diff --git a/output/toy.dblp.v12.json/bnn/t31.s11.m13.l[100].lr0.1.b4096.e20.s1/rerank/f0.test.pred.det_cons.10.utileval.csv b/output/toy.dblp.v12.json/bnn/t31.s11.m13.l[100].lr0.1.b4096.e20.s1/rerank/f0.test.pred.det_cons.10.utileval.csv
index df2940c..3b93fd4 100644
--- a/output/toy.dblp.v12.json/bnn/t31.s11.m13.l[100].lr0.1.b4096.e20.s1/rerank/f0.test.pred.det_cons.10.utileval.csv
+++ b/output/toy.dblp.v12.json/bnn/t31.s11.m13.l[100].lr0.1.b4096.e20.s1/rerank/f0.test.pred.det_cons.10.utileval.csv
@@ -1,5 +1,14 @@
 metric,mean.before,mean.after
-map_cut_2,0.21666666666666665,0.15
-map_cut_5,0.30333333333333334,0.2033333333333333
+P_2,0.3,
+P_5,0.24,
+P_10,0.1399999999999999,
+recall_2,0.2166666666666666,
+recall_5,0.4833333333333333,
+recall_10,0.5833333333333333,
+ndcg_cut_2,0.367888315659275,
+ndcg_cut_5,0.4257486373712523,
+ndcg_cut_10,0.4666251168889495,
+map_cut_2,0.2166666666666666,0.15
+map_cut_5,0.3033333333333333,0.2033333333333333
 map_cut_10,0.3176190476190476,0.3005555555555556
 aucroc,0.6316568047337279,0.6494082840236687
diff --git a/output/toy.dblp.v12.json/bnn/t31.s11.m13.l[100].lr0.1.b4096.e20.s1/rerank/f1.test.pred.det_cons.10.utileval.csv b/output/toy.dblp.v12.json/bnn/t31.s11.m13.l[100].lr0.1.b4096.e20.s1/rerank/f1.test.pred.det_cons.10.utileval.csv
index d2511b0..ca80a9a 100644
--- a/output/toy.dblp.v12.json/bnn/t31.s11.m13.l[100].lr0.1.b4096.e20.s1/rerank/f1.test.pred.det_cons.10.utileval.csv
+++ b/output/toy.dblp.v12.json/bnn/t31.s11.m13.l[100].lr0.1.b4096.e20.s1/rerank/f1.test.pred.det_cons.10.utileval.csv
@@ -1,5 +1,14 @@
 metric,mean.before,mean.after
-map_cut_2,0.10833333333333332,0.10833333333333332
-map_cut_5,0.21944444444444441,0.175
+P_2,0.3,
+P_5,0.24,
+P_10,0.2,
+recall_2,0.2166666666666666,
+recall_5,0.4333333333333333,
+recall_10,0.8,
+ndcg_cut_2,0.232111684340725,
+ndcg_cut_5,0.333128018266259,
+ndcg_cut_10,0.4926083647102192,
+map_cut_2,0.1083333333333333,0.10833333333333332
+map_cut_5,0.2194444444444444,0.175
 map_cut_10,0.3319444444444444,0.3256547619047619
 aucroc,0.6442307692307693,0.6479289940828403
diff --git a/output/toy.dblp.v12.json/bnn/t31.s11.m13.l[100].lr0.1.b4096.e20.s1/rerank/f2.test.pred.det_cons.10.utileval.csv b/output/toy.dblp.v12.json/bnn/t31.s11.m13.l[100].lr0.1.b4096.e20.s1/rerank/f2.test.pred.det_cons.10.utileval.csv
index 24fd745..ed0f462 100644
--- a/output/toy.dblp.v12.json/bnn/t31.s11.m13.l[100].lr0.1.b4096.e20.s1/rerank/f2.test.pred.det_cons.10.utileval.csv
+++ b/output/toy.dblp.v12.json/bnn/t31.s11.m13.l[100].lr0.1.b4096.e20.s1/rerank/f2.test.pred.det_cons.10.utileval.csv
@@ -1,5 +1,14 @@
 metric,mean.before,mean.after
+P_2,0.1,
+P_5,0.04,
+P_10,0.16,
+recall_2,0.05,
+recall_5,0.05,
+recall_10,0.6333333333333333,
+ndcg_cut_2,0.0773705614469083,
+ndcg_cut_5,0.049260477748146,
+ndcg_cut_10,0.3069278887774607,
 map_cut_2,0.025,0.31666666666666665
 map_cut_5,0.025,0.31666666666666665
-map_cut_10,0.14384920634920634,0.4896825396825396
-aucroc,0.45192307692307687,0.6923076923076923
+map_cut_10,0.1438492063492063,0.4896825396825396
+aucroc,0.4519230769230768,0.6923076923076923
diff --git a/src/main.py b/src/main.py
index 2556ed8..ad2f325 100644
--- a/src/main.py
+++ b/src/main.py
@@ -113,11 +113,12 @@ def reranked_preds(teamsvecs_members, splits, reranked_idx, reranked_probs, outp
         return sparse_matrix_reranked
 
