Updated list of bluebench tasks (with contributions by [email protected])
Browse files- src/about.py +16 -2
src/about.py
CHANGED
@@ -12,8 +12,22 @@ class Task:
|
|
12 |
# ---------------------------------------------------
|
13 |
class Tasks(Enum):
|
14 |
# task_key in the json file, metric_key in the json file, name to display in the leaderboard
|
15 |
-
task0 = Task("anli_r1", "acc", "ANLI")
|
16 |
-
task1 = Task("logiqa", "acc_norm", "LogiQA")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
17 |
|
18 |
NUM_FEWSHOT = 0 # Change with your few shot
|
19 |
# ---------------------------------------------------
|
|
|
12 |
# ---------------------------------------------------
|
13 |
class Tasks(Enum):
|
14 |
# task_key in the json file, metric_key in the json file, name to display in the leaderboard
|
15 |
+
# task0 = Task("anli_r1", "acc", "ANLI")
|
16 |
+
# task1 = Task("logiqa", "acc_norm", "LogiQA")
|
17 |
+
|
18 |
+
task0 = Task("bias", "score", "Bias")
|
19 |
+
task1 = Task("chatbot_abilities", "score", "Chatbot Abilities")
|
20 |
+
task2 = Task("entity_extraction", "score", "Entity Extraction")
|
21 |
+
task3 = Task("knowledge", "score", "Knowledge")
|
22 |
+
task4 = Task("legal", "score", "Legal")
|
23 |
+
task5 = Task("news_classification", "score", "News Classification")
|
24 |
+
task6 = Task("product_help", "score", "Product Help")
|
25 |
+
task7 = Task("qa_finance", "score", "QA Fianace")
|
26 |
+
task8 = Task("rag_general", "score", "RAG General")
|
27 |
+
task9 = Task("reasoning", "score", "Reasoning")
|
28 |
+
task10 = Task("safety", "score", "Safety")
|
29 |
+
task11 = Task("summarization", "score", "Summarization")
|
30 |
+
task12 = Task("translation", "score", "Translation")
|
31 |
|
32 |
NUM_FEWSHOT = 0 # Change with your few shot
|
33 |
# ---------------------------------------------------
|