diff --git a/docs.html b/docs.html
index daae044..f9f9f7d 100644
--- a/docs.html
+++ b/docs.html
@@ -108,7 +108,6 @@
               <a href="docs.html#add-pass" class="docs-nav__child" data-doc="add-pass" data-i18n="doc.addPass">添加 Pass</a>
               <a href="docs.html#build-methods" class="docs-nav__child" data-doc="build-methods" data-i18n="doc.buildMethods">构建方法</a>
               <a href="docs.html#dynamic-vector" class="docs-nav__child" data-doc="dynamic-vector" data-i18n="doc.dynamicVector">动态向量表示</a>
-              <a href="docs.html#conv-opt" class="docs-nav__child" data-doc="conv-opt" data-i18n="doc.convOpt">卷积优化</a>
               <a href="docs.html#adding-operators" class="docs-nav__child" data-doc="adding-operators" data-i18n="doc.addingOperators">增加算子支持</a>
               <a href="docs.html#convolution-vectorization" class="docs-nav__child" data-doc="convolution-vectorization" data-i18n="doc.convolutionVectorization">卷积向量化</a>
             </div>
diff --git a/docs.js b/docs.js
index 4c76386..c8c210d 100644
--- a/docs.js
+++ b/docs.js
@@ -103,14 +103,6 @@
       markdownUrl: "docs/compiler/DynamicVector.md",
       bodyOnly: true
     },
-    "conv-opt": {
-      title: "卷积优化",
-      titleEn: "Convolution Optimization",
-      desc: "卷积优化工具 conv-opt：系数广播算法与所需 Vector/affine 操作概览。",
-      descEn: "Convolution optimization tool conv-opt: coefficients broadcasting algorithm and required Vector/affine ops.",
-      markdownUrl: "docs/compiler/conv-opt.md",
-      bodyOnly: true
-    },
     "adding-operators": {
       title: "增加算子支持",
       titleEn: "Adding Operators",
diff --git a/docs/compiler/conv-opt-CoefficientsBroadcasting.png b/docs/compiler/conv-opt-CoefficientsBroadcasting.png
deleted file mode 100644
index b748c4d..0000000
Binary files a/docs/compiler/conv-opt-CoefficientsBroadcasting.png and /dev/null differ
diff --git a/docs/compiler/conv-opt.en.md b/docs/compiler/conv-opt.en.md
deleted file mode 100644
index 224f0dd..0000000
--- a/docs/compiler/conv-opt.en.md
+++ /dev/null
@@ -1,29 +0,0 @@
-# Convolution Optimization Tool - conv - opt
-
-## Covolution Optimization Algorithm
-
-### Coefficient Broadcasting Algorithm
-
-Coefficients broadcasting (CB) algorithm is a high-performance implementation of 2D convolution.
-The algorithm has been implemented with the intrinsic programming model and evaluated on Intel’s advanced vector extensions. [[Ref]](https://ieeexplore.ieee.org/abstract/document/8324097)
-However, the intrinsic programming model is specific to ISA extensions,
-which means other targets cannot reuse the vectorization.
-Thus, we explored implementing the CB algorithm using MLIR.
-In this case, we can take advantage of the reusable and extensible features of MLIR and LLVM IR.
-
-Let’s first introduce the MLIR dialects and operations that are needed by the algorithm.
-
-- `affine.for`: executes loop body for specific times iterating from a lower bound to an upper bound by a stride.
-- `affine.vector_load`:returns a vector from a specific slice of buffer (MLIR MemRef).
-- `affine.vector_store`: writes a vector into a specific slice of buffer (MLIR MemRef).
-- `vector.broadcast`: broadcasts a scalar or vector value to an n-D result vector.
-- `vector.fma`: performs fused multiply-add (FMA) on the vector type.
-
-![Graph of the Coefficients Broadcasting Algorithm](/docs/compiler/conv-opt-CoefficientsBroadcasting.png)
-
-The figure shows the steps of the algorithm:
-- Iterate over each kernel element and broadcast it to vector1.
-- Load a slice of the input into the vector2 with the iteration index.
-- Load a slice of the output into the vector3 with the outermost loop iteration index.
-- Multiply and accumulate the vector1, vector2, and vector3.
-- Store the result vector into the output buffer with the outermost loop iteration index.
diff --git a/docs/compiler/conv-opt.md b/docs/compiler/conv-opt.md
deleted file mode 100644
index 34639b6..0000000
--- a/docs/compiler/conv-opt.md
+++ /dev/null
@@ -1,29 +0,0 @@
-# 卷积优化工具 conv-opt
-
-## 卷积优化算法
-
-### 系数广播（Coefficient Broadcasting）算法
-
-系数广播（Coefficients Broadcasting, CB）算法是一种高性能的二维卷积实现方式。  
-该算法最初基于 intrinsic 编程模型实现，并在 Intel 的高级向量扩展上进行了性能评估。[[参考]](https://ieeexplore.ieee.org/abstract/document/8324097)  
-然而，intrinsic 编程模型强烈依赖具体 ISA 扩展，这意味着其他目标架构难以复用同样的向量化实现。
-
-因此，我们尝试基于 MLIR 实现 CB 算法，  
-从而充分利用 MLIR 与 LLVM IR 在可复用性与可扩展性方面的优势。
-
-在介绍算法细节之前，先简要列出算法所依赖的一些 MLIR 方言与算子：
-
-- `affine.for`：带上下界与步长的循环结构，用于执行多重循环。
-- `affine.vector_load`：从缓冲区（MLIR `memref`）的指定切片中读取一个向量。
-- `affine.vector_store`：将一个向量写入缓冲区的指定切片。
-- `vector.broadcast`：将标量或向量广播为 n 维结果向量。
-- `vector.fma`：在向量类型上执行融合乘加（FMA）运算。
-
-![Graph of the Coefficients Broadcasting Algorithm](/docs/compiler/conv-opt-CoefficientsBroadcasting.png)
-
-上图展示了系数广播算法的大致执行流程：
-- 遍历卷积核中的每个元素，并将其广播到向量 `vector1`；
-- 按当前迭代索引，从输入特征图中加载一段切片到 `vector2`；
-- 按最外层循环索引，从输出特征图中加载一段切片到 `vector3`；
-- 对 `vector1`、`vector2` 和 `vector3` 进行乘加累积；
-- 将结果向量写回到输出缓冲区对应的位置。
diff --git a/i18n.js b/i18n.js
index a8db1fd..8e65fb9 100644
--- a/i18n.js
+++ b/i18n.js
@@ -40,7 +40,6 @@
       "doc.addPass": "添加 Pass",
       "doc.buildMethods": "构建方法",
       "doc.dynamicVector": "动态向量表示",
-      "doc.convOpt": "卷积优化",
       "doc.convolutionVectorization": "卷积向量化",
       "doc.addingOperators": "增加算子支持",
       "doc.operatorLang": "RuyiAI 算子编程语言",
@@ -96,7 +95,6 @@
       "doc.addPass": "Add Pass",
       "doc.buildMethods": "Build Methods",
       "doc.dynamicVector": "Dynamic Vector Representation",
-      "doc.convOpt": "Convolution Optimization",
       "doc.convolutionVectorization": "Convolution Vectorization",
       "doc.addingOperators": "Adding Operators",
       "doc.operatorLang": "RuyiAI Operator Language",