Upload 2 files

app.py

@@ -0,0 +1,140 @@
+import gradio as gr
+import pandas as pd
+from constants import *
+
+# ... 其他导入 ...
+
+# 定义自定义 CSS
+custom_css = """
+h1  { /* 根据需要选择正确的标题标签 */
+    background-color: blue; /* 蓝色背景 */
+    color: white; /* 白色文字 */
+    padding: 10px; /* 内边距 */
+    text-align: center; /* 文本居中 */
+}
+h2  { /* 根据需要选择正确的标题标签 */
+    color: white; /* 白色文字 */
+    padding: 10px; /* 内边距 */
+    text-align: center; /* 文本居中 */
+}
+
+"""
+
+def get_preview_data():
+    df = pd.read_json(DATA_DIR)
+    df=df.head(4)
+    return df
+def get_result_data():
+    df={
+    "DataSet": ["WikiData_recent", "WikiData_recent", "WikiData_recent", "WikiData_recent",
+                "ZsRE", "ZsRE", "ZsRE", "ZsRE",
+                "WikiBio", "WikiBio", "WikiBio",
+                "WikiData_counterfact", "WikiData_counterfact", "WikiData_counterfact", "WikiData_counterfact",
+                "ConvSent", "ConvSent", "ConvSent",
+                "Sanitation", "Sanitation", "Sanitation"],
+    "Metric": ["Edit Succ. ↑", "Portability ↑", "Locality ↑", "Fluency ↑",
+               "Edit Succ. ↑", "Portability ↑", "Locality ↑", "Fluency ↑",
+               "Edit Succ. ↑", "Locality ↑", "Fluency ↑",
+               "Edit Succ. ↑", "Portability ↑", "Locality ↑", "Fluency ↑",
+               "Edit Succ. ↑", "Locality ↓", "Fluency ↑",
+               "Edit Succ. ↑", "Locality ↑", "Fluency ↑"],
+    "SERAC": [98.68, 63.52, 100.00, 553.19,
+              99.67, 56.48, 30.23, 410.89,
+              99.69, 69.79, 606.95,
+              99.99, 76.07, 98.96, 549.91,
+              62.75, 0.26, 458.21,
+              0.00, 100.00, 416.29],
+    "ICE": [60.74, 36.93, 33.34, 531.01,
+            66.01, 63.94, 23.14, 541.14,
+            95.53, 47.90, 632.92,
+            69.83, 45.32, 32.38, 547.22,
+            52.78, 49.73, 621.45,
+            72.50, 56.58, 794.15],
+    "AdaLoRA": [65.61, 47.22, 55.78, 537.51,
+                69.86, 52.95, 72.21, 532.82,
+                97.02, 57.87, 615.86,
+                72.14, 55.17, 66.78, 553.85,
+                44.89, 0.18, 606.42,
+                2.50, 65.50, 330.44],
+    "MEND": [76.88, 50.11, 92.87, 586.34,
+             96.74, 60.41, 92.79, 524.33,
+             93.66, 69.51, 609.39,
+             78.82, 57.53, 94.16, 588.94,
+             50.76, 3.42, 379.43,
+             0.00, 5.29, 407.18],
+    "ROME": [85.08, 37.45, 66.2, 574.28,
+             96.57, 52.20, 27.14, 570.47,
+             95.05, 46.96, 617.25,
+             83.21, 38.69, 65.4, 578.84,
+             45.79, 0.00, 606.32,
+             85.00, 50.31, 465.12],
+    "MEMIT": [85.32, 37.94, 64.78, 566.66,
+              83.07, 51.43, 25.46, 559.72,
+              94.29, 51.56, 616.65,
+              83.41, 40.09, 63.68, 568.58,
+              44.75, 0.00, 602.62,
+              48.75, 67.47, 466.10],
+    "FT-L": [71.18, 48.71, 63.7, 549.35,
+             54.65, 45.02, 71.12, 474.18,
+             83.41, 40.09, 63.68, 568.58,
+             66.27, 60.14, 604.00,
+             51.12, 39.07, 62.51,
+             48.75, 67.47, 466.10]
+    }
+    df=pd.DataFrame(df)
+    return df
+
+block = gr.Blocks(css=custom_css)  # 应用自定义 CSS
+
+with block:
+    gr.Markdown(TITLE)
+    
+    gr.Markdown("## BACKGROUND")
+    gr.Markdown(
+        BACKGROUND
+    )
+    gr.Image('./img/demo.gif')
+            
+    gr.Markdown("## DATA PREVIEW")
+    gr.Markdown(LEADERBORAD_INTRODUCTION)
+
+    with gr.Tabs(elem_classes="tab-buttons") as tabs:
+        with gr.TabItem("🏅 Data preview ", elem_id="ke-benchmark-tab-table", id=0):
+            # 创建数据帧组件
+            ke_data_component = gr.components.Dataframe(
+                value=get_preview_data(), 
+                headers=DATA_COLUMN_NAMES,
+                type="pandas", 
+            )
+        with gr.TabItem("data Structure", elem_id="about-struct-tab-table", id=3):
+            gr.Markdown(DATA_STRUCT, elem_classes="markdown-text")
+            
+        with gr.TabItem("📝 data schema", elem_id="about-benchmark-tab-table", id=4):
+            gr.Markdown(DATA_SCHEMA, elem_classes="markdown-text")
+       
+       
+       
+        
+    gr.Markdown("## EXPERIMENT RESULTS")
+    gr.Markdown("We list the results of current knowledge editing methods on Llama2-7b-chat in Table")
+    with gr.Tabs(elem_classes="tab-buttons") as tabs:
+        with gr.TabItem("🏅 result", elem_id="ke-benchmark-tab-table", id=0):
+            # 创建数据帧组件
+            ke_data_component = gr.components.Dataframe(
+                value=get_result_data(), 
+                headers=RESULT_COLUMN_NAMES,
+                type="pandas", 
+            )
+        # About tab
+        with gr.TabItem("📝 About", elem_id="about-benchmark-tab-table", id=4):
+            gr.Markdown("Results of existing knowledge edit methods on the constructed benchmark. The symbol indicates that higher numbers correspond to better performance, while ��� denotes the opposite, with lower numbers indicating better performance. For WikiBio and Convsent, we do not test the portability as they are about specific topics. ", elem_classes="markdown-text")
+            
+    with gr.Row():
+        with gr.Accordion("Citation", open=False):
+            citation_button = gr.Textbox(
+                value=CITATION_BUTTON_TEXT,
+                label=CITATION_BUTTON_LABEL,
+                elem_id="citation-button",
+            ).style(show_copy_button=True)       
+
+block.launch(share=True)

