函数 <https://zh.wikipedia.org/wiki/%E7%BA%AF%E5%87%BD%E6%95%B0>
 static unsigned long long fibonacci(int i) {
     // TODO: 为缓存设置正确的初始值
-    static unsigned long long cache[96], cached;
+    static unsigned long long cache[96]{0, 1};
+    static int cached = 2;
+    
     // TODO: 设置正确的循环条件
-    for (; false; ++cached) {
+    for (; cached<=i; ++cached) {
         cache[cached] = cache[cached - 1] + cache[cached - 2];
     }
     return cache[i];
diff --git a/exercises/08_pointer/main.cpp b/exercises/08_pointer/main.cpp
index ba37173f5..622a2fd87 100644
--- a/exercises/08_pointer/main.cpp
+++ b/exercises/08_pointer/main.cpp
@@ -4,6 +4,17 @@
 bool is_fibonacci(int *ptr, int len, int stride) {
     ASSERT(len >= 3, "`len` should be at least 3");
     // TODO: 编写代码判断从 ptr 开始，每 stride 个元素取 1 个元素，组成长度为 n 的数列是否满足
+    for (int i = 0; i <= len - 3; i++)
+    {
+        int a = ptr[i * stride];
+        int b = ptr[(i+1)*stride];
+        int c = ptr[(i+2)*stride];
+        if (a + b != c)
+        {
+            return false;
+        }
+
+    }
     // arr[i + 2] = arr[i] + arr[i + 1]
     return true;
 }
diff --git a/exercises/09_enum&union/main.cpp b/exercises/09_enum&union/main.cpp
index 3f2cec768..5779dd314 100644
--- a/exercises/09_enum&union/main.cpp
+++ b/exercises/09_enum&union/main.cpp
@@ -37,6 +37,7 @@ ColorEnum convert_by_pun(Color c) {
 
     TypePun pun;
     // TODO: 补全类型双关转换
+    pun.c = c;
 
     return pun.e;
 }
diff --git a/exercises/10_trivial/main.cpp b/exercises/10_trivial/main.cpp
index 6ba23e48e..173ebbdd6 100644
--- a/exercises/10_trivial/main.cpp
+++ b/exercises/10_trivial/main.cpp
@@ -9,8 +9,9 @@ struct FibonacciCache {
 
 // TODO: 实现正确的缓存优化斐波那契计算
 static unsigned long long fibonacci(FibonacciCache &cache, int i) {
-    for (; false; ++cached) {
-        cache[cached] = cache[cached - 1] + cache[cached - 2];
+    int cached = cache.cached;
+    for (cached; cached<=i; ++cached) {
+        cache.cache[cached] = cache.cache[cached - 1] + cache.cache[cached - 2];
     }
     return cache.cache[i];
 }
@@ -20,6 +21,9 @@ int main(int argc, char **argv) {
     // NOTICE: C/C++ 中，读取未初始化的变量（包括结构体变量）是未定义行为
     // READ: 初始化的各种写法 <https://zh.cppreference.com/w/cpp/language/initialization>
     FibonacciCache fib;
+    fib.cache[0] = 0;
+    fib.cache[1] = 1;
+    fib.cached = 2;
     ASSERT(fibonacci(fib, 10) == 55, "fibonacci(10) should be 55");
     std::cout << "fibonacci(10) = " << fibonacci(fib, 10) << std::endl;
     return 0;
diff --git a/exercises/11_method/main.cpp b/exercises/11_method/main.cpp
index 0e08e0a36..57d04651d 100644
--- a/exercises/11_method/main.cpp
+++ b/exercises/11_method/main.cpp
@@ -6,7 +6,7 @@ struct Fibonacci {
 
     // TODO: 实现正确的缓存优化斐波那契计算
     unsigned long long get(int i) {
-        for (; false; ++cached) {
+        for (; cached<=i; ++cached) {
             cache[cached] = cache[cached - 1] + cache[cached - 2];
         }
         return cache[i];
@@ -16,6 +16,9 @@ struct Fibonacci {
 int main(int argc, char **argv) {
     // TODO: 初始化缓存结构体，使计算正确
     Fibonacci fib;
+    fib.cache[0] = 0;
+    fib.cache[1] = 1;
+    fib.cached = 2;
     ASSERT(fib.get(10) == 55, "fibonacci(10) should be 55");
     std::cout << "fibonacci(10) = " << fib.get(10) << std::endl;
     return 0;
diff --git a/exercises/12_method_const/main.cpp b/exercises/12_method_const/main.cpp
index 5521be4da..65c079be1 100644
--- a/exercises/12_method_const/main.cpp
+++ b/exercises/12_method_const/main.cpp
@@ -5,7 +5,8 @@
 struct Fibonacci {
     int numbers[11];
     // TODO: 修改方法签名和实现，使测试通过
-    int get(int i) {
+    int get(int i) const {
+        return numbers[i];
     }
 };
 
diff --git a/exercises/13_class/main.cpp b/exercises/13_class/main.cpp
index 9afa98c5b..b6dc868d1 100644
--- a/exercises/13_class/main.cpp
+++ b/exercises/13_class/main.cpp
@@ -15,10 +15,16 @@ class Fibonacci {
 public:
     // TODO: 实现构造器
     // Fibonacci()
+    Fibonacci(const size_t inputCache[16], int cached) {
+        for (int i = 0; i < 16; ++i) {
+            cache[i] = inputCache[i];
+        }
+        this->cached = cached;
+    }
 
     // TODO: 实现正确的缓存优化斐波那契计算
     size_t get(int i) {
-        for (; false; ++cached) {
+        for (; cached<=i; ++cached) {
             cache[cached] = cache[cached - 1] + cache[cached - 2];
         }
         return cache[i];
@@ -28,7 +34,8 @@ class Fibonacci {
 int main(int argc, char **argv) {
     // 现在类型拥有无参构造器，声明时会直接调用。
     // 这个写法不再是未定义行为了。
-    Fibonacci fib;
+    size_t cache[16]{0, 1};
+    Fibonacci fib(cache, 2);
     ASSERT(fib.get(10) == 55, "fibonacci(10) should be 55");
     std::cout << "fibonacci(10) = " << fib.get(10) << std::endl;
     return 0;
diff --git a/exercises/14_class_destruct/main.cpp b/exercises/14_class_destruct/main.cpp
index 42150e8ca..f00bcb789 100644
--- a/exercises/14_class_destruct/main.cpp
+++ b/exercises/14_class_destruct/main.cpp
@@ -11,20 +11,29 @@ class DynFibonacci {
 
 public:
     // TODO: 实现动态设置容量的构造器
-    DynFibonacci(int capacity): cache(new ?), cached(?) {}
+    DynFibonacci(int capacity) {
+        cache = new size_t[capacity];
+        cache[0] = 0;
+        cache[1] = 1;
+        cached = 2;
+    }
 
     // TODO: 实现析构器，释放缓存空间
     ~DynFibonacci();
 
     // TODO: 实现正确的缓存优化斐波那契计算
     size_t get(int i) {
-        for (; false; ++cached) {
+        for (; cached <= i; ++cached) {
             cache[cached] = cache[cached - 1] + cache[cached - 2];
         }
         return cache[i];
     }
 };
 
+DynFibonacci ::~DynFibonacci() {
+    delete[] cache;
+}
+
 int main(int argc, char **argv) {
     DynFibonacci fib(12);
     ASSERT(fib.get(10) == 55, "fibonacci(10) should be 55");
diff --git a/exercises/15_class_clone/main.cpp b/exercises/15_class_clone/main.cpp
index f74b70391..38737265a 100644
--- a/exercises/15_class_clone/main.cpp
+++ b/exercises/15_class_clone/main.cpp
@@ -10,17 +10,26 @@ class DynFibonacci {
 
 public:
     // TODO: 实现动态设置容量的构造器
-    DynFibonacci(int capacity): cache(new ?), cached(?) {}
+    DynFibonacci(int capacity) : cache(new size_t[capacity]), cached(2) {
+        cache[0] = 0;
+        cache[1] = 1;
+    }
 
     // TODO: 实现复制构造器
-    DynFibonacci(DynFibonacci const &) = delete;
+    DynFibonacci(DynFibonacci const &c) {
+        this->cache = new size_t[c.cached];
+        this->cached = c.cached;
+        for (int i = 0; i < c.cached; i++) {
+            this->cache[i] = c.cache[i];
+        }
+    }
 
     // TODO: 实现析构器，释放缓存空间
     ~DynFibonacci();
 
     // TODO: 实现正确的缓存优化斐波那契计算
     size_t get(int i) {
-        for (; false; ++cached) {
+        for (; cached <= i; ++cached) {
             cache[cached] = cache[cached - 1] + cache[cached - 2];
         }
         return cache[i];
@@ -38,6 +47,11 @@ class DynFibonacci {
     }
 };
 
+
+DynFibonacci::~DynFibonacci() {
+    delete[] cache;
+}
+
 int main(int argc, char **argv) {
     DynFibonacci fib(12);
     ASSERT(fib.get(10) == 55, "fibonacci(10) should be 55");
diff --git a/exercises/16_class_move/main.cpp b/exercises/16_class_move/main.cpp
index 8d2c421da..2269a80b8 100644
--- a/exercises/16_class_move/main.cpp
+++ b/exercises/16_class_move/main.cpp
@@ -15,21 +15,45 @@ class DynFibonacci {
 
 public:
     // TODO: 实现动态设置容量的构造器
-    DynFibonacci(int capacity): cache(new ?), cached(?) {}
+    DynFibonacci(int capacity): cache(new size_t[capacity]), cached(2) 
+    {
+        cache[0] = 0;
+        cache[1] = 1;
+    }
 
     // TODO: 实现移动构造器
-    DynFibonacci(DynFibonacci &&) noexcept = delete;
+    DynFibonacci(DynFibonacci &&c) noexcept : cache(c.cache), cached(c.cached)
+    {
+        c.cache =nullptr;
+        c.cached = 0;
+    }
 
     // TODO: 实现移动赋值
     // NOTICE: ⚠ 注意移动到自身问题 ⚠
-    DynFibonacci &operator=(DynFibonacci &&) noexcept = delete;
+    DynFibonacci& operator=(DynFibonacci&& c) noexcept
+    {
+        if (this != &c)
+        {
+            delete[] cache;
+            cache = c.cache;
+            cached = c.cached;
+
+            c.cache = nullptr;
+            c.cached = 0;
+        }
+        return *this;
+    }
 
     // TODO: 实现析构器，释放缓存空间
-    ~DynFibonacci();
+    ~DynFibonacci()
+    {
+        delete[]cache;
+        cache = nullptr;
+    }
 
     // TODO: 实现正确的缓存优化斐波那契计算
     size_t operator[](int i) {
-        for (; false; ++cached) {
+        for (; cached<=i; ++cached) {
             cache[cached] = cache[cached - 1] + cache[cached - 2];
         }
         return cache[i];
diff --git a/exercises/17_class_derive/main.cpp b/exercises/17_class_derive/main.cpp
index 819ae72fc..93014cfa6 100644
--- a/exercises/17_class_derive/main.cpp
+++ b/exercises/17_class_derive/main.cpp
@@ -50,9 +50,9 @@ int main(int argc, char **argv) {
     B b = B(3);
 
     // TODO: 补全三个类型的大小
-    static_assert(sizeof(X) == ?, "There is an int in X");
-    static_assert(sizeof(A) == ?, "There is an int in A");
-    static_assert(sizeof(B) == ?, "B is an A with an X");
+    static_assert(sizeof(X) == sizeof(int), "There is an int in X");
+    static_assert(sizeof(A) == sizeof(int), "There is an int in A");
+    static_assert(sizeof(B) == 2*sizeof(int), "B is an A with an X");
 
     i = 0;
     std::cout << std::endl
diff --git a/exercises/18_class_virtual/main.cpp b/exercises/18_class_virtual/main.cpp
index ac6382413..344709e4c 100644
--- a/exercises/18_class_virtual/main.cpp
+++ b/exercises/18_class_virtual/main.cpp
@@ -42,38 +42,38 @@ int main(int argc, char **argv) {
     C c;
     D d;
 
-    ASSERT(a.virtual_name() == '?', MSG);
-    ASSERT(b.virtual_name() == '?', MSG);
-    ASSERT(c.virtual_name() == '?', MSG);
-    ASSERT(d.virtual_name() == '?', MSG);
-    ASSERT(a.direct_name() == '?', MSG);
-    ASSERT(b.direct_name() == '?', MSG);
-    ASSERT(c.direct_name() == '?', MSG);
-    ASSERT(d.direct_name() == '?', MSG);
+    ASSERT(a.virtual_name() == 'A', MSG);
+    ASSERT(b.virtual_name() == 'B', MSG);
+    ASSERT(c.virtual_name() == 'C', MSG);
+    ASSERT(d.virtual_name() == 'C', MSG);
+    ASSERT(a.direct_name() == 'A', MSG);
+    ASSERT(b.direct_name() == 'B', MSG);
+    ASSERT(c.direct_name() == 'C', MSG);
+    ASSERT(d.direct_name() == 'D', MSG);
 
     A &rab = b;
     B &rbc = c;
     C &rcd = d;
 
-    ASSERT(rab.virtual_name() == '?', MSG);
-    ASSERT(rbc.virtual_name() == '?', MSG);
-    ASSERT(rcd.virtual_name() == '?', MSG);
-    ASSERT(rab.direct_name() == '?', MSG);
-    ASSERT(rbc.direct_name() == '?', MSG);
-    ASSERT(rcd.direct_name() == '?', MSG);
+    ASSERT(rab.virtual_name() == 'B', MSG);
+    ASSERT(rbc.virtual_name() == 'C', MSG);
+    ASSERT(rcd.virtual_name() == 'C', MSG);
+    ASSERT(rab.direct_name() == 'A', MSG);
+    ASSERT(rbc.direct_name() == 'B', MSG);
+    ASSERT(rcd.direct_name() == 'C', MSG);
 
     A &rac = c;
     B &rbd = d;
 
-    ASSERT(rac.virtual_name() == '?', MSG);
-    ASSERT(rbd.virtual_name() == '?', MSG);
-    ASSERT(rac.direct_name() == '?', MSG);
-    ASSERT(rbd.direct_name() == '?', MSG);
+    ASSERT(rac.virtual_name() == 'C', MSG);
+    ASSERT(rbd.virtual_name() == 'C', MSG);
+    ASSERT(rac.direct_name() == 'A', MSG);
+    ASSERT(rbd.direct_name() == 'B', MSG);
 
     A &rad = d;
 
