step4: fix small bugs

step4_test_reward_model.py

@@ -100,7 +100,5 @@ if __name__ == "__main__":
     )
 
     groups = group_results(results)
-    eval_results = []
-    for k in range(16):
-        eval_results.append(score_pass_at_k(groups, 1))
+    eval_results = [score_pass_at_k(groups, k, orm_test_model) for k in range(16)]
     save_jsonl(eval_results, cfg["orm"][orm_test_model]["eval_result_path"])

utils_metric.py

@@ -70,7 +70,7 @@ def pass_at_k(groups, k):
 
 
 def score_pass_at_k(groups, k, strategy):
-    result = {"strategy": f"{strategy} * pass@k={k}"}
+    result = {"strategy": strategy, "k": k}
 
     for difficulty, problems in groups.items():
         num_samples, num_correct = [], []