app.py

app.py

@@ -164,26 +164,13 @@ def calculate_spelling_error_density(sentences):
 
 # ======================== Leaderboard数据处理 ========================
 def load_leaderboard_data():
-    """从JSON加载Leaderboard数据并添加类型分类"""
+    """从JSON加载Leaderboard数据（现在包含Category字段）"""
     json_path = "leaderboard.json"
     try:
         with open(json_path, 'r', encoding='utf-8') as f:
             data = json.load(f)
         
-        # 添加类型分类 (示例分类规则)
-        for item in data:
-            benchmark_name = item['Benchmark'].lower()
-            if 'mmlu' in benchmark_name or 'arc' in benchmark_name:
-                item['Type'] = 'A'
-            elif 'gsm' in benchmark_name or 'math' in benchmark_name:
-                item['Type'] = 'B'
-            elif 'med' in benchmark_name or 'bio' in benchmark_name:
-                item['Type'] = 'C'
-            elif 'code' in benchmark_name or 'human' in benchmark_name:
-                item['Type'] = 'D'
-            else:
-                item['Type'] = 'E'
-        
+        # Category已经在JSON中定义，直接加载即可
         return pd.DataFrame(data)
     except Exception as e:
         print(f"Error loading leaderboard: {e}")
@@ -197,11 +184,11 @@ def make_clickable_download(download_text):
     return download_text
 
 def filter_leaderboard(df, query):
-    """根据类型筛选Leaderboard"""
+    """根据Category筛选Leaderboard"""
     if query == "all":
         return df
     else:
-        return df[df['Type'] == query]
+        return df[df['Category'] == query]
 
 def search_leaderboard(df, query):
     """搜索Leaderboard"""
@@ -395,6 +382,15 @@ ABOUT_TEXT = """
 - **WAR (Whitespace Anomaly Rate)**: 空白符异常率
 - **SED (Spelling Error Density)**: 拼写错误密度
 
+### 数据集分类
+
+- **BT (Basic Tasks)**: 基础任务 - MRPC, RTE, SST2
+- **RA (Reasoning Abilities)**: 推理能力 - ARC, GSM8K, MMLU
+- **TG (Text Generation)**: 文本生成 - CoQA, DROP, Truthful_QA
+- **SU (Speech Understanding)**: 语音理解 - WNLI, Natural_questions
+- **ME (Medical)**: 医学领域 - MedMCQA, MedQA, PubMedQA
+- **GR (Grammatical)**: 语法领域 - BEA-2019, CoNLL-2014
+
 ### 使用说明
 
 1. **配置 API Key**: Settings → Repository secrets → `DEEPSEEK_API_KEY`
@@ -427,11 +423,12 @@ SUBMISSION_TEXT = """
 ### 数据格式要求
 
 提交的数据需要包含以下字段:
+- ID: 序号
+- Category: 类别 (BT/RA/TG/SU/ME/GR)
 - Benchmark名称
-- 去噪方法
 - WAR (%)
 - SED
-- 下载链接
+- Download: 下载链接
 
 ### 联系方式
 
@@ -467,17 +464,17 @@ with demo:
                         show_label=False,
                         elem_id="search-bar",
                     )
-                    filter_types = gr.Radio(
-                        label="⏚ 筛选Benchmark类型",
-                        choices=["all", "A", "B", "C", "D", "E"],
+                    filter_categories = gr.Radio(
+                        label="📂 筛选Benchmark类别",
+                        choices=["all", "BT", "RA", "TG", "SU", "ME", "GR"],
                         value="all",
                         elem_id="filter-columns",
                     )
                 
                 leaderboard_table = gr.Dataframe(
-                    value=leaderboard_data[['ID', 'Benchmark', 'WAR', 'SED', 'Download']],
-                    headers=['ID', 'Benchmark', 'WAR (%)', 'SED', '下载'],
-                    datatype=['number', 'str', 'number', 'number', 'markdown'],
+                    value=leaderboard_data[['ID', 'Category', 'Benchmark', 'WAR', 'SED', 'Download']],
+                    headers=['ID', 'Category', 'Benchmark', 'WAR (%)', 'SED', '下载'],
+                    datatype=['number', 'str', 'str', 'number', 'number', 'markdown'],
                     elem_id="leaderboard-table",
                     interactive=False,
                 )
@@ -489,23 +486,23 @@ with demo:
                 
                 # 绑定搜索和筛选
                 search_bar.submit(
-                    lambda df, query: search_leaderboard(df, query)[['ID', 'Benchmark', 'WAR', 'SED', 'Download']],
+                    lambda df, query: search_leaderboard(df, query)[['ID', 'Category', 'Benchmark', 'WAR', 'SED', 'Download']],
                     [hidden_leaderboard, search_bar],
                     leaderboard_table
                 )
                 