-    ASSERT(rad.virtual_name() == '?', MSG);
-    ASSERT(rad.direct_name() == '?', MSG);
+    ASSERT(rad.virtual_name() == 'C', MSG);
+    ASSERT(rad.direct_name() == 'A', MSG);
 
     return 0;
 }
diff --git a/exercises/19_class_virtual_destruct/main.cpp b/exercises/19_class_virtual_destruct/main.cpp
index cdd54f74f..8d66e8edf 100644
--- a/exercises/19_class_virtual_destruct/main.cpp
+++ b/exercises/19_class_virtual_destruct/main.cpp
@@ -3,14 +3,15 @@
 // READ: 静态字段 <https://zh.cppreference.com/w/cpp/language/static>
 // READ: 虚析构函数 <https://zh.cppreference.com/w/cpp/language/destructor>
 
+
 struct A {
     // TODO: 正确初始化静态字段
-    static int num_a = 0;
+    static int num_a;
 
     A() {
         ++num_a;
     }
-    ~A() {
+    virtual ~A() {
         --num_a;
     }
 
@@ -20,12 +21,12 @@ struct A {
 };
 struct B final : public A {
     // TODO: 正确初始化静态字段
-    static int num_b = 0;
+    static int num_b;
 
     B() {
         ++num_b;
     }
-    ~B() {
+    ~B() override {
         --num_b;
     }
 
@@ -34,13 +35,16 @@ struct B final : public A {
     }
 };
 
+int A::num_a = 0;
+int B::num_b = 0;
+
 int main(int argc, char **argv) {
     auto a = new A;
     auto b = new B;
-    ASSERT(A::num_a == ?, "Fill in the correct value for A::num_a");
-    ASSERT(B::num_b == ?, "Fill in the correct value for B::num_b");
-    ASSERT(a->name() == '?', "Fill in the correct value for a->name()");
-    ASSERT(b->name() == '?', "Fill in the correct value for b->name()");
+    ASSERT(A::num_a == 2, "Fill in the correct value for A::num_a");
+    ASSERT(B::num_b == 1, "Fill in the correct value for B::num_b");
+    ASSERT(a->name() == 'A', "Fill in the correct value for a->name()");
+    ASSERT(b->name() == 'B', "Fill in the correct value for b->name()");
 
     delete a;
     delete b;
@@ -48,13 +52,14 @@ int main(int argc, char **argv) {
     ASSERT(B::num_b == 0, "Every B was destroyed");
 
     A *ab = new B;// 派生类指针可以随意转换为基类指针
-    ASSERT(A::num_a == ?, "Fill in the correct value for A::num_a");
-    ASSERT(B::num_b == ?, "Fill in the correct value for B::num_b");
-    ASSERT(ab->name() == '?', "Fill in the correct value for ab->name()");
+    ASSERT(A::num_a == 1, "Fill in the correct value for A::num_a");
+    ASSERT(B::num_b == 1, "Fill in the correct value for B::num_b");
+    ASSERT(ab->name() == 'B', "Fill in the correct value for ab->name()");
 
     // TODO: 基类指针无法随意转换为派生类指针，补全正确的转换语句
-    B &bb = *ab;
-    ASSERT(bb.name() == '?', "Fill in the correct value for bb->name()");
+    //B &bb = *ab;
+    B &bb = dynamic_cast<B &>(*ab);
+    ASSERT(bb.name() == 'B', "Fill in the correct value for bb->name()");
 
     // TODO: ---- 以下代码不要修改，通过改正类定义解决编译问题 ----
     delete ab;// 通过指针可以删除指向的对象，即使是多态对象
diff --git a/exercises/20_function_template/main.cpp b/exercises/20_function_template/main.cpp
index cb6d978d3..7196e8930 100644
--- a/exercises/20_function_template/main.cpp
+++ b/exercises/20_function_template/main.cpp
@@ -2,7 +2,8 @@
 
 // READ: 函数模板 <https://zh.cppreference.com/w/cpp/language/function_template>
 // TODO: 将这个函数模板化
-int plus(int a, int b) {
+template<typename T>
+T plus(T a, T b) {
     return a + b;
 }
 
@@ -14,7 +15,7 @@ int main(int argc, char **argv) {
     ASSERT(plus(1.25f, 2.5f) == 3.75f, "Plus two float");
     ASSERT(plus(1.25, 2.5) == 3.75, "Plus two double");
     // TODO: 修改判断条件使测试通过
-    ASSERT(plus(0.1, 0.2) == 0.3, "How to make this pass?");
+    ASSERT(std::abs(plus(0.1, 0.2) - 0.3) < 1e-9, "How to make this pass?");
 
     return 0;
 }
diff --git a/exercises/21_runtime_datatype/main.cpp b/exercises/21_runtime_datatype/main.cpp
index 9c4bf376a..153ab5b2d 100644
--- a/exercises/21_runtime_datatype/main.cpp
+++ b/exercises/21_runtime_datatype/main.cpp
@@ -18,13 +18,22 @@ struct TaggedUnion {
 };
 
 // TODO: 将这个函数模板化用于 sigmoid_dyn
-float sigmoid(float x) {
+template<typename T>
+T sigmoid(T x) {
     return 1 / (1 + std::exp(-x));
 }
 
 TaggedUnion sigmoid_dyn(TaggedUnion x) {
     TaggedUnion ans{x.type};
     // TODO: 根据 type 调用 sigmoid
+    if (DataType::Float == x.type)
+    {
+        ans.f = sigmoid(x.f);
+    }
+    else
+    {
+        ans.d = sigmoid(x.d);
+    }
     return ans;
 }
 
diff --git a/exercises/22_class_template/main.cpp b/exercises/22_class_template/main.cpp
index d4985d904..890d7cb7c 100644
--- a/exercises/22_class_template/main.cpp
+++ b/exercises/22_class_template/main.cpp
@@ -1,8 +1,12 @@
 #include "../exercise.h"
 #include <cstring>
 // READ: 类模板 <https://zh.cppreference.com/w/cpp/language/class_template>
+#include <cassert>
+
+
+template<typename T>
+// Tensor4D(const unsigned int[4], const T *) -> Tensor4D<T>;
 
-template<class T>
 struct Tensor4D {
     unsigned int shape[4];
     T *data;
@@ -10,6 +14,10 @@ struct Tensor4D {
     Tensor4D(unsigned int const shape_[4], T const *data_) {
         unsigned int size = 1;
         // TODO: 填入正确的 shape 并计算 size
+        for (int i = 0; i < 4; ++i) {
+            shape[i] = shape_[i];
+            size *= shape[i];
+        }
         data = new T[size];
         std::memcpy(data, data_, size * sizeof(T));
     }
@@ -28,6 +36,36 @@ struct Tensor4D {
     // 则 `this` 与 `others` 相加时，3 个形状为 `[1, 2, 1, 4]` 的子张量各自与 `others` 对应项相加。
     Tensor4D &operator+=(Tensor4D const &others) {
         // TODO: 实现单向广播的加法
+        unsigned int total = 1;
+        for (int i = 0; i < 4; ++i) {
+            // others维度要么等于当前维度，要么为1（广播）
+            assert((others.shape[i] == shape[i]) || (others.shape[i] == 1));
+            total *= shape[i];
+        }
+
+        for (unsigned int i = 0; i < total; ++i) {
+            // 将一维索引i解码成4维索引
+            unsigned int index[4];
+            unsigned int stride = total;
+            unsigned int offset = i;
+            for (int d = 0; d < 4; ++d) {
+                stride /= shape[d];
+                index[d] = offset / stride;
+                offset %= stride;
+            }
+
+            // 计算others中对应的广播后索引
+            unsigned int other_index = 0;
+            unsigned int other_stride = 1;
+            for (int d = 3; d >= 0; --d) {
+                unsigned int idx = (others.shape[d] == 1) ? 0 : index[d];
+                other_index += idx * other_stride;
+                other_stride *= others.shape[d];
+            }
+
+            data[i] += others.data[other_index];
+        }
+
         return *this;
     }
 };
diff --git a/exercises/23_template_const/main.cpp b/exercises/23_template_const/main.cpp
index e0105e168..f2f49c3d1 100644
--- a/exercises/23_template_const/main.cpp
+++ b/exercises/23_template_const/main.cpp
@@ -11,6 +11,10 @@ struct Tensor {
     Tensor(unsigned int const shape_[N]) {
         unsigned int size = 1;
         // TODO: 填入正确的 shape 并计算 size
+        for (unsigned int i = 0; i < N; ++i) {
+            shape[i] = shape_[i];
+            size *= shape[i];
+        }
         data = new T[size];
         std::memset(data, 0, size * sizeof(T));
     }
@@ -20,7 +24,13 @@ struct Tensor {
 
     // 为了保持简单，禁止复制和移动
     Tensor(Tensor const &) = delete;
-    Tensor(Tensor &&) noexcept = delete;
+    //Tensor(Tensor&&) noexcept = delete;
+    Tensor(Tensor &&other) noexcept {
+        std::memcpy(shape, other.shape, sizeof(shape));
+        data = other.data;
+        other.data = nullptr;// 转移所有权
+    }
+
 
     T &operator[](unsigned int const indices[N]) {
         return data[data_index(indices)];
@@ -32,10 +42,18 @@ struct Tensor {
 private:
     unsigned int data_index(unsigned int const indices[N]) const {
         unsigned int index = 0;
+        unsigned int stride[N];
+        stride[N - 1] = 1;
+        for (int i = N - 2; i >= 0; --i) {
+            stride[i] = stride[i + 1] * shape[i + 1];
+        }
+
+        // 再正序遍历，用 stride 乘索引累加
         for (unsigned int i = 0; i < N; ++i) {
             ASSERT(indices[i] < shape[i], "Invalid index");
-            // TODO: 计算 index
+            index += indices[i] * stride[i];
         }
+
         return index;
     }
 };
diff --git a/exercises/24_std_array/main.cpp b/exercises/24_std_array/main.cpp
index c29718d9d..8ac430609 100644
--- a/exercises/24_std_array/main.cpp
+++ b/exercises/24_std_array/main.cpp
@@ -8,21 +8,22 @@
 int main(int argc, char **argv) {
     {
         std::array<int, 5> arr{{1, 2, 3, 4, 5}};
-        ASSERT(arr.size() == ?, "Fill in the correct value.");
-        ASSERT(sizeof(arr) == ?, "Fill in the correct value.");
+        ASSERT(arr.size() == 5, "Fill in the correct value.");
+        ASSERT(sizeof(arr) == 20, "Fill in the correct value.");
         int ans[]{1, 2, 3, 4, 5};
-        ASSERT(std::memcmp(arr.?, ans, ?) == 0, "Fill in the correct values.");
+        ASSERT(std::memcmp(arr.data(), ans, sizeof(ans)) == 0, "Fill in the correct values.");
     }
     {
         std::array<double, 8> arr;
-        ASSERT(arr.size() == ?, "Fill in the correct value.");
-        ASSERT(sizeof(arr) == ?, "Fill in the correct value.");
+        ASSERT(arr.size() == 8, "Fill in the correct value.");
+        ASSERT(sizeof(arr) == 64, "Fill in the correct value.");
     }
     {
         std::array<char, 21> arr{"Hello, InfiniTensor!"};
-        ASSERT(arr.size() == ?, "Fill in the correct value.");
-        ASSERT(sizeof(arr) == ?, "Fill in the correct value.");
-        ASSERT(std::strcmp(arr.?, "Hello, InfiniTensor!") == 0, "Fill in the correct value.");
+        ASSERT(arr.size() == 21, "Fill in the correct value.");
+        //std::cout << sizeof(char) << std::endl;
+        ASSERT(sizeof(arr) == 21, "Fill in the correct value.");
+        ASSERT(std::strcmp(arr.data(), "Hello, InfiniTensor!") == 0, "Fill in the correct value.");
     }
     return 0;
 }
diff --git a/exercises/25_std_vector/main.cpp b/exercises/25_std_vector/main.cpp
index f9e41bb78..3872b36d5 100644
--- a/exercises/25_std_vector/main.cpp
+++ b/exercises/25_std_vector/main.cpp
@@ -5,84 +5,84 @@
 // READ: std::vector <https://zh.cppreference.com/w/cpp/container/vector>
 
