Update app.py

src/app.py

@@ -231,7 +231,7 @@ if 'feedback_given' not in st.session_state:
 # Streamlit Layout
 # -----------------
 st.markdown(
-    "<h1 style='text-align: center;'> 🕵️ Adaptive LLM-texts Detector</h1>",
+    "<h1 style='text-align: center;'> Detect AI-Generated Texts 🕵️ </h1>",
     unsafe_allow_html=True,
 )
 
@@ -249,9 +249,10 @@ if not model_loaded:
 # -----------------
 # --- Two columns: Input text & button | Result displays ---
 text_input = st.text_area(
-    label="",
+    label="📝 Input Text to be Detected",
     placeholder="Paste your text to be detected here. Typically, providing text with a longer content would get a more reliable result.",
     height=240,
+    label_visibility="hidden",
 )
 
 subcol11, subcol12, subcol13 = st.columns((1, 1, 1))
@@ -267,7 +268,7 @@ selected_domain = subcol11.selectbox(
 detect_clicked = subcol12.button("🔍 Detect", type="primary", use_container_width=True)
 
 selected_level = subcol13.slider(
-    label="Significance (α)",
+    label="Significance level (α)",
     min_value=0.01,
     max_value=0.2,
     value=0.05,
@@ -275,33 +276,33 @@ selected_level = subcol13.slider(
     # label_visibility="collapsed",
 )
 
-col2, col3, col4 = st.columns((1, 1, 2))
-with col2:
-    statistics_ph = st.empty()
-    statistics_ph.text_input(
-        label="Statistic",
-        value="",
-        placeholder="",
-        disabled=True,
-    )
+# col2, col3, col4 = st.columns((1, 1, 2))
+# with col2:
+#     statistics_ph = st.empty()
+#     statistics_ph.text_input(
+#         label="Statistic",
+#         value="",
+#         placeholder="",
+#         disabled=True,
+#     )
 
-with col3:
-    pvalue_ph = st.empty()
-    pvalue_ph.text_input(
-        label="p-value",
-        value="",
-        placeholder="",
-        disabled=True,
-    )
+# with col3:
+#     pvalue_ph = st.empty()
+#     pvalue_ph.text_input(
+#         label="p-value",
+#         value="",
+#         placeholder="",
+#         disabled=True,
+#     )
 
-with col4:
-    conclusion_ph = st.empty()
-    conclusion_ph.text_input(
-        label="Conclusion",
-        value="",
-        placeholder="",
-        disabled=True,
-    )
+# with col4:
+#     conclusion_ph = st.empty()
+#     conclusion_ph.text_input(
+#         label="Conclusion",
+#         value="",
+#         placeholder="",
+#         disabled=True,
+#     )
 
 # -----------------
 # Detection Logic
@@ -347,36 +348,67 @@ if detect_clicked:
                 'elapsed_time': elapsed_time
             }
 
-            # Update score displays
-            statistics_ph.text_input(
-                label="Statistics",
-                value=f"{crit:.6f}",
-                disabled=True,
-                help="Detection statistics will appear here after clicking Detect.",
-            )
+            # # Update score displays
+            # statistics_ph.text_input(
+            #     label="Statistics",
+            #     value=f"{crit:.6f}",
+            #     disabled=True,
+            # )
             
-            pvalue_ph.text_input(
-                label="p-value",
-                value=f"{p_value:.6f}",
-                disabled=True,
-                help="p-value will appear here after clicking Detect.",
-            )
-
-            conclusion_ph.text_input(
-                label="Conclusion",
-                value="Reject H0: Text is likely LLM-generated." if p_value < selected_level else "Fail to Reject H0: Text is likely human-written.",
-                disabled=True,
-                help="Conclusion will appear here after clicking Detect.",
-            )
+            # pvalue_ph.text_input(
+            #     label="p-value",
+            #     value=f"{p_value:.6f}",
+            #     disabled=True,
+            # )
+
+            # conclusion_ph.text_input(
+            #     label="Conclusion",
+            #     value="Reject H0: Text is likely LLM-generated." if p_value < selected_level else "Fail to Reject H0: Text is likely human-written.",
+            #     disabled=True,
+            # )
 