-                filter_types.change(
-                    lambda df, query: filter_leaderboard(df, query)[['ID', 'Benchmark', 'WAR', 'SED', 'Download']],
-                    [hidden_leaderboard, filter_types],
+                filter_categories.change(
+                    lambda df, query: filter_leaderboard(df, query)[['ID', 'Category', 'Benchmark', 'WAR', 'SED', 'Download']],
+                    [hidden_leaderboard, filter_categories],
                     leaderboard_table
                 )
                 
                 gr.Markdown("""
                 **说明:**
-                - WAR: 空白符异常率变化 (正值表示改善)
-                - SED: 拼写错误密度变化 (负值表示改善)
+                - **Category**: BT=基础任务, RA=推理能力, TG=文本生成, SU=语音理解, ME=医学领域, GR=语法领域
+                - **WAR**: 空白符异常率变化 (正值表示改善)
+                - **SED**: 拼写错误密度变化 (负值表示改善)
                 - 绿色: 正向提升 | 红色: 负向影响
-                - 类型分类: A=知识问答, B=数学推理, C=医学领域, D=代码生成, E=其他
                 """, elem_classes="markdown-text")
         
         # ==================== Tab 2: Performance Plot ====================

leaderboard.json

@@ -1,6 +1,7 @@
 [
     {
         "ID": 1,
+        "Category": "RA",
         "Benchmark": "ARC_deepseek_r1_denoising",
         "WAR": 0.00,
         "SED": 0.67,
@@ -8,6 +9,7 @@
     },
     {
         "ID": 2,
+        "Category": "RA",
         "Benchmark": "ARC_wac_gec",
         "WAR": 0.00,
         "SED": 0.66,
@@ -15,6 +17,7 @@
     },
     {
         "ID": 3,
+        "Category": "TG",
         "Benchmark": "COQA_deepseek_r1_denoising",
         "WAR": 4.18,
         "SED": 2.57,
@@ -22,6 +25,7 @@
     },
     {
         "ID": 4,
+        "Category": "TG",
         "Benchmark": "COQA_wac_gec",
         "WAR": 4.70,
         "SED": 2.56,
@@ -29,6 +33,7 @@
     },
     {
         "ID": 5,
+        "Category": "TG",
         "Benchmark": "DROP_deepseek_r1_denoising",
         "WAR": 0.02,
         "SED": 3.24,
@@ -36,6 +41,7 @@
     },
     {
         "ID": 6,
+        "Category": "TG",
         "Benchmark": "DROP_wac_gec",
         "WAR": 0.64,
         "SED": 3.25,
@@ -43,27 +49,31 @@
     },
     {
         "ID": 7,
+        "Category": "BT",
         "Benchmark": "MRPC_deepseek_r1_denoising",
         "WAR": 3.80,
         "SED": 4.70,
         "Download": "[下载](https://huggingface.co/datasets/lllouo/BD-benchmarks/tree/main/GLUE/glue_deepseek_r1_denoising/mrpc)"
     },
-        {
+    {
         "ID": 8,
+        "Category": "BT",
         "Benchmark": "MRPC_wac_gec",
         "WAR": 1.84,
         "SED": 4.50,
         "Download": "[下载](https://huggingface.co/datasets/lllouo/BD-benchmarks/tree/main/GLUE/glue_wac_gec/mrpc)"
     },
-        {
+    {
         "ID": 9,
+        "Category": "BT",
         "Benchmark": "RTE_deepseek_r1_denoising",
         "WAR": 0.36,
         "SED": 4.50,
         "Download": "[下载](https://huggingface.co/datasets/lllouo/BD-benchmarks/tree/main/GLUE/glue_deepseek_r1_denoising/rte)"
     },
-        {
+    {
         "ID": 10,
+        "Category": "BT",
         "Benchmark": "RTE_wac_gec",
         "WAR": 0.72,
         "SED": 4.43,
@@ -71,13 +81,15 @@
     },
     {
         "ID": 11,
+        "Category": "BT",
         "Benchmark": "SST2_deepseek_r1_denoising",
         "WAR": 7.22,
         "SED": 3.66,
         "Download": "[下载](https://huggingface.co/datasets/lllouo/BD-benchmarks/tree/main/GLUE/glue_deepseek_r1_denoising/sst2)"
     },
-        {
+    {
         "ID": 12,
+        "Category": "BT",
         "Benchmark": "SST2_wac_gec",
         "WAR": 5.39,
         "SED": 3.52,
@@ -85,13 +97,15 @@
     },
     {
         "ID": 13,
+        "Category": "SU",