 // TODO: 将下列 `?` 替换为正确的代码
-int main(int argc, char **argv) {
+int main(int argc, char** argv) {
     {
-        std::vector<int> vec{1, 2, 3, 4, 5};
-        ASSERT(vec.size() == ?, "Fill in the correct value.");
+        std::vector<int> vec{ 1, 2, 3, 4, 5 };
+        ASSERT(vec.size() == 5, "Fill in the correct value.");
         // THINK: `std::vector` 的大小是什么意思？与什么有关？
-        ASSERT(sizeof(vec) == ?, "Fill in the correct value.");
-        int ans[]{1, 2, 3, 4, 5};
-        ASSERT(std::memcmp(vec.?, ans, sizeof(ans)) == 0, "Fill in the correct values.");
+        ASSERT(sizeof(vec) == sizeof(std::vector<int>), "Fill in the correct value.");
+        int ans[]{ 1, 2, 3, 4, 5 };
+        ASSERT(std::memcmp(vec.data(), ans, sizeof(ans)) == 0, "Fill in the correct values.");
     }
     {
-        std::vector<double> vec{1, 2, 3, 4, 5};
+        std::vector<double> vec{ 1, 2, 3, 4, 5 };
         {
-            ASSERT(vec.size() == ?, "Fill in the correct value.");
-            ASSERT(sizeof(vec) == ?, "Fill in the correct value.");
-            double ans[]{1, 2, 3, 4, 5};
-            ASSERT(std::memcmp(vec.?, ans, sizeof(ans)) == 0, "Fill in the correct values.");
+            ASSERT(vec.size() == 5, "Fill in the correct value.");
+            ASSERT(sizeof(vec) == sizeof(std::vector<double>), "Fill in the correct value.");
+            double ans[]{ 1, 2, 3, 4, 5 };
+            ASSERT(std::memcmp(vec.data(), ans, sizeof(ans)) == 0, "Fill in the correct values.");
         }
         {
             vec.push_back(6);
-            ASSERT(vec.size() == ?, "Fill in the correct value.");
-            ASSERT(sizeof(vec) == ?, "Fill in the correct value.");
+            ASSERT(vec.size() == 6, "Fill in the correct value.");
+            ASSERT(sizeof(vec) == sizeof(std::vector<double>), "Fill in the correct value.");
             vec.pop_back();
-            ASSERT(vec.size() == ?, "Fill in the correct value.");
-            ASSERT(sizeof(vec) == ?, "Fill in the correct value.");
+            ASSERT(vec.size() == 5, "Fill in the correct value.");
+            ASSERT(sizeof(vec) == sizeof(std::vector<double>), "Fill in the correct value.");
         }
         {
             vec[4] = 6;
-            ASSERT(vec[0] == ?, "Fill in the correct value.");
-            ASSERT(vec[1] == ?, "Fill in the correct value.");
-            ASSERT(vec[2] == ?, "Fill in the correct value.");
-            ASSERT(vec[3] == ?, "Fill in the correct value.");
-            ASSERT(vec[4] == ?, "Fill in the correct value.");
+            ASSERT(vec[0] == 1, "Fill in the correct value.");
+            ASSERT(vec[1] == 2, "Fill in the correct value.");
+            ASSERT(vec[2] == 3, "Fill in the correct value.");
+            ASSERT(vec[3] == 4, "Fill in the correct value.");
+            ASSERT(vec[4] == 6, "Fill in the correct value.");
         }
         {
             // THINK: `std::vector` 插入删除的时间复杂度是什么？
-            vec.insert(?, 1.5);
+            vec.insert(vec.begin() + 1, 1.5);
             ASSERT((vec == std::vector<double>{1, 1.5, 2, 3, 4, 6}), "Make this assertion pass.");
-            vec.erase(?);
+            vec.erase(vec.begin() + 3);
             ASSERT((vec == std::vector<double>{1, 1.5, 2, 4, 6}), "Make this assertion pass.");
         }
         {
             vec.shrink_to_fit();
-            ASSERT(vec.capacity() == ?, "Fill in the correct value.");
+            ASSERT(vec.capacity() == 5, "Fill in the correct value.");
             vec.clear();
             ASSERT(vec.empty(), "`vec` is empty now.");
-            ASSERT(vec.size() == ?, "Fill in the correct value.");
-            ASSERT(vec.capacity() == ?, "Fill in the correct value.");
+            ASSERT(vec.size() == 0, "Fill in the correct value.");
+            ASSERT(vec.capacity() == 5, "Fill in the correct value.");
         }
     }
     {
-        std::vector<char> vec(?, ?); // TODO: 调用正确的构造函数
+        std::vector<char> vec(48, 'z');// TODO: 调用正确的构造函数
         ASSERT(vec[0] == 'z', "Make this assertion pass.");
         ASSERT(vec[47] == 'z', "Make this assertion pass.");
         ASSERT(vec.size() == 48, "Make this assertion pass.");
-        ASSERT(sizeof(vec) == ?, "Fill in the correct value.");
+        ASSERT(sizeof(vec) == sizeof(std::vector<char>), "Fill in the correct value.");
         {
             auto capacity = vec.capacity();
             vec.resize(16);
-            ASSERT(vec.size() == ?, "Fill in the correct value.");
-            ASSERT(vec.capacity() == ?, "Fill in a correct identifier.");
+            ASSERT(vec.size() == 16, "Fill in the correct value.");
+            ASSERT(vec.capacity() == capacity, "Fill in a correct identifier.");
         }
         {
             vec.reserve(256);
-            ASSERT(vec.size() == ?, "Fill in the correct value.");
-            ASSERT(vec.capacity() == ?, "Fill in the correct value.");
+            ASSERT(vec.size() == 16, "Fill in the correct value.");
+            ASSERT(vec.capacity() == 256, "Fill in the correct value.");
         }
         {
             vec.push_back('a');
             vec.push_back('b');
             vec.push_back('c');
             vec.push_back('d');
-            ASSERT(vec.size() == ?, "Fill in the correct value.");
-            ASSERT(vec.capacity() == ?, "Fill in the correct value.");
-            ASSERT(vec[15] == ?, "Fill in the correct value.");
-            ASSERT(vec[?] == 'a', "Fill in the correct value.");
-            ASSERT(vec[?] == 'b', "Fill in the correct value.");
-            ASSERT(vec[?] == 'c', "Fill in the correct value.");
-            ASSERT(vec[?] == 'd', "Fill in the correct value.");
+            ASSERT(vec.size() == 20, "Fill in the correct value.");
+            ASSERT(vec.capacity() == 256, "Fill in the correct value.");
+            ASSERT(vec[15] == 'z', "Fill in the correct value.");
+            ASSERT(vec[16] == 'a', "Fill in the correct value.");
+            ASSERT(vec[17] == 'b', "Fill in the correct value.");
+            ASSERT(vec[18] == 'c', "Fill in the correct value.");
+            ASSERT(vec[19] == 'd', "Fill in the correct value.");
         }
     }
 
diff --git a/exercises/26_std_vector_bool/main.cpp b/exercises/26_std_vector_bool/main.cpp
index b4ab4f9c4..972a80b08 100644
--- a/exercises/26_std_vector_bool/main.cpp
+++ b/exercises/26_std_vector_bool/main.cpp
@@ -6,29 +6,29 @@
 
 // TODO: 将下列 `?` 替换为正确的代码
 int main(int argc, char **argv) {
-    std::vector<bool> vec(?, ?);// TODO: 正确调用构造函数
+    std::vector<bool> vec(100, true);// TODO: 正确调用构造函数
     ASSERT(vec[0], "Make this assertion pass.");
     ASSERT(vec[99], "Make this assertion pass.");
     ASSERT(vec.size() == 100, "Make this assertion pass.");
     // NOTICE: 平台相关！注意 CI:Ubuntu 上的值。
     std::cout << "sizeof(std::vector<bool>) = " << sizeof(std::vector<bool>) << std::endl;
-    ASSERT(sizeof(vec) == ?, "Fill in the correct value.");
+    ASSERT(sizeof(vec) == sizeof(std::vector<bool>), "Fill in the correct value.");
     {
         vec[20] = false;
-        ASSERT(?vec[20], "Fill in `vec[20]` or `!vec[20]`.");
+        ASSERT(!vec[20], "Fill in `vec[20]` or `!vec[20]`.");
     }
     {
         vec.push_back(false);
-        ASSERT(vec.size() == ?, "Fill in the correct value.");
-        ASSERT(?vec[100], "Fill in `vec[100]` or `!vec[100]`.");
+        ASSERT(vec.size() == 101, "Fill in the correct value.");
+        ASSERT(!vec[100], "Fill in `vec[100]` or `!vec[100]`.");
     }
     {
         auto ref = vec[30];
-        ASSERT(?ref, "Fill in `ref` or `!ref`");
+        ASSERT(ref, "Fill in `ref` or `!ref`");
         ref = false;
-        ASSERT(?ref, "Fill in `ref` or `!ref`");
+        ASSERT(!ref, "Fill in `ref` or `!ref`");
         // THINK: WHAT and WHY?
-        ASSERT(?vec[30], "Fill in `vec[30]` or `!vec[30]`.");
+        ASSERT(!vec[30], "Fill in `vec[30]` or `!vec[30]`.");
     }
     return 0;
 }
diff --git a/exercises/27_strides/main.cpp b/exercises/27_strides/main.cpp
index baceaf2a9..13feaf5ae 100644
--- a/exercises/27_strides/main.cpp
+++ b/exercises/27_strides/main.cpp
@@ -18,6 +18,13 @@ std::vector<udim> strides(std::vector<udim> const &shape) {
     // TODO: 完成函数体，根据张量形状计算张量连续存储时的步长。
     // READ: 逆向迭代器 std::vector::rbegin <https://zh.cppreference.com/w/cpp/container/vector/rbegin>
     //       使用逆向迭代器可能可以简化代码
+    if (shape.empty()) return strides;
+
+    strides.back() = 1;// 最后一维步长为1
+    for (int i = (int) shape.size() - 2; i >= 0; --i) {
+        strides[i] = shape[i + 1] * strides[i + 1];
+    }
+
     return strides;
 }
 
diff --git a/exercises/28_std_string/main.cpp b/exercises/28_std_string/main.cpp
index d8b276274..5fcfdf31d 100644
--- a/exercises/28_std_string/main.cpp
+++ b/exercises/28_std_string/main.cpp
@@ -10,9 +10,9 @@ int main(int argc, char **argv) {
     auto world = "world";
     // READ: `decltype` 表达式 <https://zh.cppreference.com/w/cpp/language/decltype>
     // READ: `std::is_same_v` 元编程判别 <https://zh.cppreference.com/w/cpp/types/is_same>
-    ASSERT((std::is_same_v<decltype(hello), ?>), "Fill in the missing type.");
-    ASSERT((std::is_same_v<decltype(world), ?>), "Fill in the missing type.");
+    ASSERT((std::is_same_v<decltype(hello), std::string>), "Fill in the missing type.");
+    ASSERT((std::is_same_v<decltype(world), const char*>), "Fill in the missing type.");
     // TODO: 将 `?` 替换为正确的字符串
-    ASSERT(hello + ", " + world + '!' == "?", "Fill in the missing string.");
+    ASSERT(hello + ", " + world + '!' == "Hello, world!", "Fill in the missing string.");
     return 0;
 }
diff --git a/exercises/29_std_map/main.cpp b/exercises/29_std_map/main.cpp
index fcccca347..68122a5fc 100644
--- a/exercises/29_std_map/main.cpp
+++ b/exercises/29_std_map/main.cpp
@@ -7,11 +7,13 @@
 template<class k, class v>
 bool key_exists(std::map<k, v> const &map, k const &key) {
     // TODO: 实现函数
+    return map.find(key) != map.end();
 }
 
 template<class k, class v>
 void set(std::map<k, v> &map, k key, v value) {
     // TODO: 实现函数
+    map[key] = value;
 }
 
 // ---- 不要修改以下代码 ----
diff --git a/exercises/30_std_unique_ptr/main.cpp b/exercises/30_std_unique_ptr/main.cpp
index 9b98b5794..d9489ac7b 100644
--- a/exercises/30_std_unique_ptr/main.cpp
+++ b/exercises/30_std_unique_ptr/main.cpp
@@ -44,17 +44,19 @@ int main(int argc, char **argv) {
     forward(drop(reset(forward(forward(reset(nullptr))))));
     problems[1] = std::move(RECORDS);
 
+
     drop(drop(reset(drop(reset(reset(nullptr))))));
     problems[2] = std::move(RECORDS);
 
+
     // ---- 不要修改以上代码 ----
 
     std::vector<const char *> answers[]{
         {"fd"},
         // TODO: 分析 problems[1] 中资源的生命周期，将记录填入 `std::vector`
         // NOTICE: 此题结果依赖对象析构逻辑，平台相关，提交时以 CI 实际运行平台为准
-        {"", "", "", "", "", "", "", ""},
-        {"", "", "", "", "", "", "", ""},
+        {"ffr", "d"},
+        {"r", "d", "d"},
     };
 
     // ---- 不要修改以下代码 ----
diff --git a/exercises/31_std_shared_ptr/main.cpp b/exercises/31_std_shared_ptr/main.cpp
index febbbcc6f..d35134457 100644
--- a/exercises/31_std_shared_ptr/main.cpp
+++ b/exercises/31_std_shared_ptr/main.cpp
@@ -10,36 +10,36 @@ int main(int argc, char **argv) {
     std::shared_ptr<int> ptrs[]{shared, shared, shared};
 
     std::weak_ptr<int> observer = shared;
-    ASSERT(observer.use_count() == ?, "");
+    ASSERT(observer.use_count() == 4, "");
 
     ptrs[0].reset();
-    ASSERT(observer.use_count() == ?, "");
+    ASSERT(observer.use_count() == 3, "");
 
     ptrs[1] = nullptr;
-    ASSERT(observer.use_count() == ?, "");
+    ASSERT(observer.use_count() == 2, "");
 
     ptrs[2] = std::make_shared<int>(*shared);
-    ASSERT(observer.use_count() == ?, "");
+    ASSERT(observer.use_count() == 1, "");
 
     ptrs[0] = shared;
     ptrs[1] = shared;
     ptrs[2] = std::move(shared);
-    ASSERT(observer.use_count() == ?, "");
+    ASSERT(observer.use_count() == 3, "");
 
     std::ignore = std::move(ptrs[0]);
     ptrs[1] = std::move(ptrs[1]);
     ptrs[1] = std::move(ptrs[2]);
-    ASSERT(observer.use_count() == ?, "");
+    ASSERT(observer.use_count() == 2, "");
 
     shared = observer.lock();
-    ASSERT(observer.use_count() == ?, "");
+    ASSERT(observer.use_count() == 3, "");
 
     shared = nullptr;
     for (auto &ptr : ptrs) ptr = nullptr;
-    ASSERT(observer.use_count() == ?, "");
+    ASSERT(observer.use_count() == 0, "");
 
     shared = observer.lock();
-    ASSERT(observer.use_count() == ?, "");
+    ASSERT(observer.use_count() == 0, "");
 
     return 0;
 }
diff --git a/exercises/32_std_transform/main.cpp b/exercises/32_std_transform/main.cpp
index f4dc25a5c..ad386804c 100644
--- a/exercises/32_std_transform/main.cpp
+++ b/exercises/32_std_transform/main.cpp
@@ -6,10 +6,16 @@
 // READ: `std::transform` <https://zh.cppreference.com/w/cpp/algorithm/transform>
 // READ: `std::vector::begin` <https://zh.cppreference.com/w/cpp/container/vector/begin>
 
