ikala-ray commited on
Commit
40d66eb
1 Parent(s): 761cc30

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +34 -0
README.md CHANGED
@@ -27,6 +27,37 @@ We present TMMLU+ a traditional Chinese massive multitask language understanding
27
 
28
  TMMLU+ dataset is 6 times larger and contains more balanced subjects compared to the previous version, [TMMLU](https://github.com/mtkresearch/MR-Models/tree/main/TC-Eval/data/TMMLU). We included benchmark results in TMMLU+ from closed-source models and 20 open-weight Chinese large language models of parameters ranging from 1.8B to 72B. Benchmark results show Traditional Chinese variants still lag behind those trained on Simplified Chinese major models.
29
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
30
  ## Benchmark on direct prompting
31
 
32
  | model | STEM | Social Science | Humanities | Other | Average |
@@ -59,3 +90,6 @@ TMMLU+ dataset is 6 times larger and contains more balanced subjects compared to
59
  | [yentinglin/Taiwan-LLM-7B-v2.1-chat](https://huggingface.co/yentinglin/Taiwan-LLM-7B-v2.1-chat) | 14.99 | 16.23 | 15.00 | 16.22 |15.61|
60
  | [FlagAlpha/Atom-7B](https://huggingface.co/FlagAlpha/Atom-7B) | 5.60 | 13.57 | 7.71 | 11.84 |9.68|
61
 
 
 
 
 
27
 
28
  TMMLU+ dataset is 6 times larger and contains more balanced subjects compared to the previous version, [TMMLU](https://github.com/mtkresearch/MR-Models/tree/main/TC-Eval/data/TMMLU). We included benchmark results in TMMLU+ from closed-source models and 20 open-weight Chinese large language models of parameters ranging from 1.8B to 72B. Benchmark results show Traditional Chinese variants still lag behind those trained on Simplified Chinese major models.
29
 
30
+
31
+ ```python
32
+ from datasets import load_dataset
33
+ task_list = [
34
+ 'engineering_math', 'dentistry', 'traditional_chinese_medicine_clinical_medicine', 'clinical_psychology', 'technical', 'culinary_skills', 'mechanical', 'logic_reasoning', 'real_estate',
35
+ 'general_principles_of_law', 'finance_banking', 'anti_money_laundering', 'ttqav2', 'marketing_management', 'business_management', 'organic_chemistry', 'advance_chemistry',
36
+ 'physics', 'secondary_physics', 'human_behavior', 'national_protection', 'jce_humanities', 'politic_science', 'agriculture', 'official_document_management',
37
+ 'financial_analysis', 'pharmacy', 'educational_psychology', 'statistics_and_machine_learning', 'management_accounting', 'introduction_to_law', 'computer_science', 'veterinary_pathology',
38
+ 'accounting', 'fire_science', 'optometry', 'insurance_studies', 'pharmacology', 'taxation', 'trust_practice', 'geography_of_taiwan', 'physical_education', 'auditing', 'administrative_law',
39
+ 'education_(profession_level)', 'economics', 'veterinary_pharmacology', 'nautical_science', 'occupational_therapy_for_psychological_disorders',
40
+ 'basic_medical_science', 'macroeconomics', 'trade', 'chinese_language_and_literature', 'tve_design', 'junior_science_exam', 'junior_math_exam', 'junior_chinese_exam',
41
+ 'junior_social_studies', 'tve_mathematics', 'tve_chinese_language', 'tve_natural_sciences', 'junior_chemistry', 'music', 'education', 'three_principles_of_people',
42
+ 'taiwanese_hokkien'
43
+ ]
44
+ for task in task_list:
45
+ val = load_dataset('ikala/tmmluplus', task)['validation']
46
+ dev = load_dataset('ikala/tmmluplus', task)['train']
47
+ test = load_dataset('ikala/tmmluplus', task)['test']
48
+ ```
49
+
50
+ Statistic on datasets result
51
+
52
+ | Category | Test | Dev | Validation |
53
+ |----------------------------------|-------|------|------------|
54
+ | STEM | 3458 | 70 | 385 |
55
+ | Other (Business, Health, Misc.) | 8939 | 135 | 995 |
56
+ | Social Sciences | 5958 | 90 | 665 |
57
+ | Humanities | 1763 | 35 | 197 |
58
+ | **Total** | 20118 | 330 | 2242 |
59
+
60
+
61
  ## Benchmark on direct prompting
62
 
63
  | model | STEM | Social Science | Humanities | Other | Average |
 
90
  | [yentinglin/Taiwan-LLM-7B-v2.1-chat](https://huggingface.co/yentinglin/Taiwan-LLM-7B-v2.1-chat) | 14.99 | 16.23 | 15.00 | 16.22 |15.61|
91
  | [FlagAlpha/Atom-7B](https://huggingface.co/FlagAlpha/Atom-7B) | 5.60 | 13.57 | 7.71 | 11.84 |9.68|
92
 
93
+
94
+
95
+