         "Benchmark": "WNLI_deepseek_r1_denoising",
         "WAR": 0.00,
         "SED": 0.59,
         "Download": "[下载](https://huggingface.co/datasets/lllouo/BD-benchmarks/tree/main/GLUE/glue_deepseek_r1_denoising/wnli)"
     },
-        {
+    {
         "ID": 14,
+        "Category": "SU",
         "Benchmark": "WNLI_wac_gec",
         "WAR": 0.00,
         "SED": 0.64,
@@ -99,13 +113,15 @@
     },
     {
         "ID": 15,
+        "Category": "RA",
         "Benchmark": "GSM8K_deepseek_r1_denoising",
         "WAR": 0.30,
         "SED": 1.13,
         "Download": "[下载](https://huggingface.co/datasets/lllouo/BD-benchmarks/tree/main/GSM8K/gsm8k_deepseek_r1_denoising)"
     },
-        {
+    {
         "ID": 16,
+        "Category": "RA",
         "Benchmark": "GSM8K_wac_gec",
         "WAR": 1.97,
         "SED": 1.11,
@@ -113,13 +129,15 @@
     },
     {
         "ID": 17,
+        "Category": "RA",
         "Benchmark": "MMLU_deepseek_r1_denoising",
         "WAR": 6.56,
         "SED": 2.15,
         "Download": "[下载](https://huggingface.co/datasets/lllouo/BD-benchmarks/tree/main/MMLU/mmlu_deepseek_r1_denoising)"
     },
-        {
+    {
         "ID": 18,
+        "Category": "RA",
         "Benchmark": "MMLU_wac_gec",
         "WAR": 2.98,
         "SED": 2.08,
@@ -127,13 +145,15 @@
     },
     {
         "ID": 19,
+        "Category": "ME",
         "Benchmark": "MedMCQA_deepseek_r1_denoising",
         "WAR": 3.44,
         "SED": 5.70,
         "Download": "[下载](https://huggingface.co/datasets/lllouo/BD-benchmarks/tree/main/MedMCQA/medmcqa_deepseek_r1_denoising)"
     },
-        {
+    {
         "ID": 20,
+        "Category": "ME",
         "Benchmark": "MedMCQA_wac_gec",
         "WAR": 2.44,
         "SED": 5.91,
@@ -141,13 +161,15 @@
     },
     {
         "ID": 21,
+        "Category": "ME",
         "Benchmark": "MedQA_deepseek_r1_denoising",
         "WAR": 16.26,
         "SED": 6.49,
         "Download": "[下载](https://huggingface.co/datasets/lllouo/BD-benchmarks/tree/main/MedQA/MedQA_deepseek_r1_denoising)"
     },
-        {
+    {
         "ID": 22,
+        "Category": "ME",
         "Benchmark": "MedQA_wac_gec",
         "WAR": 0.79,
         "SED": 6.51,
@@ -155,13 +177,15 @@
     },
     {
         "ID": 23,
+        "Category": "SU",
         "Benchmark": "Natural_questions_deepseek_r1_denoising",
         "WAR": 0.06,
         "SED": 3.06,
         "Download": "[下载](https://huggingface.co/datasets/lllouo/BD-benchmarks/tree/main/Natural_questions/nq_open_deepseek_r1_denoising)"
     },
-        {
+    {
         "ID": 24,
+        "Category": "SU",
         "Benchmark": "Natural_questions_wac_gec",
         "WAR": 0.28,
         "SED": 2.93,
@@ -169,13 +193,15 @@
     },
     {
         "ID": 25,
+        "Category": "ME",
         "Benchmark": "PubMedQA_deepseek_r1_denoising",
         "WAR": 0.20,
         "SED": 8.19,
         "Download": "[下载](https://huggingface.co/datasets/lllouo/BD-benchmarks/tree/main/PubMedQA/pubmed_qa_deepseek_r1_denoising)"
     },
-        {
+    {
         "ID": 26,
+        "Category": "ME",
         "Benchmark": "PubMedQA_wac_gec",
         "WAR": 0.00,
         "SED": 8.10,
@@ -183,13 +209,15 @@
     },
     {
         "ID": 27,
+        "Category": "TG",
         "Benchmark": "Truthful_QA_deepseek_r1_denoising",
         "WAR": 0.00,
         "SED": 1.73,
         "Download": "[下载](https://huggingface.co/datasets/lllouo/BD-benchmarks/tree/main/Truthful_QA/truthful_qa_deepseek_r1_denoising)"
     },
-        {
+    {
         "ID": 28,
+        "Category": "TG",
         "Benchmark": "Truthful_QA_wac_gec",
         "WAR": 0.00,
         "SED": 1.53,