Merge branch 'main' of https://huggingface.co/spaces/allenai/reward-bench-v2

app.py

@@ -41,8 +41,8 @@ repo = snapshot_download(
 ###########################################
 
 def avg_over_rewardbench_v2(dataframe_core):
-    domain_cols = ["chat", "factuality", "safety", "math", "precise if", "ties"]
-    domain_weights = [0, 1, 1, 1, 1, 1]
+    domain_cols = ["factuality", "precise if", "math", "safety", "chat", "ties"]
+    domain_weights = [1, 1, 1, 1, 1, 1]
     new_df = dataframe_core.copy()
 
     # for main subsets, keys in subset_mapping, take the weighted avg by example_counts and store for the models
@@ -58,6 +58,8 @@ def avg_over_rewardbench_v2(dataframe_core):
     keep_columns = ["model", "model_type", "average"] + domain_cols
     new_df = new_df[keep_columns]
 
+    # TODO: update domain_cols and comment this out if final dataset version changes names
+    new_df = new_df.rename(columns={"factuality": "Factuality", "precise if": "Precise IF", "math": "Math", "safety": "Safety", "chat": "Focus", "ties": "Ties"})
     return new_df
 
 def avg_over_rewardbench(dataframe_core, dataframe_prefs):