constants.py

@@ -0,0 +1,69 @@
+# this is .py for store constants 
+
+DATA_DIR="./data/data.json"
+
+
+
+MODEL_INFO = ["Model Name", "Language Model"]
+AVG_INFO = ["Avg. All"]
+ME_INFO=["Method Name", "Language Model"]
+
+# KE 固定信息
+KE_Data_INFO = ["FewNERD", "FewRel", "InstructIE-en", "MAVEN","WikiEvents"]
+
+KE_TASK_INFO = ["Avg. All", "FewNERD", "FewRel", "InstructIE-en", "MAVEN","WikiEvents"]
+KE_CSV_DIR = "./ke_files/result-kgc.csv"
+DATA_COLUMN_NAMES =["locality","labels","concept","text"]
+KE_TABLE_INTRODUCTION = """In the table below, we summarize each task performance of all the models. We use F1 score(%) as the primary evaluation metric for each tasks. 
+    """
+RESULT_COLUMN_NAMES= ["DataSet","Metric","Metric","ICE","AdaLoRA","MEND","ROME","MEMIT","FT-L","FT"]
+DATA_STRUCT="""
+Datasets	ZsRE	Wikirecent	Wikicounterfact	WikiBio
+Train	10,000	570	1455	592
+Test	1230	1266	885	1392
+"""
+TITLE = """# KnowEdit: a dataset for knowledge editing"""
+
+BACKGROUND="""
+Large Language Models (LLMs) have shown extraordinary capabilities in understanding and generating text that closely mirrors human communication. However, a primary limitation lies in the significant computational demands during training, arising from their extensive parameterization.There is an increasing interest in efficient, lightweight methods for onthe-fly model modifications. To this end, recent years have seen a burgeoning in the techniques of knowledge editing for LLMs, which aim to efficiently modify LLMs’ behaviors within specific domains while preserving overall performance across various inputs. 
+"""
+
+LEADERBORAD_INTRODUCTION = """
+     This is the dataset for knowledge editing. It contains six tasks: ZsRE, Wiki<sub>recent</sub>, Wiki<sub>counterfact</sub>, WikiBio, ConvSent and Sanitation. This repo shows the former 4 tasks and you can get the data for ConvSent and Sanitation from their original papers.
+    """
+DATA_SCHEMA =""" {
+  "subject": xxx,
+  "target_new": xxx,
+  "prompt": xxx,
+  "portability":{
+      "Logical_Generalization": [],
+      ...
+  }
+  "locality":{
+      "Relation_Specificity": [],
+      ...
+  }
+}"""
+
+
+
+CITATION_BUTTON_LABEL = "Copy the following snippet to cite these results"
+CITATION_BUTTON_TEXT = r"""@article{tan2023evaluation,
+    title={Evaluation of ChatGPT as a question answering system for answering complex questions},
+    author={Yiming Tan and Dehai Min and Yu Li and Wenbo Li and Nan Hu and Yongrui Chen and Guilin Qi},
+    journal={arXiv preprint arXiv:2303.07992},
+    year={2023}
+}
+@article{gui2023InstructIE,
+  author       = {Honghao Gui and Jintian Zhang and Hongbin Ye and Ningyu Zhang},
+  title        = {InstructIE: {A} Chinese Instruction-based Information Extraction Dataset},
+  journal      = {arXiv preprint arXiv:2303.07992},
+  year         = {2023}
+}
+@article{yao2023edit,
+  author       = {Yunzhi Yao and Peng Wang and Bozhong Tian and Siyuan Cheng and Zhoubo Li and Shumin Deng and Huajun Chen and  Ningyu Zhang},
+  title        = {Editing Large Language Models: Problems, Methods, and Opportunities},
+  journal      = {arXiv preprint arXiv:2305.13172},
+  year         = {2023}
+}
+"""