SUFE-AIFLM-Lab commited on
Commit
f684d2f
1 Parent(s): 0e52f9c

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +42 -4
README.md CHANGED
@@ -8,7 +8,7 @@ language:
8
  - zh
9
  pretty_name: FinEval
10
  size_categories:
11
- - 10K<n<100K
12
  viewer: false
13
  ---
14
  <p><h1> The FinEval Dataset! </h1></p>
@@ -21,10 +21,48 @@ FinEval is a collection of high-quality multiple-choice questions covering vario
21
 
22
  Each subject consists of three splits: dev, val, and test. The dev set per subject consists of five exemplars with explanations for few-shot evaluation. The val set is intended to be used for hyperparameter tuning. And the test set is for model evaluation. Labels on the test split are not released, users are required to submit their results to automatically obtain test accuracy.
23
 
24
- ### Load the data
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
25
  ```python
26
  from datasets import load_dataset
27
  dataset=load_dataset(r"SUFE-AIFLM-Lab/FinEval",name="finance")
28
 
29
- ```
30
-
 
8
  - zh
9
  pretty_name: FinEval
10
  size_categories:
11
+ - 1K<n<10K
12
  viewer: false
13
  ---
14
  <p><h1> The FinEval Dataset! </h1></p>
 
21
 
22
  Each subject consists of three splits: dev, val, and test. The dev set per subject consists of five exemplars with explanations for few-shot evaluation. The val set is intended to be used for hyperparameter tuning. And the test set is for model evaluation. Labels on the test split are not released, users are required to submit their results to automatically obtain test accuracy.
23
 
24
+ # Language
25
+
26
+ The language of the data is Chinese.
27
+
28
+ # Performance Leaderboard
29
+ We divide the evaluation into Answer Only and Chain of Thought. For examples of prompts for both methods, please refer to zero-shot for Answer Only, few-shot for Answer Only, and Chain of Thought.
30
+
31
+ Below is the average accuracy(%) on the test split. We report the average accuracy over the subjects within each category. "Average" column indicates the average accuracy over all the subjects. Notably, we only report the results from each model under the best setting, which is determined by the highest average accuracy achieved among four settings (i.e., zero- and few-shot learning with and without CoT):
32
+
33
+ | Model | Size | Finance | Economy | Accounting | Certificate | Average |
34
+ |------------------------|---------|:-------:|:-------:|:----------:|:-----------:|:-------:|
35
+ | GPT-4 | unknown | 71.0 | 74.5 | 59.3 | 70.4 | 68.6 |
36
+ | ChatGPT | 175B | 59.3 | 61.6 | 45.2 | 55.1 | 55.0 |
37
+ | Qwen-7B | 7B | 54.5 | 54.4 | 50.3 | 55.8 | 53.8 |
38
+ | Qwen-Chat-7B | 7B | 51.5 | 52.1 | 44.5 | 53.6 | 50.5 |
39
+ | Baichuan-13B-Base | 13B | 52.6 | 50.2 | 43.4 | 53.5 | 50.1 |
40
+ | Baichuan-13B-Chat | 13B | 51.6 | 51.1 | 41.7 | 52.8 | 49.4 |
41
+ | ChatGLM2-6B | 6B | 46.5 | 46.4 | 44.5 | 51.5 | 47.4 |
42
+ | InternLM-7B | 7B | 49.0 | 49.2 | 40.5 | 49.4 | 47.1 |
43
+ | InternLM-Chat-7B | 7B | 48.4 | 49.1 | 40.8 | 49.5 | 47.0 |
44
+ | LLaMA-2-Chat-70B | 70B | 47.1 | 46.7 | 41.5 | 45.7 | 45.2 |
45
+ | Falcon-40B | 40B | 45.4 | 43.2 | 35.8 | 44.8 | 42.4 |
46
+ | Baichuan-7B | 7B | 44.9 | 41.5 | 34.9 | 45.6 | 42.0 |
47
+ | LLaMA-2-Chat-13B | 13B | 41.6 | 38.4 | 34.1 | 42.1 | 39.3 |
48
+ | Ziya-LLaMA-13B-v1 | 13B | 43.3 | 36.9 | 34.3 | 41.2 | 39.3 |
49
+ | Bloomz-7b1-mt | 7B | 41.4 | 42.1 | 32.5 | 39.7 | 38.8 |
50
+ | LLaMA-2-13B | 13B | 39.5 | 38.6 | 31.6 | 39.6 | 37.4 |
51
+ | ChatGLM-6B | 6B | 38.8 | 36.2 | 33.8 | 39.1 | 37.2 |
52
+ | Chinese-Llama-2-7B | 7B | 37.8 | 37.8 | 31.4 | 36.7 | 35.9 |
53
+ | Chinese-Alpaca-Plus-7B | 7B | 30.5 | 33.4 | 32.7 | 38.5 | 34.0 |
54
+ | moss-moon-003-sft | 16B | 35.6 | 34.3 | 28.7 | 35.6 | 33.7 |
55
+ | LLaMA-2-Chat-7B | 7B | 35.6 | 31.8 | 31.9 | 34.0 | 33.5 |
56
+ | LLaMA-2-7B | 7B | 34.9 | 36.4 | 31.4 | 31.6 | 33.4 |
57
+ | AquilaChat-7B | 7B | 34.2 | 31.3 | 29.8 | 36.2 | 33.1 |
58
+ | moss-moon-003-base | 16B | 32.2 | 33.1 | 29.2 | 30.7 | 31.2 |
59
+ | Aquila-7B | 7B | 27.1 | 31.6 | 32.4 | 33.6 | 31.2 |
60
+ | LLaMA-13B | 13B | 33.1 | 29.7 | 27.2 | 33.6 | 31.1 |
61
+ | Falcon-7B | 7B | 28.5 | 28.2 | 27.5 | 27.4 | 27.9 |
62
+
63
+ # Load the data
64
  ```python
65
  from datasets import load_dataset
66
  dataset=load_dataset(r"SUFE-AIFLM-Lab/FinEval",name="finance")
67
 
68
+ ```