+
+
 int main(int argc, char **argv) {
     std::vector<int> val{8, 13, 21, 34, 55};
     // TODO: 调用 `std::transform`，将 `v` 中的每个元素乘以 2，并转换为字符串，存入 `ans`
     // std::vector<std::string> ans
+    std::vector<std::string> ans(val.size());
+    std::transform(val.begin(), val.end(), ans.begin(), [](int x) {
+        return std::to_string(x * 2);
+    });
     ASSERT(ans.size() == val.size(), "ans size should be equal to val size");
     ASSERT(ans[0] == "16", "ans[0] should be 16");
     ASSERT(ans[1] == "26", "ans[1] should be 26");
diff --git a/exercises/33_std_accumulate/main.cpp b/exercises/33_std_accumulate/main.cpp
index 6326929d5..53e9af330 100644
--- a/exercises/33_std_accumulate/main.cpp
+++ b/exercises/33_std_accumulate/main.cpp
@@ -12,6 +12,8 @@ int main(int argc, char **argv) {
     //       - 连续存储；
     //       的张量占用的字节数
     // int size =
+    int size = sizeof(DataType) * std::accumulate(
+                                      shape, shape + sizeof(shape) / sizeof(int), 1, std::multiplies<int>());
     ASSERT(size == 602112, "4x1x3x224x224 = 602112");
     return 0;
 }
diff --git a/exercises/exercises/00_hello_world/main.cpp b/exercises/exercises/00_hello_world/main.cpp
new file mode 100644
index 000000000..eb1a725a7
--- /dev/null
+++ b/exercises/exercises/00_hello_world/main.cpp
@@ -0,0 +1,11 @@
+#include "../exercise.h"
+
+// READ: std streams <https://zh.cppreference.com/w/cpp/io/c/std_streams>
+// READ: 流修饰符 <https://zh.cppreference.com/w/cpp/io/manip>
+// READ: format in cxx20 <https://zh.cppreference.com/w/cpp/utility/format/format>
+
+int main(int argc, char **argv) {
+    // TODO: 在控制台输出 "Hello, InfiniTensor!" 并换行
+    std::cout<< "Hello, InfiniTensor!" << std::endl;
+    return 0;
+}
diff --git a/exercises/exercises/01_variable&add/main.cpp b/exercises/exercises/01_variable&add/main.cpp
new file mode 100644
index 000000000..fb05c57f3
--- /dev/null
+++ b/exercises/exercises/01_variable&add/main.cpp
@@ -0,0 +1,11 @@
+#include "../exercise.h"
+
+// READ: 运算符 <https://zh.cppreference.com/w/cpp/language/expressions#.E8.BF.90.E7.AE.97.E7.AC.A6>
+
+int main(int argc, char **argv) {
+    // TODO: 补全变量定义并打印加法运算
+    // x ?
+    int x = 10;
+    std::cout << x << " + " << x << " = " << x + x << std::endl;
+    return 0;
+}
diff --git a/exercises/exercises/02_function/main.cpp b/exercises/exercises/02_function/main.cpp
new file mode 100644
index 000000000..603a9ec04
--- /dev/null
+++ b/exercises/exercises/02_function/main.cpp
@@ -0,0 +1,21 @@
+#include "../exercise.h"
+
+// READ: 声明 <https://zh.cppreference.com/w/cpp/language/declarations>
+// NOTICE: cppreference 中的示例中指出了复杂声明的解读法，建议认真阅读。
+// NOTICE: 补充由内而外读法的机翻解释 <https://learn.microsoft.com/zh-cn/cpp/c-language/interpreting-more-complex-declarators?view=msvc-170>
+
+// TODO: 在这里声明函数
+int add(int a, int b);
+
+int main(int argc, char **argv) {
+    ASSERT(add(123, 456) == 123 + 456, "add(123, 456) should be 123 + 456");
+
+    auto x = 1, y = 2;
+    std::cout << x << " + " << y << " = " << add(x, y) << std::endl;
+    return 0;
+}
+
+int add(int a, int b) {
+    // TODO: 补全函数定义，但不要移动代码行
+    return a+b;
+}
diff --git a/exercises/exercises/03_argument&parameter/main.cpp b/exercises/exercises/03_argument&parameter/main.cpp
new file mode 100644
index 000000000..76136eae6
--- /dev/null
+++ b/exercises/exercises/03_argument&parameter/main.cpp
@@ -0,0 +1,26 @@
+#include "../exercise.h"
+
+// READ: <https://stackoverflow.com/questions/156767/whats-the-difference-between-an-argument-and-a-parameter>
+// THINK: 参数都有哪些传递方式？如何选择传递方式？
+
+void func(int);
+
+// TODO: 为下列 ASSERT 填写正确的值
+int main(int argc, char **argv) {
+    auto arg = 99;
+    ASSERT(arg == 99, "arg should be ?");
+    std::cout << "befor func call: " << arg << std::endl;
+    func(arg);
+    ASSERT(arg == 99, "arg should be ?");
+    std::cout << "after func call: " << arg << std::endl;
+    return 0;
+}
+
+// TODO: 为下列 ASSERT 填写正确的值
+void func(int param) {
+    ASSERT(param == 99, "param should be ?");
+    std::cout << "befor add: " << param << std::endl;
+    param += 1;
+    ASSERT(param == 100, "param should be ?");
+    std::cout << "after add: " << param << std::endl;
+}
diff --git a/exercises/exercises/04_static/main.cpp b/exercises/exercises/04_static/main.cpp
new file mode 100644
index 000000000..64f417694
--- /dev/null
+++ b/exercises/exercises/04_static/main.cpp
@@ -0,0 +1,19 @@
+#include "../exercise.h"
+
+// READ: `static` 关键字 <https://zh.cppreference.com/w/cpp/language/storage_duration>
+// THINK: 这个函数的两个 `static` 各自的作用是什么？
+static int func(int param) {
+    static int static_ = param;
+    // std::cout << "static_ = " << static_ << std::endl;
+    return static_++;
+}
+
+int main(int argc, char **argv) {
+    // TODO: 将下列 `?` 替换为正确的数字
+    ASSERT(func(5) == 5, "static variable value incorrect");
+    ASSERT(func(4) == 6, "static variable value incorrect");
+    ASSERT(func(3) == 7, "static variable value incorrect");
+    ASSERT(func(2) == 8, "static variable value incorrect");
+    ASSERT(func(1) == 9, "static variable value incorrect");
+    return 0;
+}
diff --git a/exercises/exercises/05_constexpr/main.cpp b/exercises/exercises/05_constexpr/main.cpp
new file mode 100644
index 000000000..2f6a5451a
--- /dev/null
+++ b/exercises/exercises/05_constexpr/main.cpp
@@ -0,0 +1,28 @@
+#include "../exercise.h"
+
+constexpr unsigned long long fibonacci(int i) {
+    switch (i) {
+        case 0:
+            return 0;
+        case 1:
+            return 1;
+        default:
+            return fibonacci(i - 1) + fibonacci(i - 2);
+    }
+}
+
+int main(int argc, char **argv) {
+    constexpr auto FIB20 = fibonacci(20);
+    ASSERT(FIB20 == 6765, "fibonacci(20) should be 6765");
+    std::cout << "fibonacci(20) = " << FIB20 << std::endl;
+
+    // TODO: 观察错误信息，修改一处，使代码编译运行
+    // PS: 编译运行，但是不一定能算出结果……
+    constexpr auto ANS_N = 9;
+    //constexpr auto ANS = fibonacci(ANS_N);
+    auto ANS = fibonacci(ANS_N);
+
+    std::cout << "fibonacci(" << ANS_N << ") = " << ANS << std::endl;
+
+    return 0;
+}
diff --git a/exercises/exercises/06_array/main.cpp b/exercises/exercises/06_array/main.cpp
new file mode 100644
index 000000000..bbce324cf
--- /dev/null
+++ b/exercises/exercises/06_array/main.cpp
@@ -0,0 +1,26 @@
+#include "../exercise.h"
+
+// READ: 数组 <https://zh.cppreference.com/w/cpp/language/array>
+
+unsigned long long arr[90]{0, 1};
+unsigned long long fibonacci(int i) {
+    switch (i) {
+        case 0:
+            return 0;
+        case 1:
+            return 1;
+        default:
+            // TODO: 补全三目表达式缺失的部分
+            return arr[i] ? arr[i] : (arr[i] = fibonacci(i - 1) + fibonacci(i - 2));
+    }
+}
+
+int main(int argc, char **argv) {
+    // TODO: 为此 ASSERT 填写正确的值
+    ASSERT(sizeof(arr) == 720, "sizeof array is size of all its elements");
+    // ---- 不要修改以下代码 ----
+    ASSERT(fibonacci(2) == 1, "fibonacci(2) should be 1");
+    ASSERT(fibonacci(20) == 6765, "fibonacci(20) should be 6765");
+    ASSERT(fibonacci(80) == 23416728348467685, "fibonacci(80) should be 23416728348467685");
+    return 0;
+}
diff --git a/exercises/exercises/07_loop/main.cpp b/exercises/exercises/07_loop/main.cpp
new file mode 100644
index 000000000..ed485308b
--- /dev/null
+++ b/exercises/exercises/07_loop/main.cpp
@@ -0,0 +1,30 @@
+#include "../exercise.h"
+
+// TODO: 改正函数实现，实现正确的缓存优化斐波那契计算
+// THINk: 这个函数是一个纯函数（pure function）吗？
+// READ: 纯函数 <https://zh.wikipedia.org/wiki/%E7%BA%AF%E5%87%BD%E6%95%B0>
+static unsigned long long fibonacci(int i) {
+    // TODO: 为缓存设置正确的初始值
+    static unsigned long long cache[96]{0, 1};
+    static int cached = 2;
+    
+    // TODO: 设置正确的循环条件
+    for (; cached<=i; ++cached) {
+        cache[cached] = cache[cached - 1] + cache[cached - 2];
+    }
+    return cache[i];
+}
+
+// ---- 不要修改以下代码 ----
+int main(int argc, char **argv) {
+    ASSERT(fibonacci(0) == 0, "fibonacci(0) should be 0");
+    ASSERT(fibonacci(1) == 1, "fibonacci(1) should be 1");
+    ASSERT(fibonacci(2) == 1, "fibonacci(2) should be 1");
+    ASSERT(fibonacci(3) == 2, "fibonacci(3) should be 2");
+    ASSERT(fibonacci(10) == 55, "fibonacci(10) should be 55");
+
+    auto fib90 = fibonacci(90);
+    std::cout << "fibonacci(90) = " << fib90 << std::endl;
+    ASSERT(fib90 == 2880067194370816120, "fibonacci(90) should be 2880067194370816120");
+    return 0;
+}
diff --git a/exercises/exercises/08_pointer/main.cpp b/exercises/exercises/08_pointer/main.cpp
new file mode 100644
index 000000000..622a2fd87
--- /dev/null
+++ b/exercises/exercises/08_pointer/main.cpp
@@ -0,0 +1,39 @@
+#include "../exercise.h"
+
+// READ: 数组向指针退化 <https://zh.cppreference.com/w/cpp/language/array#%E6%95%B0%E7%BB%84%E5%88%B0%E6%8C%87%E9%92%88%E7%9A%84%E9%80%80%E5%8C%96>
+bool is_fibonacci(int *ptr, int len, int stride) {
+    ASSERT(len >= 3, "`len` should be at least 3");
+    // TODO: 编写代码判断从 ptr 开始，每 stride 个元素取 1 个元素，组成长度为 n 的数列是否满足
+    for (int i = 0; i <= len - 3; i++)
+    {
+        int a = ptr[i * stride];
+        int b = ptr[(i+1)*stride];
+        int c = ptr[(i+2)*stride];
+        if (a + b != c)
+        {
+            return false;
+        }
+
+    }
+    // arr[i + 2] = arr[i] + arr[i + 1]
+    return true;
+}
+
+// ---- 不要修改以下代码 ----
+int main(int argc, char **argv) {
+    int arr0[]{0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55},
+        arr1[]{0, 1, 2, 3, 4, 5, 6},
+        arr2[]{99, 98, 4, 1, 7, 2, 11, 3, 18, 5, 29, 8, 47, 13, 76, 21, 123, 34, 199, 55, 322, 0, 0};
+    // clang-format off
+    ASSERT( is_fibonacci(arr0    , sizeof(arr0) / sizeof(*arr0)    , 1),         "arr0 is Fibonacci"    );
+    ASSERT( is_fibonacci(arr0 + 2, sizeof(arr0) / sizeof(*arr0) - 4, 1), "part of arr0 is Fibonacci"    );
+    ASSERT(!is_fibonacci(arr1    , sizeof(arr1) / sizeof(*arr1)    , 1),         "arr1 is not Fibonacci");
+    ASSERT( is_fibonacci(arr1 + 1,  3                              , 1), "part of arr1 is Fibonacci"    );
+    ASSERT(!is_fibonacci(arr2    , sizeof(arr2) / sizeof(*arr2)    , 1),         "arr2 is not Fibonacci");
+    ASSERT( is_fibonacci(arr2 + 2, 10                              , 2), "part of arr2 is Fibonacci"    );
+    ASSERT( is_fibonacci(arr2 + 3,  9                              , 2), "part of arr2 is Fibonacci"    );
+    ASSERT(!is_fibonacci(arr2 + 3, 10                              , 2), "guard check"                  );
+    ASSERT(!is_fibonacci(arr2 + 1, 10                              , 2), "guard check"                  );
+    // clang-format on
+    return 0;
+}
diff --git a/exercises/exercises/09_enum&union/main.cpp b/exercises/exercises/09_enum&union/main.cpp
new file mode 100644
index 000000000..5779dd314
--- /dev/null
+++ b/exercises/exercises/09_enum&union/main.cpp
@@ -0,0 +1,51 @@
+#include "../exercise.h"
+
+// READ: 枚举类型 <https://zh.cppreference.com/w/cpp/language/enum>
+
+// `enum` 是 C 的兼容类型，本质上其对应类型的常量。
+// 在 `enum` 中定义标识符等价于定义 constexpr 常量，
+// 这些标识符不需要前缀，可以直接引用。
+// 因此 `enum` 定义会污染命名空间。
+enum ColorEnum : unsigned char {
+    COLOR_RED = 31,
+    COLOR_GREEN,
+    COLOR_YELLOW,
+    COLOR_BLUE,
+};
+
+// 有作用域枚举型是 C++ 引入的类型安全枚举。
+// 其内部标识符需要带前缀引用，如 `Color::Red`。
+// 作用域枚举型可以避免命名空间污染，并提供类型安全保证。
+enum class Color : int {
+    Red = COLOR_RED,
+    Green,
+    Yellow,
+    Blue,
+};
+
+ColorEnum convert_by_pun(Color c) {
+    // READ: <https://zh.cppreference.com/w/cpp/language/union>
+    // `union` 表示在同一内存位置存储的不同类型的值。
+    // 其常见用法是实现类型双关转换，即将一种类型的值转换为另一种无关类型的值。
+    // 但这种写法实际上仅在 C 语言良定义，在 C++ 中是未定义行为。
+    // 这是比较少见的 C++ 不与 C 保持兼容的特性。
+    // READ: 类型双关 <https://tttapa.github.io/Pages/Programming/Cpp/Practices/type-punning.html>
+    union TypePun {
+        ColorEnum e;
+        Color c;
+    };
+
+    TypePun pun;
+    // TODO: 补全类型双关转换
+    pun.c = c;
+
+    return pun.e;
+}
+
+int main(int argc, char **argv) {
+    ASSERT(convert_by_pun(Color::Red) == COLOR_RED, "Type punning conversion");
+    ASSERT(convert_by_pun(Color::Green) == COLOR_GREEN, "Type punning conversion");
