Updated theme and gradient

app.py

@@ -1,7 +1,9 @@
 import gradio as gr
 import pandas as pd
+import numpy as np
 from gradio_leaderboard import Leaderboard, SelectColumns, ColumnFilter
 
+# Define constants and enums
 TITLE = "<h1>M-RewardBench Leaderboard</h1>"
 INTRODUCTION_TEXT = "https://m-rewardbench.github.io/"
 GOOGLE_SHEET_URL = "https://docs.google.com/spreadsheets/d/1qrD7plUdrBwAw7G6UeDVZAaV9ihxaNAcoiKwSaqotR4/export?gid=0&format=csv"
@@ -17,67 +19,22 @@ class AutoEvalColumn:
   }
   
   model_type = {
-    "name": "Model_Type",
+    "name": "MT",
     "type": "markdown",
     "displayed_by_default": True,
     "never_hidden": True,
-  }
-  
-  eng_Latn = {
-    "name": "eng_Latn",
-    "type": "float",
-    "displayed_by_default": True,
-    "never_hidden": False,
-  }
-  
-  Avg_Multilingual = {
-    "name": "Avg_Multilingual",
-    "type": "float",
-    "displayed_by_default": True,
-    "never_hidden": False,
-  }
+  }	
   
-  arb_Arab = {
-    "name": "arb_Arab",
-    "type": "float",
-    "displayed_by_default": True,
-    "never_hidden": False,
-  }
-  
-  tur_Latn = {
-    "name": "tur_Latn",
-    "type": "float",
-    "displayed_by_default": True,
-    "never_hidden": False,
-  }
-  
-  rus_Cyrl = {
-    "name": "rus_Cyrl",
-    "type": "float",
-    "displayed_by_default": True,
-    "never_hidden": False,
-  }
-  
-  ces_Latn = {
-    "name": "ces_Latn",
-    "type": "float",
-    "displayed_by_default": True,
-    "never_hidden": False,
-  }
-  
-  pol_Latn = {
-    "name": "pol_Latn",
-    "type": "float",
-    "displayed_by_default": True,
-    "never_hidden": False,
-  }
-  
-  kor_Hang = {
-    "name": "kor_Hang",
-    "type": "float",
-    "displayed_by_default": True,
-    "never_hidden": False,
-  }					
+  @classmethod
+  def add_columns_from_df(cls, df, columns):
+    for col in columns:
+      if col.lower() != 'model':  # Skip if it's the model column since it's predefined
+        setattr(cls, col, {
+              "name": col,
+              "type": "markdown",
+              "displayed_by_default": True,
+              "never_hidden": False,
+        })				
 
 
 def get_result_data():
@@ -121,8 +78,39 @@ def format_model_link(row):
   #   return f'<a href="{url}" target="_blank">{model_name}</a>'
   return model_name
 
+lang_ids = "eng_Latn	arb_Arab	tur_Latn	rus_Cyrl	ces_Latn	pol_Latn	kor_Hang	zho_Hans	zho_Hant	fra_Latn	ell_Grek	deu_Latn	ron_Latn	ita_Latn	nld_Latn	pes_Arab	hin_Deva	ukr_Cyrl	por_Latn	ind_Latn	jpn_Jpan	spa_Latn	heb_Hebr	vie_Latn"
+
+emojis = "🔢 💬 🎯"
 
-demo = gr.Blocks()
+model_types = {"Generative RM": "💬", "DPO": "🎯", "Sequence Classifier": "🔢"}
+
+from functools import partial
+def format_with_color(val, min_val=50, max_val=100):
+    """
+    Formats a value with inline green color gradient CSS.
+    Returns an HTML string with bold, black text and muted green background.
+    """
+    try:
+        val = float(val)
+        if pd.isna(val):
+            return str(val)
+            
+        # Normalize value between 50 and 100 to 0-1 range
+        normalized = (val - min_val) / (max_val - min_val)
+        # Clamp value between 0 and 1
+        normalized = max(0, min(1, normalized))
+        
+        # Create color gradient with reduced brightness (max 200 instead of 255)
+        # and increased minimum intensity (50 instead of 0)
+        intensity = int(50 + (150 * (1 - normalized)))        
+        
+        # Return HTML with inline CSS - bold black text
+        return f'<div val={val} style="background-color: rgb({intensity}, 200, {intensity}); color: black; font-weight: bold; text-align: center; vertical-align: middle;">{val*100:.1f}</div>'
+      
+    except (ValueError, TypeError):
+        return str(val)
+
+demo = gr.Blocks(theme=gr.themes.Soft())
 with demo:
   gr.HTML(TITLE)
   gr.Markdown(INTRODUCTION_TEXT)
@@ -130,7 +118,41 @@ with demo:
   with gr.Tabs() as tabs:
     with gr.TabItem("🏅 Leaderboard"):
       df = get_result_data()
+      df["Model_Type"] = df["Model_Type"].map(model_types)
+    
       df["Model"] = df.apply(format_model_link, axis=1)
+      
+      df["zho"] = df[["zho_Hans", "zho_Hant"]].mean(axis=1)
+      
+      columns = lang_ids.split("\t")
+      # print(df.head())
+      df.pop("zho_Hans")
+      df.pop("zho_Hant")
+      
+      df.rename(columns={
+        "Model_Type": "MT",
+        "Avg_Multilingual": "AVG",
+      }, inplace=True)
+      df.rename(columns={col: col[:3] for col in columns}, inplace=True)
+  
+      # df = df.style.applymap(apply_color_gradient, subset=['eng'])
+      numeric_cols = df.select_dtypes(include=[np.number]).columns
+      
+      
+      for col in numeric_cols:
+        lang_format_with_color = partial(format_with_color, 
+                                        min_val=df[col].min(),
+                                        max_val=df[col].max())
+    
+        df[col] = df[col].apply(lang_format_with_color)
+        
+      
+      # for col in numeric_cols:
+      #   df[col] = (df[col] * 100).round(1).astype(str)
+      
+      AutoEvalColumn.add_columns_from_df(df, numeric_cols)
+        
       leaderboard = init_leaderboard(df)
+      
 
 demo.launch(ssr_mode=False)