xw27 commited on
Commit
7e33843
1 Parent(s): 12bf5e0

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +12 -14
README.md CHANGED
@@ -1,24 +1,22 @@
1
  # SciBench
2
- **SciBench** is a novel benchmark for college-level scientific problems consisting of _695_
3
- problems sourced from instructional textbooks. The benchmark is designed to evaluate the complex reasoning capabilities,
4
  strong domain knowledge, and advanced calculation skills of LLMs.
5
- Please refer to our paper for full description: [SciBench: Evaluating College-Level Scientific Problem-Solving Abilities of Large Language Models](https://arxiv.org/abs/2307.10635)
 
6
 
7
- We developed an innovative **evaluation protocol** for a detailed analysis of reasoning abilities. This
8
- involves instructing LLMs to self-identify and categorize their errors within a predefined set of
9
- capabilities. This process offers a fine-grained understanding of where the models are falling short.
10
-
11
- ## Data
12
- Each file is list of dictionary and can be extracted using following scripts.
13
- Each file stands for one textbook, which is fully elaborated in the paper.
14
 
 
 
15
  ```
16
- subject='atkins'
17
- with open("./data/{}.json".format(subject), encoding='utf-8') as json_file:
18
- problems=json.load(json_file)
19
-
 
 
20
  ```
21
 
 
22
  ---
23
  license: mit
24
  ---
 
1
  # SciBench
2
+ **SciBench** is a novel benchmark for college-level scientific problems sourced from instructional textbooks. The benchmark is designed to evaluate the complex reasoning capabilities,
 
3
  strong domain knowledge, and advanced calculation skills of LLMs.
4
+ Please refer to our [paper](https://arxiv.org/abs/2307.10635) or [website](https://scibench-ucla.github.io) for full description: SciBench: Evaluating College-Level Scientific Problem-Solving Abilities of Large Language Models
5
+ .
6
 
 
 
 
 
 
 
 
7
 
8
+ ## Citation
9
+ If you find our paper useful, please cite our paper
10
  ```
11
+ @inproceedings{wang2024scibench,
12
+ author = {Wang, Xiaoxuan and Hu, Ziniu and Lu, Pan and Zhu, Yanqiao and Zhang, Jieyu and Subramaniam, Satyen and Loomba, Arjun R. and Zhang, Shichang and Sun, Yizhou and Wang, Wei},
13
+ title = {{SciBench: Evaluating College-Level Scientific Problem-Solving Abilities of Large Language Models}},
14
+ booktitle = {Proceedings of the Forty-First International Conference on Machine Learning},
15
+ year = {2024},
16
+ }
17
  ```
18
 
19
+
20
  ---
21
  license: mit
22
  ---