+    ASSERT(convert_by_pun(Color::Yellow) == COLOR_YELLOW, "Type punning conversion");
+    ASSERT(convert_by_pun(Color::Blue) == COLOR_BLUE, "Type punning conversion");
+    return 0;
+}
diff --git a/exercises/exercises/10_trivial/main.cpp b/exercises/exercises/10_trivial/main.cpp
new file mode 100644
index 000000000..173ebbdd6
--- /dev/null
+++ b/exercises/exercises/10_trivial/main.cpp
@@ -0,0 +1,30 @@
+#include "../exercise.h"
+
+// READ: Trivial type <https://learn.microsoft.com/zh-cn/cpp/cpp/trivial-standard-layout-and-pod-types?view=msvc-170>
+
+struct FibonacciCache {
+    unsigned long long cache[16];
+    int cached;
+};
+
+// TODO: 实现正确的缓存优化斐波那契计算
+static unsigned long long fibonacci(FibonacciCache &cache, int i) {
+    int cached = cache.cached;
+    for (cached; cached<=i; ++cached) {
+        cache.cache[cached] = cache.cache[cached - 1] + cache.cache[cached - 2];
+    }
+    return cache.cache[i];
+}
+
+int main(int argc, char **argv) {
+    // TODO: 初始化缓存结构体，使计算正确
+    // NOTICE: C/C++ 中，读取未初始化的变量（包括结构体变量）是未定义行为
+    // READ: 初始化的各种写法 <https://zh.cppreference.com/w/cpp/language/initialization>
+    FibonacciCache fib;
+    fib.cache[0] = 0;
+    fib.cache[1] = 1;
+    fib.cached = 2;
+    ASSERT(fibonacci(fib, 10) == 55, "fibonacci(10) should be 55");
+    std::cout << "fibonacci(10) = " << fibonacci(fib, 10) << std::endl;
+    return 0;
+}
diff --git a/exercises/exercises/11_method/main.cpp b/exercises/exercises/11_method/main.cpp
new file mode 100644
index 000000000..57d04651d
--- /dev/null
+++ b/exercises/exercises/11_method/main.cpp
@@ -0,0 +1,25 @@
+#include "../exercise.h"
+
+struct Fibonacci {
+    unsigned long long cache[128];
+    int cached;
+
+    // TODO: 实现正确的缓存优化斐波那契计算
+    unsigned long long get(int i) {
+        for (; cached<=i; ++cached) {
+            cache[cached] = cache[cached - 1] + cache[cached - 2];
+        }
+        return cache[i];
+    }
+};
+
+int main(int argc, char **argv) {
+    // TODO: 初始化缓存结构体，使计算正确
+    Fibonacci fib;
+    fib.cache[0] = 0;
+    fib.cache[1] = 1;
+    fib.cached = 2;
+    ASSERT(fib.get(10) == 55, "fibonacci(10) should be 55");
+    std::cout << "fibonacci(10) = " << fib.get(10) << std::endl;
+    return 0;
+}
diff --git a/exercises/exercises/12_method_const/main.cpp b/exercises/exercises/12_method_const/main.cpp
new file mode 100644
index 000000000..65c079be1
--- /dev/null
+++ b/exercises/exercises/12_method_const/main.cpp
@@ -0,0 +1,18 @@
+#include "../exercise.h"
+
+// READ: 有 cv 限定符的成员函数 <https://zh.cppreference.com/w/cpp/language/member_functions>
+
+struct Fibonacci {
+    int numbers[11];
+    // TODO: 修改方法签名和实现，使测试通过
+    int get(int i) const {
+        return numbers[i];
+    }
+};
+
+int main(int argc, char **argv) {
+    Fibonacci constexpr FIB{{0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55}};
+    ASSERT(FIB.get(10) == 55, "fibonacci(10) should be 55");
+    std::cout << "fibonacci(10) = " << FIB.get(10) << std::endl;
+    return 0;
+}
diff --git a/exercises/exercises/13_class/main.cpp b/exercises/exercises/13_class/main.cpp
new file mode 100644
index 000000000..b6dc868d1
--- /dev/null
+++ b/exercises/exercises/13_class/main.cpp
@@ -0,0 +1,42 @@
+#include "../exercise.h"
+
+// C++ 中，`class` 和 `struct` 之间的**唯一区别**是
+// `class` 默认访问控制符是 `private`，
+// `struct` 默认访问控制符是 `public`。
+// READ: 访问说明符 <https://zh.cppreference.com/w/cpp/language/access>
+
+// 这个 class 中的字段被 private 修饰，只能在 class 内部访问。
+// 因此必须提供构造器来初始化字段。
+// READ: 构造器 <https://zh.cppreference.com/w/cpp/language/constructor>
+class Fibonacci {
+    size_t cache[16];
+    int cached;
+
+public:
+    // TODO: 实现构造器
+    // Fibonacci()
+    Fibonacci(const size_t inputCache[16], int cached) {
+        for (int i = 0; i < 16; ++i) {
+            cache[i] = inputCache[i];
+        }
+        this->cached = cached;
+    }
+
+    // TODO: 实现正确的缓存优化斐波那契计算
+    size_t get(int i) {
+        for (; cached<=i; ++cached) {
+            cache[cached] = cache[cached - 1] + cache[cached - 2];
+        }
+        return cache[i];
+    }
+};
+
+int main(int argc, char **argv) {
+    // 现在类型拥有无参构造器，声明时会直接调用。
+    // 这个写法不再是未定义行为了。
+    size_t cache[16]{0, 1};
+    Fibonacci fib(cache, 2);
+    ASSERT(fib.get(10) == 55, "fibonacci(10) should be 55");
+    std::cout << "fibonacci(10) = " << fib.get(10) << std::endl;
+    return 0;
+}
diff --git a/exercises/exercises/14_class_destruct/main.cpp b/exercises/exercises/14_class_destruct/main.cpp
new file mode 100644
index 000000000..f00bcb789
--- /dev/null
+++ b/exercises/exercises/14_class_destruct/main.cpp
@@ -0,0 +1,42 @@
+#include "../exercise.h"
+
+// READ: 析构函数 <https://zh.cppreference.com/w/cpp/language/destructor>
+// READ: RAII <https://learn.microsoft.com/zh-cn/cpp/cpp/object-lifetime-and-resource-management-modern-cpp?view=msvc-170>
+
+/// @brief 任意缓存容量的斐波那契类型。
+/// @details 可以在构造时传入缓存容量，因此需要动态分配缓存空间。
+class DynFibonacci {
+    size_t *cache;
+    int cached;
+
+public:
+    // TODO: 实现动态设置容量的构造器
+    DynFibonacci(int capacity) {
+        cache = new size_t[capacity];
+        cache[0] = 0;
+        cache[1] = 1;
+        cached = 2;
+    }
+
+    // TODO: 实现析构器，释放缓存空间
+    ~DynFibonacci();
+
+    // TODO: 实现正确的缓存优化斐波那契计算
+    size_t get(int i) {
+        for (; cached <= i; ++cached) {
+            cache[cached] = cache[cached - 1] + cache[cached - 2];
+        }
+        return cache[i];
+    }
+};
+
+DynFibonacci ::~DynFibonacci() {
+    delete[] cache;
+}
+
+int main(int argc, char **argv) {
+    DynFibonacci fib(12);
+    ASSERT(fib.get(10) == 55, "fibonacci(10) should be 55");
+    std::cout << "fibonacci(10) = " << fib.get(10) << std::endl;
+    return 0;
+}
diff --git a/exercises/exercises/15_class_clone/main.cpp b/exercises/exercises/15_class_clone/main.cpp
new file mode 100644
index 000000000..38737265a
--- /dev/null
+++ b/exercises/exercises/15_class_clone/main.cpp
@@ -0,0 +1,61 @@
+#include "../exercise.h"
+
+// READ: 复制构造函数 <https://zh.cppreference.com/w/cpp/language/copy_constructor>
+// READ: 函数定义（显式弃置）<https://zh.cppreference.com/w/cpp/language/function>
+
+
+class DynFibonacci {
+    size_t *cache;
+    int cached;
+
+public:
+    // TODO: 实现动态设置容量的构造器
+    DynFibonacci(int capacity) : cache(new size_t[capacity]), cached(2) {
+        cache[0] = 0;
+        cache[1] = 1;
+    }
+
+    // TODO: 实现复制构造器
+    DynFibonacci(DynFibonacci const &c) {
+        this->cache = new size_t[c.cached];
+        this->cached = c.cached;
+        for (int i = 0; i < c.cached; i++) {
+            this->cache[i] = c.cache[i];
+        }
+    }
+
+    // TODO: 实现析构器，释放缓存空间
+    ~DynFibonacci();
+
+    // TODO: 实现正确的缓存优化斐波那契计算
+    size_t get(int i) {
+        for (; cached <= i; ++cached) {
+            cache[cached] = cache[cached - 1] + cache[cached - 2];
+        }
+        return cache[i];
+    }
+
+    // NOTICE: 不要修改这个方法
+    // NOTICE: 名字相同参数也相同，但 const 修饰不同的方法是一对重载方法，可以同时存在
+    //         本质上，方法是隐藏了 this 参数的函数
+    //         const 修饰作用在 this 上，因此它们实际上参数不同
+    size_t get(int i) const {
+        if (i <= cached) {
+            return cache[i];
+        }
+        ASSERT(false, "i out of range");
+    }
+};
+
+
+DynFibonacci::~DynFibonacci() {
+    delete[] cache;
+}
+
+int main(int argc, char **argv) {
+    DynFibonacci fib(12);
+    ASSERT(fib.get(10) == 55, "fibonacci(10) should be 55");
+    DynFibonacci const fib_ = fib;
+    ASSERT(fib_.get(10) == fib.get(10), "Object cloned");
+    return 0;
+}
diff --git a/exercises/exercises/16_class_move/main.cpp b/exercises/exercises/16_class_move/main.cpp
new file mode 100644
index 000000000..2269a80b8
--- /dev/null
+++ b/exercises/exercises/16_class_move/main.cpp
@@ -0,0 +1,90 @@
+#include "../exercise.h"
+
+// READ: 左值右值（概念）<https://learn.microsoft.com/zh-cn/cpp/c-language/l-value-and-r-value-expressions?view=msvc-170>
+// READ: 左值右值（细节）<https://zh.cppreference.com/w/cpp/language/value_category>
+// READ: 关于移动语义 <https://learn.microsoft.com/zh-cn/cpp/cpp/rvalue-reference-declarator-amp-amp?view=msvc-170#move-semantics>
+// READ: 如果实现移动构造 <https://learn.microsoft.com/zh-cn/cpp/cpp/move-constructors-and-move-assignment-operators-cpp?view=msvc-170>
+
+// READ: 移动构造函数 <https://zh.cppreference.com/w/cpp/language/move_constructor>
+// READ: 移动赋值 <https://zh.cppreference.com/w/cpp/language/move_assignment>
+// READ: 运算符重载 <https://zh.cppreference.com/w/cpp/language/operators>
+
+class DynFibonacci {
+    size_t *cache;
+    int cached;
+
+public:
+    // TODO: 实现动态设置容量的构造器
+    DynFibonacci(int capacity): cache(new size_t[capacity]), cached(2) 
+    {
+        cache[0] = 0;
+        cache[1] = 1;
+    }
+
+    // TODO: 实现移动构造器
+    DynFibonacci(DynFibonacci &&c) noexcept : cache(c.cache), cached(c.cached)
+    {
+        c.cache =nullptr;
+        c.cached = 0;
+    }
+
+    // TODO: 实现移动赋值
+    // NOTICE: ⚠ 注意移动到自身问题 ⚠
+    DynFibonacci& operator=(DynFibonacci&& c) noexcept
+    {
+        if (this != &c)
+        {
+            delete[] cache;
+            cache = c.cache;
+            cached = c.cached;
+
+            c.cache = nullptr;
+            c.cached = 0;
+        }
+        return *this;
+    }
+
+    // TODO: 实现析构器，释放缓存空间
+    ~DynFibonacci()
+    {
+        delete[]cache;
+        cache = nullptr;
+    }
+
+    // TODO: 实现正确的缓存优化斐波那契计算
+    size_t operator[](int i) {
+        for (; cached<=i; ++cached) {
+            cache[cached] = cache[cached - 1] + cache[cached - 2];
+        }
+        return cache[i];
+    }
+
+    // NOTICE: 不要修改这个方法
+    size_t operator[](int i) const {
+        ASSERT(i <= cached, "i out of range");
+        return cache[i];
+    }
+
+    // NOTICE: 不要修改这个方法
+    bool is_alive() const {
+        return cache;
+    }
+};
+
+int main(int argc, char **argv) {
+    DynFibonacci fib(12);
+    ASSERT(fib[10] == 55, "fibonacci(10) should be 55");
+
+    DynFibonacci const fib_ = std::move(fib);
+    ASSERT(!fib.is_alive(), "Object moved");
+    ASSERT(fib_[10] == 55, "fibonacci(10) should be 55");
+
+    DynFibonacci fib0(6);
+    DynFibonacci fib1(12);
+
+    fib0 = std::move(fib1);
+    fib0 = std::move(fib0);
+    ASSERT(fib0[10] == 55, "fibonacci(10) should be 55");
+
+    return 0;
+}
diff --git a/exercises/exercises/17_class_derive/main.cpp b/exercises/exercises/17_class_derive/main.cpp
new file mode 100644
index 000000000..93014cfa6
--- /dev/null
+++ b/exercises/exercises/17_class_derive/main.cpp
@@ -0,0 +1,77 @@
+#include "../exercise.h"
+
+// READ: 派生类 <https://zh.cppreference.com/w/cpp/language/derived_class>
+
+static int i = 0;
+
+struct X {
+    int x;
+
+    X(int x_) : x(x_) {
+        std::cout << ++i << ". " << "X(" << x << ')' << std::endl;
+    }
+    X(X const &other) : x(other.x) {
+        std::cout << ++i << ". " << "X(X const &) : x(" << x << ')' << std::endl;
+    }
+    ~X() {
+        std::cout << ++i << ". " << "~X(" << x << ')' << std::endl;
+    }
+};
+struct A {
+    int a;
+
+    A(int a_) : a(a_) {
+        std::cout << ++i << ". " << "A(" << a << ')' << std::endl;
+    }
+    A(A const &other) : a(other.a) {
+        std::cout << ++i << ". " << "A(A const &) : a(" << a << ')' << std::endl;
+    }
+    ~A() {
+        std::cout << ++i << ". " << "~A(" << a << ')' << std::endl;
+    }
+};
+struct B : public A {
+    X x;
+
+    B(int b) : A(1), x(b) {
+        std::cout << ++i << ". " << "B(" << a << ", X(" << x.x << "))" << std::endl;
+    }
+    B(B const &other) : A(other.a), x(other.x) {
+        std::cout << ++i << ". " << "B(B const &) : A(" << a << "), x(X(" << x.x << "))" << std::endl;
+    }
+    ~B() {
+        std::cout << ++i << ". " << "~B(" << a << ", X(" << x.x << "))" << std::endl;
+    }
+};
+
+int main(int argc, char **argv) {
+    X x = X(1);
+    A a = A(2);
+    B b = B(3);
+
+    // TODO: 补全三个类型的大小
+    static_assert(sizeof(X) == sizeof(int), "There is an int in X");
+    static_assert(sizeof(A) == sizeof(int), "There is an int in A");
+    static_assert(sizeof(B) == 2*sizeof(int), "B is an A with an X");
+
+    i = 0;