     @staticmethod
-    def eval_utility(teamsvecs_members, reranked_preds, preds, splits, metrics, output, algorithm, k_max) -> None:
+    def eval_utility(teamsvecs_members, reranked_preds, fpred, preds, splits, metrics, output, algorithm, k_max) -> None:
         """
         Args:
             teamsvecs_members: teamsvecs pickle file
             reranked_preds: re-ranked teams
+            fpred: .pred filename (to see if .pred.eval.mean.csv exists)
             preds: loaded predictions from a .pred file
             splits: indices of test and train samples
             metrics: desired utility metrics
@@ -128,7 +129,11 @@ def eval_utility(teamsvecs_members, reranked_preds, preds, splits, metrics, outp
         """
         # predictions = torch.load(self.predictions_address)
         y_test = teamsvecs_members[splits['test']]
-        _, df_mean_before, _, _ = calculate_metrics(y_test, preds, False, metrics) #although we already have this at test.pred.eval.mean.csv
+        try:
+            df_mean_before = pd.read_csv(f'{fpred}.eval.mean.csv', names=['mean'], header=0)#we should already have it at f*.test.pred.eval.mean.csv
+        except FileNotFoundError:
+            _, df_mean_before, _, _ = calculate_metrics(y_test, preds, False, metrics)
+            df_mean_before.to_csv(f'{fpred}.eval.mean.csv', columns=['mean'])
         df_mean_before.rename(columns={'mean': 'mean.before'}, inplace=True)
         _, df_mean_after, _, _ = calculate_metrics(y_test, reranked_preds.toarray(), False, metrics)
         df_mean_after.rename(columns={'mean': 'mean.after'}, inplace=True)
@@ -190,24 +195,23 @@ def run(fpred, output, fteamsvecs, fsplits, np_ratio, algorithm='det_cons', k_ma
             #not sure os handles file locking for append during parallel run ...
             # with open(f'{new_output}.rerank.time', 'a') as file: file.write(f'{elapsed_time} {new_output} {algorithm} {k_max}\n')
             with open(f'{output}/rerank.time', 'a') as file: file.write(f'{elapsed_time} {new_output} {algorithm} {k_max}\n')
+        try:
+            with open(f'{new_output}.{algorithm}.{k_max}.rerank.pred', 'rb') as f: reranked_preds = pickle.load(f)
+        except FileNotFoundError: reranked_preds = Reranking.reranked_preds(teamsvecs['member'], splits, reranked_idx, probs, new_output, algorithm, k_max)
 
         try:
-            print('Loading fairness evaluation results ...')
+            print('Loading fairness evaluation results before and after reranking ...')
             fairness_eval = pd.read_csv(f'{new_output}.{algorithm}.{k_max}.faireval.csv')
         except FileNotFoundError:
             print(f'Loading fairness results failed! Evaluating fairness metric {fairness_metrics} ...') #for now, it's hardcoded for 'ndkl'
             Reranking.eval_fairness(preds, labels, reranked_idx, ratios, new_output, algorithm, k_max)
 
         try:
-            with open(f'{new_output}.{algorithm}.{k_max}.rerank.pred', 'rb') as f: reranked_preds = pickle.load(f)
-        except FileNotFoundError:
-            reranked_preds = Reranking.reranked_preds(teamsvecs['member'], splits, reranked_idx, probs, new_output, algorithm, k_max)
-        try:
-            print('Loading utility metric evaluation results ...')
+            print('Loading utility metric evaluation results before and after reranking ...')
             utility_before = pd.read_csv(f'{new_output}.{algorithm}.{k_max}.utileval.csv')
         except:
             print(f' Loading utility metric results failed! Evaluating utility metric {utility_metrics} ...')
-            Reranking.eval_utility(teamsvecs['member'], reranked_preds, preds, splits, utility_metrics, new_output, algorithm, k_max=k_max)
+            Reranking.eval_utility(teamsvecs['member'], reranked_preds, fpred, preds, splits, utility_metrics, new_output, algorithm, k_max=k_max)
 
         print(f'Pipeline for the baseline {fpred} completed by {multiprocessing.current_process()}! {time() - st}')
         print('#'*100)
@@ -233,6 +237,7 @@ def addargs(parser):
 
 """
 A running example of arguments
+# single *.pred file
 python -u main.py 
 -fteamsvecs ../data/preprocessed/dblp/toy.dblp.v12.json/teamsvecs.pkl
 -fsplit ../output/toy.dblp.v12.json/splits.json
@@ -240,6 +245,7 @@ def addargs(parser):
 -reranker det_cons
 -output ../output/toy.dblp.v12.json/
 
+# root folder containing many *.pred files.
 python -u main.py 
 -fteamsvecs ../data/preprocessed/dblp/toy.dblp.v12.json/teamsvecs.pkl
 -fsplit ../output/toy.dblp.v12.json/splits.json