Update app.py

app.py

@@ -18,7 +18,7 @@ def make_default_md(arena_df, elo_results):
 
     leaderboard_md = f"""
 # NeurIPS LLM Merging Competition Leaderboard
-[Website]() | [GitHub]() | [Discord]() |
+[Website](https://llm-merging.github.io/index) | [Starter Kit (Github)]() | [Discord](https://discord.com/invite/dPBHEVnV) |
 
 """
     return leaderboard_md
@@ -223,7 +223,6 @@ def recompute_final_ranking(arena_df):
     
 def get_arena_table(arena_df, model_table_df, arena_subset_df=None):
     arena_df = arena_df.sort_values(by=["final_ranking", "rating"], ascending=[True, False])
-    arena_df = arena_df[arena_df["num_battles"] > 2000]
     arena_df["final_ranking"] = recompute_final_ranking(arena_df)
     arena_df = arena_df.sort_values(by=["final_ranking"], ascending=True)
 
@@ -234,7 +233,6 @@ def get_arena_table(arena_df, model_table_df, arena_subset_df=None):
         arena_subset_df = arena_subset_df[arena_subset_df.index.isin(arena_df.index)]
         arena_subset_df = arena_subset_df.sort_values(by=["rating"], ascending=False)
         # arena_subset_df = arena_subset_df.sort_values(by=["final_ranking"], ascending=True)
-        # arena_subset_df = arena_subset_df[arena_subset_df["num_battles"] > 500]
         arena_subset_df["final_ranking"] = recompute_final_ranking(arena_subset_df)
         # keep only the models in the subset in arena_df and recompute final_ranking
         arena_df = arena_df[arena_df.index.isin(arena_subset_df.index)]
@@ -248,10 +246,6 @@ def get_arena_table(arena_df, model_table_df, arena_subset_df=None):
         arena_df = arena_subset_df.join(arena_df["final_ranking"], rsuffix="_global", how="inner")
         arena_df["ranking_difference"] =  arena_df["final_ranking_global"] - arena_df["final_ranking"]
 
-        # no tie version
-        # arena_df = arena_subset_df.join(arena_df["final_ranking_no_tie"], rsuffix="_global", how="inner")
-        # arena_df["ranking_difference"] =  arena_df["final_ranking_no_tie_global"] - arena_df["final_ranking_no_tie"]
-
         arena_df = arena_df.sort_values(by=["final_ranking", "rating"], ascending=[True, False])
         arena_df["final_ranking"] = arena_df.apply(lambda x: create_ranking_str(x["final_ranking"], x["ranking_difference"]), axis=1)
 
@@ -272,15 +266,6 @@ def get_arena_table(arena_df, model_table_df, arena_subset_df=None):
             row.append(model_name)
             # elo rating
             row.append(round(arena_df.iloc[i]["rating"]))
-            upper_diff = round(
-                arena_df.iloc[i]["rating_q975"] - arena_df.iloc[i]["rating"]
-            )
-            lower_diff = round(
-                arena_df.iloc[i]["rating"] - arena_df.iloc[i]["rating_q025"]
-            )
-            row.append(f"+{upper_diff}/-{lower_diff}")
-            # num battles
-            row.append(round(arena_df.iloc[i]["num_battles"]))
             # Organization
             row.append(
                 model_table_df[model_table_df["key"] == model_key]["Organization"].values[0]
@@ -289,11 +274,6 @@ def get_arena_table(arena_df, model_table_df, arena_subset_df=None):
             row.append(
                 model_table_df[model_table_df["key"] == model_key]["License"].values[0]
             )
-            cutoff_date = model_table_df[model_table_df["key"] == model_key]["Knowledge cutoff date"].values[0]
-            if cutoff_date == "-":
-                row.append("Unknown")
-            else:
-                row.append(cutoff_date)
             values.append(row)
         except Exception as e:
             print(f"{model_key} - {e}")
@@ -301,23 +281,9 @@ def get_arena_table(arena_df, model_table_df, arena_subset_df=None):
 
 key_to_category_name = {
     "full": "Overall",
-    "coding": "Coding",
-    "long_user": "Longer Query",
-    "english": "English",
-    "chinese": "Chinese",
-    "french": "French",
-    "no_tie": "Exclude Ties",
-    "no_short": "Exclude Short",
 }
 cat_name_to_explanation = {
     "Overall": "Overall Questions",
-    "Coding": "Coding: whether conversation contains code snippets",
-    "Longer Query": "Longer Query (>= 500 tokens)",
-    "English": "English Prompts",
-    "Chinese": "Chinese Prompts",
-    "French": "French Prompts",
-    "Exclude Ties": "Exclude Ties and Bothbad",
-    "Exclude Short": "User Query >= 5 tokens",
 }
 
 def build_leaderboard_tab(elo_results_file, leaderboard_table_file, show_plot=False):
@@ -364,7 +330,6 @@ def build_leaderboard_tab(elo_results_file, leaderboard_table_file, show_plot=Fa
                         "⭐ Arena Elo",
                         "Organization",
                         "License",
-                        "Knowledge Cutoff",
                     ],
                     datatype=[
                         "number",
@@ -372,7 +337,6 @@ def build_leaderboard_tab(elo_results_file, leaderboard_table_file, show_plot=Fa
                         "number",
                         "str",
                         "str",
-                        "str",
                     ],
                     value=arena_table_vals,
                     elem_id="arena_leaderboard_dataframe",
@@ -419,7 +383,7 @@ def build_leaderboard_tab(elo_results_file, leaderboard_table_file, show_plot=Fa
         pass
 
     def update_leaderboard_df(arena_table_vals):
-        elo_datarame = pd.DataFrame(arena_table_vals, columns=[ "Rank", "🤖 Model", "⭐ Arena Elo", "Organization", "License", "Knowledge Cutoff"])
+        elo_datarame = pd.DataFrame(arena_table_vals, columns=[ "Rank", "🤖 Model", "⭐ Arena Elo", "Organization", "License"])
 
         # goal: color the rows based on the rank with styler
         def highlight_max(s):
@@ -446,7 +410,6 @@ def build_leaderboard_tab(elo_results_file, leaderboard_table_file, show_plot=Fa
                         "⭐ Arena Elo",
                         "Organization",
                         "License",
-                        "Knowledge Cutoff",
                     ],
                     datatype=[
                         "number",
@@ -454,7 +417,6 @@ def build_leaderboard_tab(elo_results_file, leaderboard_table_file, show_plot=Fa
                         "number",
                         "str",
                         "str",
-                        "str",
                     ],
                     value=arena_values,
                     elem_id="arena_leaderboard_dataframe",
@@ -470,7 +432,6 @@ def build_leaderboard_tab(elo_results_file, leaderboard_table_file, show_plot=Fa
                         "⭐ Arena Elo",
                         "Organization",
                         "License",
-                        "Knowledge Cutoff",
                     ],
                     datatype=[
                         "number",
@@ -478,7 +439,6 @@ def build_leaderboard_tab(elo_results_file, leaderboard_table_file, show_plot=Fa
                         "number",
                         "str",
                         "str",
-                        "str",
                     ],
                     value=arena_values,
                     elem_id="arena_leaderboard_dataframe",