+    std::cout << std::endl
+              << "-------------------------" << std::endl
+              << std::endl;
+
+    // 这是不可能的，A 无法提供 B 增加的成员变量的值
+    // B ba = A(4);
+
+    // 这也是不可能的，因为 A 是 B 的一部分，就好像不可以把套娃的外层放进内层里。
+    A ab = B(5);// 然而这个代码可以编译和运行！
+    // THINK: 观察打印出的信息，推测把大象放进冰箱分几步？
+    // THINK: 这样的代码是“安全”的吗？
+    // NOTICE: 真实场景中不太可能出现这样的代码
+
+    i = 0;
+    std::cout << std::endl
+              << "-------------------------" << std::endl
+              << std::endl;
+
+    return 0;
+}
diff --git a/exercises/exercises/18_class_virtual/main.cpp b/exercises/exercises/18_class_virtual/main.cpp
new file mode 100644
index 000000000..344709e4c
--- /dev/null
+++ b/exercises/exercises/18_class_virtual/main.cpp
@@ -0,0 +1,82 @@
+#include "../exercise.h"
+
+// READ: 虚函数 <https://zh.cppreference.com/w/cpp/language/virtual>
+
+struct A {
+    virtual char virtual_name() const {
+        return 'A';
+    }
+    char direct_name() const {
+        return 'A';
+    }
+};
+struct B : public A {
+    // READ: override <https://zh.cppreference.com/w/cpp/language/override>
+    char virtual_name() const override {
+        return 'B';
+    }
+    char direct_name() const {
+        return 'B';
+    }
+};
+struct C : public B {
+    // READ: final <https://zh.cppreference.com/w/cpp/language/final>
+    char virtual_name() const final {
+        return 'C';
+    }
+    char direct_name() const {
+        return 'C';
+    }
+};
+struct D : public C {
+    char direct_name() const {
+        return 'D';
+    }
+};
+
+int main(int argc, char **argv) {
+    constexpr auto MSG = "Replace '?' with its correct name.";
+
+    A a;
+    B b;
+    C c;
+    D d;
+
+    ASSERT(a.virtual_name() == 'A', MSG);
+    ASSERT(b.virtual_name() == 'B', MSG);
+    ASSERT(c.virtual_name() == 'C', MSG);
+    ASSERT(d.virtual_name() == 'C', MSG);
+    ASSERT(a.direct_name() == 'A', MSG);
+    ASSERT(b.direct_name() == 'B', MSG);
+    ASSERT(c.direct_name() == 'C', MSG);
+    ASSERT(d.direct_name() == 'D', MSG);
+
+    A &rab = b;
+    B &rbc = c;
+    C &rcd = d;
+
+    ASSERT(rab.virtual_name() == 'B', MSG);
+    ASSERT(rbc.virtual_name() == 'C', MSG);
+    ASSERT(rcd.virtual_name() == 'C', MSG);
+    ASSERT(rab.direct_name() == 'A', MSG);
+    ASSERT(rbc.direct_name() == 'B', MSG);
+    ASSERT(rcd.direct_name() == 'C', MSG);
+
+    A &rac = c;
+    B &rbd = d;
+
+    ASSERT(rac.virtual_name() == 'C', MSG);
+    ASSERT(rbd.virtual_name() == 'C', MSG);
+    ASSERT(rac.direct_name() == 'A', MSG);
+    ASSERT(rbd.direct_name() == 'B', MSG);
+
+    A &rad = d;
+
+    ASSERT(rad.virtual_name() == 'C', MSG);
+    ASSERT(rad.direct_name() == 'A', MSG);
+
+    return 0;
+}
+
+// READ: 扩展阅读-纯虚、抽象 <https://zh.cppreference.com/w/cpp/language/abstract_class>
+// READ: 扩展阅读-虚继承 <https://zh.cppreference.com/w/cpp/language/derived_class>
diff --git a/exercises/exercises/19_class_virtual_destruct/main.cpp b/exercises/exercises/19_class_virtual_destruct/main.cpp
new file mode 100644
index 000000000..8d66e8edf
--- /dev/null
+++ b/exercises/exercises/19_class_virtual_destruct/main.cpp
@@ -0,0 +1,70 @@
+#include "../exercise.h"
+
+// READ: 静态字段 <https://zh.cppreference.com/w/cpp/language/static>
+// READ: 虚析构函数 <https://zh.cppreference.com/w/cpp/language/destructor>
+
+
+struct A {
+    // TODO: 正确初始化静态字段
+    static int num_a;
+
+    A() {
+        ++num_a;
+    }
+    virtual ~A() {
+        --num_a;
+    }
+
+    virtual char name() const {
+        return 'A';
+    }
+};
+struct B final : public A {
+    // TODO: 正确初始化静态字段
+    static int num_b;
+
+    B() {
+        ++num_b;
+    }
+    ~B() override {
+        --num_b;
+    }
+
+    char name() const final {
+        return 'B';
+    }
+};
+
+int A::num_a = 0;
+int B::num_b = 0;
+
+int main(int argc, char **argv) {
+    auto a = new A;
+    auto b = new B;
+    ASSERT(A::num_a == 2, "Fill in the correct value for A::num_a");
+    ASSERT(B::num_b == 1, "Fill in the correct value for B::num_b");
+    ASSERT(a->name() == 'A', "Fill in the correct value for a->name()");
+    ASSERT(b->name() == 'B', "Fill in the correct value for b->name()");
+
+    delete a;
+    delete b;
+    ASSERT(A::num_a == 0, "Every A was destroyed");
+    ASSERT(B::num_b == 0, "Every B was destroyed");
+
+    A *ab = new B;// 派生类指针可以随意转换为基类指针
+    ASSERT(A::num_a == 1, "Fill in the correct value for A::num_a");
+    ASSERT(B::num_b == 1, "Fill in the correct value for B::num_b");
+    ASSERT(ab->name() == 'B', "Fill in the correct value for ab->name()");
+
+    // TODO: 基类指针无法随意转换为派生类指针，补全正确的转换语句
+    //B &bb = *ab;
+    B &bb = dynamic_cast<B &>(*ab);
+    ASSERT(bb.name() == 'B', "Fill in the correct value for bb->name()");
+
+    // TODO: ---- 以下代码不要修改，通过改正类定义解决编译问题 ----
+    delete ab;// 通过指针可以删除指向的对象，即使是多态对象
+    ASSERT(A::num_a == 0, "Every A was destroyed");
+    ASSERT(B::num_b == 0, "Every B was destroyed");
+
+    return 0;
+}
diff --git a/exercises/exercises/20_function_template/main.cpp b/exercises/exercises/20_function_template/main.cpp
new file mode 100644
index 000000000..7196e8930
--- /dev/null
+++ b/exercises/exercises/20_function_template/main.cpp
@@ -0,0 +1,21 @@
+#include "../exercise.h"
+
+// READ: 函数模板 <https://zh.cppreference.com/w/cpp/language/function_template>
+// TODO: 将这个函数模板化
+template<typename T>
+T plus(T a, T b) {
+    return a + b;
+}
+
+int main(int argc, char **argv) {
+    ASSERT(plus(1, 2) == 3, "Plus two int");
+    ASSERT(plus(1u, 2u) == 3u, "Plus two unsigned int");
+
+    // THINK: 浮点数何时可以判断 ==？何时必须判断差值？
+    ASSERT(plus(1.25f, 2.5f) == 3.75f, "Plus two float");
+    ASSERT(plus(1.25, 2.5) == 3.75, "Plus two double");
+    // TODO: 修改判断条件使测试通过
+    ASSERT(std::abs(plus(0.1, 0.2) - 0.3) < 1e-9, "How to make this pass?");
+
+    return 0;
+}
diff --git a/exercises/exercises/21_runtime_datatype/main.cpp b/exercises/exercises/21_runtime_datatype/main.cpp
new file mode 100644
index 000000000..153ab5b2d
--- /dev/null
+++ b/exercises/exercises/21_runtime_datatype/main.cpp
@@ -0,0 +1,54 @@
+#include "../exercise.h"
+#include <cmath>
+
+enum class DataType {
+    Float,
+    Double,
+};
+
+/// @brief Tagged union 即标签化联合体，是联合体的一种常见应用。
+///        Rust enum 在实现上就是标签化联合体。
+struct TaggedUnion {
+    DataType type;
+    // NOTICE: struct/union 可以相互任意嵌套。
+    union {
+        float f;
+        double d;
+    };
+};
+
+// TODO: 将这个函数模板化用于 sigmoid_dyn
+template<typename T>
+T sigmoid(T x) {
+    return 1 / (1 + std::exp(-x));
+}
+
+TaggedUnion sigmoid_dyn(TaggedUnion x) {
+    TaggedUnion ans{x.type};
+    // TODO: 根据 type 调用 sigmoid
+    if (DataType::Float == x.type)
+    {
+        ans.f = sigmoid(x.f);
+    }
+    else
+    {
+        ans.d = sigmoid(x.d);
+    }
+    return ans;
+}
+
+// ---- 不要修改以下代码 ----
+int main(int argc, char **argv) {
+    TaggedUnion xf{DataType::Float};
+    xf.f = 5.f;
+    auto yf = sigmoid_dyn(xf);
+    ASSERT(yf.type == DataType::Float, "type mismatch");
+    ASSERT(yf.f == 1 / (1 + std::exp(-5.f)), "sigmoid float");
+
+    TaggedUnion xd{DataType::Double};
+    xd.d = 5.0;
+    auto yd = sigmoid_dyn(xd);
+    ASSERT(yd.type == DataType::Double, "type mismatch");
+    ASSERT(yd.d == 1 / (1 + std::exp(-5.0)), "sigmoid double");
+    return 0;
+}
diff --git a/exercises/exercises/22_class_template/main.cpp b/exercises/exercises/22_class_template/main.cpp
new file mode 100644
index 000000000..890d7cb7c
--- /dev/null
+++ b/exercises/exercises/22_class_template/main.cpp
@@ -0,0 +1,147 @@
+#include "../exercise.h"
+#include <cstring>
+// READ: 类模板 <https://zh.cppreference.com/w/cpp/language/class_template>
+#include <cassert>
+
+
+template<typename T>
+// Tensor4D(const unsigned int[4], const T *) -> Tensor4D<T>;
+
+struct Tensor4D {
+    unsigned int shape[4];
+    T *data;
+
+    Tensor4D(unsigned int const shape_[4], T const *data_) {
+        unsigned int size = 1;
+        // TODO: 填入正确的 shape 并计算 size
+        for (int i = 0; i < 4; ++i) {
+            shape[i] = shape_[i];
+            size *= shape[i];
+        }
+        data = new T[size];
+        std::memcpy(data, data_, size * sizeof(T));
+    }
+    ~Tensor4D() {
+        delete[] data;
+    }
+
+    // 为了保持简单，禁止复制和移动
+    Tensor4D(Tensor4D const &) = delete;
+    Tensor4D(Tensor4D &&) noexcept = delete;
+
+    // 这个加法需要支持“单向广播”。
+    // 具体来说，`others` 可以具有与 `this` 不同的形状，形状不同的维度长度必须为 1。
+    // `others` 长度为 1 但 `this` 长度不为 1 的维度将发生广播计算。
+    // 例如，`this` 形状为 `[1, 2, 3, 4]`，`others` 形状为 `[1, 2, 1, 4]`，
+    // 则 `this` 与 `others` 相加时，3 个形状为 `[1, 2, 1, 4]` 的子张量各自与 `others` 对应项相加。
+    Tensor4D &operator+=(Tensor4D const &others) {
+        // TODO: 实现单向广播的加法
+        unsigned int total = 1;
+        for (int i = 0; i < 4; ++i) {
+            // others维度要么等于当前维度，要么为1（广播）
+            assert((others.shape[i] == shape[i]) || (others.shape[i] == 1));
+            total *= shape[i];
+        }
+
+        for (unsigned int i = 0; i < total; ++i) {
+            // 将一维索引i解码成4维索引
+            unsigned int index[4];
+            unsigned int stride = total;
+            unsigned int offset = i;
+            for (int d = 0; d < 4; ++d) {
+                stride /= shape[d];
+                index[d] = offset / stride;
+                offset %= stride;
+            }
+
+            // 计算others中对应的广播后索引
+            unsigned int other_index = 0;
+            unsigned int other_stride = 1;
+            for (int d = 3; d >= 0; --d) {
+                unsigned int idx = (others.shape[d] == 1) ? 0 : index[d];
+                other_index += idx * other_stride;
+                other_stride *= others.shape[d];
+            }
+
+            data[i] += others.data[other_index];
+        }
+
+        return *this;
+    }
+};
+
+// ---- 不要修改以下代码 ----
+int main(int argc, char **argv) {
+    {
+        unsigned int shape[]{1, 2, 3, 4};
+        // clang-format off
+        int data[]{
+             1,  2,  3,  4,
+             5,  6,  7,  8,
+             9, 10, 11, 12,
+
+            13, 14, 15, 16,
+            17, 18, 19, 20,
+            21, 22, 23, 24};
+        // clang-format on
+        auto t0 = Tensor4D(shape, data);
+        auto t1 = Tensor4D(shape, data);
+        t0 += t1;
+        for (auto i = 0u; i < sizeof(data) / sizeof(*data); ++i) {
+            ASSERT(t0.data[i] == data[i] * 2, "Tensor doubled by plus its self.");
+        }
+    }
+    {
+        unsigned int s0[]{1, 2, 3, 4};
+        // clang-format off
+        float d0[]{
+            1, 1, 1, 1,
+            2, 2, 2, 2,
+            3, 3, 3, 3,
+
+            4, 4, 4, 4,
+            5, 5, 5, 5,
+            6, 6, 6, 6};
+        // clang-format on
+        unsigned int s1[]{1, 2, 3, 1};
+        // clang-format off
+        float d1[]{
+            6,
+            5,
+            4,
+
+            3,
+            2,
+            1};
+        // clang-format on
+
+        auto t0 = Tensor4D(s0, d0);
+        auto t1 = Tensor4D(s1, d1);
+        t0 += t1;
+        for (auto i = 0u; i < sizeof(d0) / sizeof(*d0); ++i) {
+            ASSERT(t0.data[i] == 7.f, "Every element of t0 should be 7 after adding t1 to it.");
+        }
+    }
+    {
+        unsigned int s0[]{1, 2, 3, 4};
+        // clang-format off
+        double d0[]{
+             1,  2,  3,  4,
+             5,  6,  7,  8,
+             9, 10, 11, 12,
+
+            13, 14, 15, 16,
+            17, 18, 19, 20,
+            21, 22, 23, 24};
+        // clang-format on
+        unsigned int s1[]{1, 1, 1, 1};
+        double d1[]{1};
+
+        auto t0 = Tensor4D(s0, d0);
+        auto t1 = Tensor4D(s1, d1);
+        t0 += t1;
+        for (auto i = 0u; i < sizeof(d0) / sizeof(*d0); ++i) {
+            ASSERT(t0.data[i] == d0[i] + 1, "Every element of t0 should be incremented by 1 after adding t1 to it.");
+        }
+    }
+}
diff --git a/exercises/exercises/23_template_const/main.cpp b/exercises/exercises/23_template_const/main.cpp
new file mode 100644
index 000000000..f2f49c3d1
--- /dev/null
+++ b/exercises/exercises/23_template_const/main.cpp