             st.info(
-                """
-                **📊 p-value:**
-                - **Lower p-value** (closer to 0) indicates text is **more likely AI-generated**
-                - **Higher p-value** (closer to 1) indicates text is **more likely human-written**
+                f"""
+                **Conclusion**: 
+
+                    {'Text is likely LLM-generated.' if p_value < selected_level else 'Fail to reject hypothesis that text is human-written.'}
+
+                based on the observation that $p$-value {p_value:.3f} is {'less' if p_value < selected_level else 'greater'} than significance level {selected_level:.2f} 📊
                 """,
                 icon="💡"
             )
+            st.markdown(
+                """
+                <style>
+                /* Tighten spacing inside Clarification / Citation expanders */
+                div[data-testid="stExpander"] {
+                    margin-top: -1.3rem;
+                }
+                div[data-testid="stExpander"] p,
+                div[data-testid="stExpander"] li {
+                    line-height: 1.35;
+                    margin-bottom: 0.1rem;
+                }
+
+                div[data-testid="stExpander"] ul {
+                    margin-top: 0.1rem;
+                }
+                </style>
+                """,
+                unsafe_allow_html=True
+            )
+            with st.expander("📋 Illustration and Suggestions"):
+                st.markdown(
+                    """
+                    + Illustration:
+                        - $p$-value: Lower $p$-value (closer to 0) indicates text is **more likely AI-generated**; Higher $p$-value (closer to 1) indicates text is **more likely human-written**.
+                        - Significance Level (α): a threshold set by the user to determine the sensitivity of the detection. Lower α means stricter criteria for claiming the text is AI-generated.
+
+                    + Suggestions for better detection:
+                        - Provide longer text inputs for more reliable detection results.
+                        - Select the domain that best matches the content of your text to improve detection accuracy.
+                    """
+                )
 
             # ========== 🆕 Feedback buttons (moved here for better UX) ==========
             st.markdown("**📝 Result Feedback**: Does this detection result meet your expectations?")
@@ -440,20 +472,53 @@ if detect_clicked:
             st.error(f"❌ Error during detection: {str(e)}")
             st.exception(e)
 
-# ========== 🆕 Citation and paper reference section ==========
-# st.markdown("---")
+
+# st.markdown("<div style='height: 4rem;'></div>", unsafe_allow_html=True)
 # st.markdown(
 #     """
-#      📄 **Citation** If you find this tool useful for you, please cite our paper: **[AdaDetectGPT: Adaptive Detection of LLM-Generated Text with Statistical Guarantees](https://arxiv.org/abs/2510.01268)**
-#     """
+#     <style>
+#     /* Tighten spacing inside Clarification / Citation expanders */
+#     div[data-testid="stExpander"] p,
+#     div[data-testid="stExpander"] li {
+#         line-height: 1.35;
+#         margin-bottom: 0.3rem;
+#     }
+
+#     div[data-testid="stExpander"] ul {
+#         margin-top: 0.3rem;
+#         margin-bottom: 0.3rem;
+#     }
+#     </style>
+#     """,
+#     unsafe_allow_html=True
 # )
-# with st.expander("📋 BibTeX Citation"):
+# with st.expander("📋 Illustration and Suggestions"):
+#     st.markdown(
+#         """
+#         + Illustration:
+#             - Statistic: a numerical measure computed from the input text. Higher values typically indicate a greater likelihood of being AI-generated.
+#             - p-value: Lower p-value (closer to 0) indicates text is **more likely AI-generated**; Higher p-value (closer to 1) indicates text is **more likely human-written**.
+#             - Conclusion: when the p-value is less than the significance level (α), the text is classified as AI-generated.
+#             - Significance Level (α): a threshold set by the user to determine the sensitivity of the detection. Lower α means stricter criteria for claiming the text is AI-generated.
+
+#         + Suggestions for better detection:
+#             - Provide longer text inputs for more reliable detection results.
+#             - Select the domain that best matches the content of your text to improve detection accuracy.
+#         """
+#     )
+
+# with st.expander("📋 Citation"):
+#     st.markdown(
+#         """
+#         If you find this tool useful for you, please cite our paper: **[AdaDetectGPT: Adaptive Detection of LLM-Generated Text with Statistical Guarantees](https://arxiv.org/abs/2510.01268)**
+#         """
+#     )
 #     st.code(
 #         """
 # @inproceedings{zhou2024adadetectgpt,
 #   title={AdaDetectGPT: Adaptive Detection of LLM-Generated Text with Statistical Guarantees},
 #   author={Hongyi Zhou and Jin Zhu and Pingfan Su and Kai Ye and Ying Yang and Shakeel A O B Gavioli-Akilagun and Chengchun Shi},
-#   booktitle={The Thirty-Ninth Annual Conference on Neural Information Processing Systems (Accepted)},
+#   booktitle={The Thirty-Ninth Annual Conference on Neural Information Processing Systems},
 #   year={2025},
 # }
 #         """,
@@ -485,7 +550,7 @@ st.markdown(
     }
     </style>
     <div class='footer'>
-        <small>Powered by Adaptive LLM Text Detection | For research purposes only</small>
+        <small> This tool is developed for research purposes only. The detection results are not 100% accurate and should not be used as the sole basis for any critical decisions. Users are advised to use this tool responsibly and ethically. </small>
     </div>
     """,
     unsafe_allow_html=True