rename avg->avg(qa1-5) and pin model_name column

app.py

@@ -10,7 +10,7 @@ st.markdown(PAGE_MARKDOWN, unsafe_allow_html=True)
 def draw_leaderboard():
     df = load_results()
 
-    tasks = ['avg'] + [f"qa{i}" for i in range(1, 11)]
+    tasks = ['avg(qa1-5)'] + [f"qa{i}" for i in range(1, 11)]
     columns = ["model_name", "≤32k", "≤128k"] + LENGTHS
 
     st.title("🔎📚🪡📚❓ BABILong Leaderboard 🏆")
@@ -21,7 +21,7 @@ def draw_leaderboard():
                 '[BABILong evals](https://huggingface.co/datasets/RMT-team/babilong_evals)')
     search_term = st.text_input("Search models:", "")
 
-    tabs = st.tabs([str(task) for task in tasks])
+    tabs = st.tabs(tasks)
     for i, tab in enumerate(tabs):
         with tab:
             task_df = df[df.task == tasks[i]][columns]
@@ -39,8 +39,11 @@ def draw_leaderboard():
 
             st.dataframe(
                 styled_df,
-                width=1100,
+                width=1070,
                 height=height,
+                column_config={
+                    "model_name": st.column_config.Column(width=260, pinned=True)
+                }
             )
 
 

draw_utils.py

@@ -40,6 +40,8 @@ def load_results():
     res = pd.concat([old_results_filtered, new_results])
     res.drop('normalized_name', axis=1, inplace=True)
 
+    res['task'] = res['task'].str.replace('avg', 'avg(qa1-5)')
+
     res.replace(-1, np.nan, inplace=True)
     res['≤32k'] = res[LENGTHS_32k].mean(axis=1)
     res['≤128k'] = res[LENGTHS_128k].mean(axis=1)