@@ -0,0 +1,92 @@
+#include "../exercise.h"
+#include <cstring>
+
+// READ: 模板非类型实参 <https://zh.cppreference.com/w/cpp/language/template_parameters#%E6%A8%A1%E6%9D%BF%E9%9D%9E%E7%B1%BB%E5%9E%8B%E5%AE%9E%E5%8F%82>
+
+template<unsigned int N, class T>
+struct Tensor {
+    unsigned int shape[N];
+    T *data;
+
+    Tensor(unsigned int const shape_[N]) {
+        unsigned int size = 1;
+        // TODO: 填入正确的 shape 并计算 size
+        for (unsigned int i = 0; i < N; ++i) {
+            shape[i] = shape_[i];
+            size *= shape[i];
+        }
+        data = new T[size];
+        std::memset(data, 0, size * sizeof(T));
+    }
+    ~Tensor() {
+        delete[] data;
+    }
+
+    // 为了保持简单，禁止复制和移动
+    Tensor(Tensor const &) = delete;
+    //Tensor(Tensor&&) noexcept = delete;
+    Tensor(Tensor &&other) noexcept {
+        std::memcpy(shape, other.shape, sizeof(shape));
+        data = other.data;
+        other.data = nullptr;// 转移所有权
+    }
+
+
+    T &operator[](unsigned int const indices[N]) {
+        return data[data_index(indices)];
+    }
+    T const &operator[](unsigned int const indices[N]) const {
+        return data[data_index(indices)];
+    }
+
+private:
+    unsigned int data_index(unsigned int const indices[N]) const {
+        unsigned int index = 0;
+        unsigned int stride[N];
+        stride[N - 1] = 1;
+        for (int i = N - 2; i >= 0; --i) {
+            stride[i] = stride[i + 1] * shape[i + 1];
+        }
+
+        // 再正序遍历，用 stride 乘索引累加
+        for (unsigned int i = 0; i < N; ++i) {
+            ASSERT(indices[i] < shape[i], "Invalid index");
+            index += indices[i] * stride[i];
+        }
+
+        return index;
+    }
+};
+
+// ---- 不要修改以下代码 ----
+int main(int argc, char **argv) {
+    {
+        unsigned int shape[]{2, 3, 4, 5};
+        auto tensor = Tensor<4, int>(shape);
+
+        unsigned int i0[]{0, 0, 0, 0};
+        tensor[i0] = 1;
+        ASSERT(tensor[i0] == 1, "tensor[i0] should be 1");
+        ASSERT(tensor.data[0] == 1, "tensor[i0] should be 1");
+
+        unsigned int i1[]{1, 2, 3, 4};
+        tensor[i1] = 2;
+        ASSERT(tensor[i1] == 2, "tensor[i1] should be 2");
+        ASSERT(tensor.data[119] == 2, "tensor[i1] should be 2");
+    }
+    {
+        unsigned int shape[]{7, 8, 128};
+        auto tensor = Tensor<3, float>(shape);
+
+        unsigned int i0[]{0, 0, 0};
+        tensor[i0] = 1.f;
+        ASSERT(tensor[i0] == 1.f, "tensor[i0] should be 1");
+        ASSERT(tensor.data[0] == 1.f, "tensor[i0] should be 1");
+
+        unsigned int i1[]{3, 4, 99};
+        tensor[i1] = 2.f;
+        ASSERT(tensor[i1] == 2.f, "tensor[i1] should be 2");
+        ASSERT(tensor.data[3683] == 2.f, "tensor[i1] should be 2");
+    }
+    return 0;
+}
diff --git a/exercises/exercises/24_std_array/main.cpp b/exercises/exercises/24_std_array/main.cpp
new file mode 100644
index 000000000..8ac430609
--- /dev/null
+++ b/exercises/exercises/24_std_array/main.cpp
@@ -0,0 +1,29 @@
+#include "../exercise.h"
+#include <array>
+#include <cstring>
+
+// READ: std::array <https://zh.cppreference.com/w/cpp/container/array>
+
+// TODO: 将下列 `?` 替换为正确的代码
+int main(int argc, char **argv) {
+    {
+        std::array<int, 5> arr{{1, 2, 3, 4, 5}};
+        ASSERT(arr.size() == 5, "Fill in the correct value.");
+        ASSERT(sizeof(arr) == 20, "Fill in the correct value.");
+        int ans[]{1, 2, 3, 4, 5};
+        ASSERT(std::memcmp(arr.data(), ans, sizeof(ans)) == 0, "Fill in the correct values.");
+    }
+    {
+        std::array<double, 8> arr;
+        ASSERT(arr.size() == 8, "Fill in the correct value.");
+        ASSERT(sizeof(arr) == 64, "Fill in the correct value.");
+    }
+    {
+        std::array<char, 21> arr{"Hello, InfiniTensor!"};
+        ASSERT(arr.size() == 21, "Fill in the correct value.");
+        //std::cout << sizeof(char) << std::endl;
+        ASSERT(sizeof(arr) == 21, "Fill in the correct value.");
+        ASSERT(std::strcmp(arr.data(), "Hello, InfiniTensor!") == 0, "Fill in the correct value.");
+    }
+    return 0;
+}
diff --git a/exercises/exercises/25_std_vector/main.cpp b/exercises/exercises/25_std_vector/main.cpp
new file mode 100644
index 000000000..3872b36d5
--- /dev/null
+++ b/exercises/exercises/25_std_vector/main.cpp
@@ -0,0 +1,90 @@
+#include "../exercise.h"
+#include <cstring>
+#include <vector>
+
+// READ: std::vector <https://zh.cppreference.com/w/cpp/container/vector>
+
+// TODO: 将下列 `?` 替换为正确的代码
+int main(int argc, char** argv) {
+    {
+        std::vector<int> vec{ 1, 2, 3, 4, 5 };
+        ASSERT(vec.size() == 5, "Fill in the correct value.");
+        // THINK: `std::vector` 的大小是什么意思？与什么有关？
+        ASSERT(sizeof(vec) == sizeof(std::vector<int>), "Fill in the correct value.");
+        int ans[]{ 1, 2, 3, 4, 5 };
+        ASSERT(std::memcmp(vec.data(), ans, sizeof(ans)) == 0, "Fill in the correct values.");
+    }
+    {
+        std::vector<double> vec{ 1, 2, 3, 4, 5 };
+        {
+            ASSERT(vec.size() == 5, "Fill in the correct value.");
+            ASSERT(sizeof(vec) == sizeof(std::vector<double>), "Fill in the correct value.");
+            double ans[]{ 1, 2, 3, 4, 5 };
+            ASSERT(std::memcmp(vec.data(), ans, sizeof(ans)) == 0, "Fill in the correct values.");
+        }
+        {
+            vec.push_back(6);
+            ASSERT(vec.size() == 6, "Fill in the correct value.");
+            ASSERT(sizeof(vec) == sizeof(std::vector<double>), "Fill in the correct value.");
+            vec.pop_back();
+            ASSERT(vec.size() == 5, "Fill in the correct value.");
+            ASSERT(sizeof(vec) == sizeof(std::vector<double>), "Fill in the correct value.");
+        }
+        {
+            vec[4] = 6;
+            ASSERT(vec[0] == 1, "Fill in the correct value.");
+            ASSERT(vec[1] == 2, "Fill in the correct value.");
+            ASSERT(vec[2] == 3, "Fill in the correct value.");
+            ASSERT(vec[3] == 4, "Fill in the correct value.");
+            ASSERT(vec[4] == 6, "Fill in the correct value.");
+        }
+        {
+            // THINK: `std::vector` 插入删除的时间复杂度是什么？
+            vec.insert(vec.begin() + 1, 1.5);
+            ASSERT((vec == std::vector<double>{1, 1.5, 2, 3, 4, 6}), "Make this assertion pass.");
+            vec.erase(vec.begin() + 3);
+            ASSERT((vec == std::vector<double>{1, 1.5, 2, 4, 6}), "Make this assertion pass.");
+        }
+        {
+            vec.shrink_to_fit();
+            ASSERT(vec.capacity() == 5, "Fill in the correct value.");
+            vec.clear();
+            ASSERT(vec.empty(), "`vec` is empty now.");
+            ASSERT(vec.size() == 0, "Fill in the correct value.");
+            ASSERT(vec.capacity() == 5, "Fill in the correct value.");
+        }
+    }
+    {
+        std::vector<char> vec(48, 'z');// TODO: 调用正确的构造函数
+        ASSERT(vec[0] == 'z', "Make this assertion pass.");
+        ASSERT(vec[47] == 'z', "Make this assertion pass.");
+        ASSERT(vec.size() == 48, "Make this assertion pass.");
+        ASSERT(sizeof(vec) == sizeof(std::vector<char>), "Fill in the correct value.");
+        {
+            auto capacity = vec.capacity();
+            vec.resize(16);
+            ASSERT(vec.size() == 16, "Fill in the correct value.");
+            ASSERT(vec.capacity() == capacity, "Fill in a correct identifier.");
+        }
+        {
+            vec.reserve(256);
+            ASSERT(vec.size() == 16, "Fill in the correct value.");
+            ASSERT(vec.capacity() == 256, "Fill in the correct value.");
+        }
+        {
+            vec.push_back('a');
+            vec.push_back('b');
+            vec.push_back('c');
+            vec.push_back('d');
+            ASSERT(vec.size() == 20, "Fill in the correct value.");
+            ASSERT(vec.capacity() == 256, "Fill in the correct value.");
+            ASSERT(vec[15] == 'z', "Fill in the correct value.");
+            ASSERT(vec[16] == 'a', "Fill in the correct value.");
+            ASSERT(vec[17] == 'b', "Fill in the correct value.");
+            ASSERT(vec[18] == 'c', "Fill in the correct value.");
+            ASSERT(vec[19] == 'd', "Fill in the correct value.");
+        }
+    }
+
+    return 0;
+}
diff --git a/exercises/exercises/26_std_vector_bool/main.cpp b/exercises/exercises/26_std_vector_bool/main.cpp
new file mode 100644
index 000000000..972a80b08
--- /dev/null
+++ b/exercises/exercises/26_std_vector_bool/main.cpp
@@ -0,0 +1,34 @@
+#include "../exercise.h"
+#include <vector>
+
+// READ: std::vector <https://zh.cppreference.com/w/cpp/container/vector_bool>
+// READ: 模板特化 <https://zh.cppreference.com/w/cpp/language/template_specialization>
+
+// TODO: 将下列 `?` 替换为正确的代码
+int main(int argc, char **argv) {
+    std::vector<bool> vec(100, true);// TODO: 正确调用构造函数
+    ASSERT(vec[0], "Make this assertion pass.");
+    ASSERT(vec[99], "Make this assertion pass.");
+    ASSERT(vec.size() == 100, "Make this assertion pass.");
+    // NOTICE: 平台相关！注意 CI:Ubuntu 上的值。
+    std::cout << "sizeof(std::vector<bool>) = " << sizeof(std::vector<bool>) << std::endl;
+    ASSERT(sizeof(vec) == sizeof(std::vector<bool>), "Fill in the correct value.");
+    {
+        vec[20] = false;
+        ASSERT(!vec[20], "Fill in `vec[20]` or `!vec[20]`.");
+    }
+    {
+        vec.push_back(false);
+        ASSERT(vec.size() == 101, "Fill in the correct value.");
+        ASSERT(!vec[100], "Fill in `vec[100]` or `!vec[100]`.");
+    }
+    {
+        auto ref = vec[30];
+        ASSERT(ref, "Fill in `ref` or `!ref`");
+        ref = false;
+        ASSERT(!ref, "Fill in `ref` or `!ref`");
+        // THINK: WHAT and WHY?
+        ASSERT(!vec[30], "Fill in `vec[30]` or `!vec[30]`.");
+    }
+    return 0;
+}
diff --git a/exercises/exercises/27_strides/main.cpp b/exercises/exercises/27_strides/main.cpp
new file mode 100644
index 000000000..13feaf5ae
--- /dev/null
+++ b/exercises/exercises/27_strides/main.cpp
@@ -0,0 +1,38 @@
+#include "../exercise.h"
+#include <vector>
+
+// 张量即多维数组。连续存储张量即逻辑结构与存储结构一致的张量。
+// 通常来说，形状为 [d0, d1, ..., dn] 的张量，第 n 维是 dn 个连续的元素，第 n-1 维是 dn-1 个连续的 dn 个元素，以此类推。
+// 张量的步长或跨度指的是张量每个维度上坐标 +1 时，数据指针跨过的范围。
+// 因此，一个连续张量，其第 n 维的步长为 1，第 n-1 维的步长为 dn，第 n-2 维的步长为 dn*dn-1，以此类推。
+// 例如，一个 2x3x4 张量，其步长为 [12, 4, 1]。
+
+// READ: 类型别名 <https://zh.cppreference.com/w/cpp/language/type_alias>
+using udim = unsigned int;
+
+/// @brief 计算连续存储张量的步长
+/// @param shape 张量的形状
+/// @return 张量每维度的访问步长
+std::vector<udim> strides(std::vector<udim> const &shape) {
+    std::vector<udim> strides(shape.size());
+    // TODO: 完成函数体，根据张量形状计算张量连续存储时的步长。
+    // READ: 逆向迭代器 std::vector::rbegin <https://zh.cppreference.com/w/cpp/container/vector/rbegin>
+    //       使用逆向迭代器可能可以简化代码
+    if (shape.empty()) return strides;
+
+    strides.back() = 1;// 最后一维步长为1
+    for (int i = (int) shape.size() - 2; i >= 0; --i) {
+        strides[i] = shape[i + 1] * strides[i + 1];
+    }
+
+    return strides;
+}
+
+// ---- 不要修改以下代码 ----
+int main(int argc, char **argv) {
+    ASSERT((strides({2, 3, 4}) == std::vector<udim>{12, 4, 1}), "Make this assertion pass.");
+    ASSERT((strides({3, 4, 5}) == std::vector<udim>{20, 5, 1}), "Make this assertion pass.");
+    ASSERT((strides({1, 3, 224, 224}) == std::vector<udim>{150528, 50176, 224, 1}), "Make this assertion pass.");
+    ASSERT((strides({7, 1, 1, 1, 5}) == std::vector<udim>{5, 5, 5, 5, 1}), "Make this assertion pass.");
+    return 0;
+}
diff --git a/exercises/exercises/28_std_string/main.cpp b/exercises/exercises/28_std_string/main.cpp
new file mode 100644
index 000000000..5fcfdf31d
--- /dev/null
+++ b/exercises/exercises/28_std_string/main.cpp
@@ -0,0 +1,18 @@
+#include "../exercise.h"
+#include <string>
+
+// READ: 字符串 <https://zh.cppreference.com/w/cpp/string/basic_string>
+
+int main(int argc, char **argv) {
+    // READ: 字符串字面量 <https://zh.cppreference.com/w/cpp/string/basic_string/operator%22%22s>
+    using namespace std::string_literals;
