Liangtai commited on
Commit
454f5f0
·
verified ·
1 Parent(s): 370117e

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +1 -0
README.md CHANGED
@@ -15,6 +15,7 @@ SCIEVAL is an evaluation benchmark for large language models in the scientific d
15
  * *scieval-dev.json* is the dev set, containing 5 samples for each task name, each ability and each category, which is specially used for few shot.
16
  * *scieval-valid.json* is the valid set, containing the answer for each question.
17
  * *scieval-test.json* is the test set.
 
18
  * *make_few_shot.py* is the code for generating the few shot data, you can modify it as you need.
19
  * *eval.py* is the evaluation code for the valid set, which is the same as the one we used for the test set. Note the the prediction should follow the format:
20
  ```
 
15
  * *scieval-dev.json* is the dev set, containing 5 samples for each task name, each ability and each category, which is specially used for few shot.
16
  * *scieval-valid.json* is the valid set, containing the answer for each question.
17
  * *scieval-test.json* is the test set.
18
+ * *scieval-test-local.json* is the test set with ground-truth answers, you can use it for local evaluation.
19
  * *make_few_shot.py* is the code for generating the few shot data, you can modify it as you need.
20
  * *eval.py* is the evaluation code for the valid set, which is the same as the one we used for the test set. Note the the prediction should follow the format:
21
  ```