update

src/backend/envs.py

@@ -24,7 +24,7 @@ class Tasks(Enum):
     task1 = Task("triviaqa", "em", "TriviaQA", 64)  # 64, as in the ATLAS paper
     task2 = Task("truthfulqa_mc1", "acc", "TruthfulQA MC1", 0)
     task3 = Task("truthfulqa_mc2", "acc", "TruthfulQA MC2", 0)  # TruthfulQA is intended as a zero-shot benchmark [5, 47]. https://owainevans.github.io/pdfs/truthfulQA_lin_evans.pdf
-    task4 = Task("halueval_qa", "em", "HaluEval QA", 0)
+    task4 = Task("halueval_qa", "acc", "HaluEval QA", 0)
 
 # NUM_FEWSHOT = 64  # Change with your few shot