+    auto hello = "Hello"s;
+    auto world = "world";
+    // READ: `decltype` 表达式 <https://zh.cppreference.com/w/cpp/language/decltype>
+    // READ: `std::is_same_v` 元编程判别 <https://zh.cppreference.com/w/cpp/types/is_same>
+    ASSERT((std::is_same_v<decltype(hello), std::string>), "Fill in the missing type.");
+    ASSERT((std::is_same_v<decltype(world), const char*>), "Fill in the missing type.");
+    // TODO: 将 `?` 替换为正确的字符串
+    ASSERT(hello + ", " + world + '!' == "Hello, world!", "Fill in the missing string.");
+    return 0;
+}
diff --git a/exercises/exercises/29_std_map/main.cpp b/exercises/exercises/29_std_map/main.cpp
new file mode 100644
index 000000000..68122a5fc
--- /dev/null
+++ b/exercises/exercises/29_std_map/main.cpp
@@ -0,0 +1,39 @@
+#include "../exercise.h"
+#include <map>
+
+// READ: `std::map` <https://zh.cppreference.com/w/cpp/container/map>
+// READ: `std::unordered_map` <https://zh.cppreference.com/w/cpp/container/unordered_map>
+
+template<class k, class v>
+bool key_exists(std::map<k, v> const &map, k const &key) {
+    // TODO: 实现函数
+    return map.find(key) != map.end();
+}
+
+template<class k, class v>
+void set(std::map<k, v> &map, k key, v value) {
+    // TODO: 实现函数
+    map[key] = value;
+}
+
+// ---- 不要修改以下代码 ----
+int main(int argc, char **argv) {
+    using namespace std::string_literals;
+
+    std::map<std::string, std::string> secrets;
+
+    set(secrets, "hello"s, "world"s);
+    ASSERT(key_exists(secrets, "hello"s), "\"hello\" shoud be in `secrets`");
+    ASSERT(!key_exists(secrets, "foo"s), "\"foo\" shoud not be in `secrets`");
+
+    set(secrets, "foo"s, "bar"s);
+    set(secrets, "Infini"s, "Tensor"s);
+    ASSERT(secrets["hello"] == "world", "hello -> world");
+    ASSERT(secrets["foo"] == "bar", "foo -> bar");
+    ASSERT(secrets["Infini"] == "Tensor", "Infini -> Tensor");
+
+    set(secrets, "hello"s, "developer"s);
+    ASSERT(secrets["hello"] == "developer", "hello -> developer");
+
+    return 0;
+}
diff --git a/exercises/exercises/30_std_unique_ptr/main.cpp b/exercises/exercises/30_std_unique_ptr/main.cpp
new file mode 100644
index 000000000..d9489ac7b
--- /dev/null
+++ b/exercises/exercises/30_std_unique_ptr/main.cpp
@@ -0,0 +1,72 @@
+#include "../exercise.h"
+#include <cstring>
+#include <memory>
+#include <string>
+#include <vector>
+
+// READ: `std::unique_ptr` <https://zh.cppreference.com/w/cpp/memory/unique_ptr>
+
+std::vector<std::string> RECORDS;
+
+class Resource {
+    std::string _records;
+
+public:
+    void record(char record) {
+        _records.push_back(record);
+    }
+
+    ~Resource() {
+        RECORDS.push_back(_records);
+    }
+};
+
+using Unique = std::unique_ptr<Resource>;
+Unique reset(Unique ptr) {
+    if (ptr) ptr->record('r');
+    return std::make_unique<Resource>();
+}
+Unique drop(Unique ptr) {
+    if (ptr) ptr->record('d');
+    return nullptr;
+}
+Unique forward(Unique ptr) {
+    if (ptr) ptr->record('f');
+    return ptr;
+}
+
+int main(int argc, char **argv) {
+    std::vector<std::string> problems[3];
+
+    drop(forward(reset(nullptr)));
+    problems[0] = std::move(RECORDS);
+
+    forward(drop(reset(forward(forward(reset(nullptr))))));
+    problems[1] = std::move(RECORDS);
+
+
+    drop(drop(reset(drop(reset(reset(nullptr))))));
+    problems[2] = std::move(RECORDS);
+
+
+    // ---- 不要修改以上代码 ----
+
+    std::vector<const char *> answers[]{
+        {"fd"},
+        // TODO: 分析 problems[1] 中资源的生命周期，将记录填入 `std::vector`
+        // NOTICE: 此题结果依赖对象析构逻辑，平台相关，提交时以 CI 实际运行平台为准
+        {"ffr", "d"},
+        {"r", "d", "d"},
+    };
+
+    // ---- 不要修改以下代码 ----
+
+    for (auto i = 0; i < 3; ++i) {
+        ASSERT(problems[i].size() == answers[i].size(), "wrong size");
+        for (auto j = 0; j < problems[i].size(); ++j) {
+            ASSERT(std::strcmp(problems[i][j].c_str(), answers[i][j]) == 0, "wrong location");
+        }
+    }
+
+    return 0;
+}
diff --git a/exercises/exercises/31_std_shared_ptr/main.cpp b/exercises/exercises/31_std_shared_ptr/main.cpp
new file mode 100644
index 000000000..d35134457
--- /dev/null
+++ b/exercises/exercises/31_std_shared_ptr/main.cpp
@@ -0,0 +1,45 @@
+#include "../exercise.h"
+#include <memory>
+
+// READ: `std::shared_ptr` <https://zh.cppreference.com/w/cpp/memory/shared_ptr>
+// READ: `std::weak_ptr` <https://zh.cppreference.com/w/cpp/memory/weak_ptr>
+
+// TODO: 将下列 `?` 替换为正确的值
+int main(int argc, char **argv) {
+    auto shared = std::make_shared<int>(10);
+    std::shared_ptr<int> ptrs[]{shared, shared, shared};
+
+    std::weak_ptr<int> observer = shared;
+    ASSERT(observer.use_count() == 4, "");
+
+    ptrs[0].reset();
+    ASSERT(observer.use_count() == 3, "");
+
+    ptrs[1] = nullptr;
+    ASSERT(observer.use_count() == 2, "");
+
+    ptrs[2] = std::make_shared<int>(*shared);
+    ASSERT(observer.use_count() == 1, "");
+
+    ptrs[0] = shared;
+    ptrs[1] = shared;
+    ptrs[2] = std::move(shared);
+    ASSERT(observer.use_count() == 3, "");
+
+    std::ignore = std::move(ptrs[0]);
+    ptrs[1] = std::move(ptrs[1]);
+    ptrs[1] = std::move(ptrs[2]);
+    ASSERT(observer.use_count() == 2, "");
+
+    shared = observer.lock();
+    ASSERT(observer.use_count() == 3, "");
+
+    shared = nullptr;
+    for (auto &ptr : ptrs) ptr = nullptr;
+    ASSERT(observer.use_count() == 0, "");
+
+    shared = observer.lock();
+    ASSERT(observer.use_count() == 0, "");
+
+    return 0;
+}
diff --git a/exercises/exercises/32_std_transform/main.cpp b/exercises/exercises/32_std_transform/main.cpp
new file mode 100644
index 000000000..ad386804c
--- /dev/null
+++ b/exercises/exercises/32_std_transform/main.cpp
@@ -0,0 +1,26 @@
+#include "../exercise.h"
+#include <algorithm>
+#include <string>
+#include <vector>
+
+// READ: `std::transform` <https://zh.cppreference.com/w/cpp/algorithm/transform>
+// READ: `std::vector::begin` <https://zh.cppreference.com/w/cpp/container/vector/begin>
+
+
+
+int main(int argc, char **argv) {
+    std::vector<int> val{8, 13, 21, 34, 55};
+    // TODO: 调用 `std::transform`，将 `v` 中的每个元素乘以 2，并转换为字符串，存入 `ans`
+    // std::vector<std::string> ans
+    std::vector<std::string> ans(val.size());
+    std::transform(val.begin(), val.end(), ans.begin(), [](int x) {
+        return std::to_string(x * 2);
+    });
+    ASSERT(ans.size() == val.size(), "ans size should be equal to val size");
+    ASSERT(ans[0] == "16", "ans[0] should be 16");
+    ASSERT(ans[1] == "26", "ans[1] should be 26");
+    ASSERT(ans[2] == "42", "ans[2] should be 42");
+    ASSERT(ans[3] == "68", "ans[3] should be 68");
+    ASSERT(ans[4] == "110", "ans[4] should be 110");
+    return 0;
+}
diff --git a/exercises/exercises/33_std_accumulate/main.cpp b/exercises/exercises/33_std_accumulate/main.cpp
new file mode 100644
index 000000000..53e9af330
--- /dev/null
+++ b/exercises/exercises/33_std_accumulate/main.cpp
@@ -0,0 +1,19 @@
+#include "../exercise.h"
+#include <numeric>
+
+// READ: `std::accumulate` <https://zh.cppreference.com/w/cpp/algorithm/accumulate>
+
+int main(int argc, char **argv) {
+    using DataType = float;
+    int shape[]{1, 3, 224, 224};
+    // TODO: 调用 `std::accumulate` 计算：
+    //       - 数据类型为 float；
+    //       - 形状为 shape；
+    //       - 连续存储；
+    //       的张量占用的字节数
+    // int size =
+    int size = sizeof(DataType) * std::accumulate(
+                                      shape, shape + sizeof(shape) / sizeof(int), 1, std::multiplies<int>());
+    ASSERT(size == 602112, "4x1x3x224x224 = 602112");
+    return 0;
+}
diff --git a/exercises/exercises/exercise.h b/exercises/exercises/exercise.h
new file mode 100644
index 000000000..38f87dc46
--- /dev/null
+++ b/exercises/exercises/exercise.h
@@ -0,0 +1,19 @@
+#ifndef __EXERCISE_H__
+#define __EXERCISE_H__
+
+#include <iostream>
+
+#define ASSERT(COND, MSG)                                                                         \
+    if (!(COND)) {                                                                                \
+        std::cerr << "\x1b[31mAssertion failed at line #" << __LINE__ << ": \x1b[0m" << std::endl \
+                  << std::endl                                                                    \
+                  << #COND << std::endl                                                           \
+                  << std::endl                                                                    \
+                  << "\x1b[34mMessage:\x1b[0m" << std::endl                                       \
+                  << std::endl                                                                    \
+                  << MSG << std::endl                                                             \
+                  << std::endl;                                                                   \
+        exit(1);                                                                                  \
+    }
+
+#endif// __EXERCISE_H__
diff --git a/exercises/exercises/xmake.lua b/exercises/exercises/xmake.lua
new file mode 100644
index 000000000..0c3b11bfd
--- /dev/null
+++ b/exercises/exercises/xmake.lua
@@ -0,0 +1,143 @@
+add_rules("mode.debug", "mode.release")
+set_encodings("utf-8")
+set_warnings("all")
+set_kind("binary")
+set_languages("cxx17")
+
+-- 格式化输出
+target("exercise00")
+    add_files("00_hello_world/main.cpp")
+
+-- 变量与运算符
+target("exercise01")
+    add_files("01_variable&add/main.cpp")
+
+-- 函数、声明和定义
+target("exercise02")
+    add_files("02_function/main.cpp")
+
+-- 形参实参
+target("exercise03")
+    add_files("03_argument&parameter/main.cpp")
+
+-- static 关键字
+target("exercise04")
+    add_files("04_static/main.cpp")
+
+-- constexpr 编译时常量和运算
+target("exercise05")
+    add_files("05_constexpr/main.cpp")
+
+-- 数组
+target("exercise06")
+    add_files("06_array/main.cpp")
+
+-- 循环
+target("exercise07")
+    add_files("07_loop/main.cpp")
+
+-- 指针
+target("exercise08")
+    add_files("08_pointer/main.cpp")
+
+-- 枚举/联合体
+target("exercise09")
+    add_files("09_enum&union/main.cpp")
+
+-- “普通”类型
+target("exercise10")
+    add_files("10_trivial/main.cpp")
+
+-- 方法
+target("exercise11")
+    add_files("11_method/main.cpp")
+
+-- const 修饰方法
+target("exercise12")
+    add_files("12_method_const/main.cpp")
+
+-- 类
+target("exercise13")
+    add_files("13_class/main.cpp")
+
+-- 析构器
+target("exercise14")
+    add_files("14_class_destruct/main.cpp")
+
+-- 复制构造函数
+target("exercise15")
+    add_files("15_class_clone/main.cpp")
+
+-- 移动语义
+target("exercise16")
+    add_files("16_class_move/main.cpp")
+
+-- 派生
+target("exercise17")
+    add_files("17_class_derive/main.cpp")
+
+-- 虚函数
+target("exercise18")
+    add_files("18_class_virtual/main.cpp")
+
+-- 虚析构函数
+target("exercise19")
+    add_files("19_class_virtual_destruct/main.cpp")
+
+-- 函数模板
+target("exercise20")
+    add_files("20_function_template/main.cpp")
+
+-- 习题：用于编译器的运行时类型
+target("exercise21")
+    add_files("21_runtime_datatype/main.cpp")
+
+-- 类模板
+target("exercise22")
+    add_files("22_class_template/main.cpp")
+
+-- 模板非类型实参
+target("exercise23")
+    add_files("23_template_const/main.cpp")
+
+-- std::array
+target("exercise24")
+    add_files("24_std_array/main.cpp")
+
+-- std::vector
+target("exercise25")
+    add_files("25_std_vector/main.cpp")
+
+-- std::vector<bool>
+target("exercise26")
+    add_files("26_std_vector_bool/main.cpp")
+
+-- 习题：步长计算
+target("exercise27")
+    add_files("27_strides/main.cpp")
+
+-- std::string
+target("exercise28")
+    add_files("28_std_string/main.cpp")
+
+-- std::map
+target("exercise29")
+    add_files("29_std_map/main.cpp")
+
+-- std::transform
+target("exercise30")
+    add_files("30_std_unique_ptr/main.cpp")
+
+-- std::accumulate
+target("exercise31")
+    add_files("31_std_shared_ptr/main.cpp")
+
+-- std::transform
+target("exercise32")
+    add_files("32_std_transform/main.cpp")
+
+-- std::accumulate
+target("exercise33")
+    add_files("33_std_accumulate/main.cpp")
+
+-- TODO: lambda; deque; forward_list; fs; thread; mutex;