File size: 6,243 Bytes
fee8f50
 
 
 
a38e13d
fee8f50
 
e22d664
fee8f50
 
 
a38e13d
fee8f50
 
e22d664
a38e13d
e22d664
a38e13d
e22d664
 
 
fee8f50
 
e22d664
8b2096c
a38e13d
24d4f12
 
 
a38e13d
 
 
fee8f50
ac7a85c
 
 
 
 
 
 
 
 
9301ba4
 
 
 
 
 
 
 
 
e22d664
 
 
9301ba4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
fee8f50
 
e22d664
fee8f50
e22d664
fee8f50
 
 
9301ba4
fee8f50
9301ba4
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
import gradio as gr
import pandas as pd
from css_html_js import custom_css

TITLE = """<h1 align="center" id="space-title">πŸ‡²πŸ‡Ύ Malay LLM Leaderboard</h1>"""

INTRODUCTION_TEXT = """
πŸ“ The πŸ‡²πŸ‡Ύ Malay LLM Leaderboard aims to track, rank and evaluate open LLMs on Malay tasks. All notebooks at https://github.com/mesolitica/llm-benchmarks, feel free to submit your own score at https://huggingface.co/spaces/mesolitica/malay-llm-leaderboard/discussions with link to the notebook.

## Dataset

πŸ“ˆ We evaluate models based on 3 datasets,

1. BM-PT3 Paper 1, contains 54 questions, https://github.com/mesolitica/malaysian-dataset/tree/master/llm-benchmark/BM-pt3
- This test is for 15 years old Malaysia student, it is about reading comprehension and general knowledge for malay language.
2. Tatabahasa, contains 349 questions, https://github.com/mesolitica/malaysian-dataset/tree/master/llm-benchmark/tatabahasabm.tripod.com
- This test is general test for malay grammar.
3. Translated IndoNLI to Malay, tested on `test_expert` dataset, https://huggingface.co/datasets/mesolitica/translated-indonli
- This test is general test to language reasoning.
4. HumanEval, https://github.com/openai/human-eval
- This test is for programming language understanding.
"""

close_source = [
    {
        'model': 'gpt-4-1106-preview',
        'BM-PT3 0-shot': 51.85185185185185,
        'BM-PT3 1-shot': 66.66666666666666,
        'BM-PT3 3-shots': 55.55555555555556,
        'Tatabahasa 0-shot': 75.64469914040114,
        'Tatabahasa 1-shot': 73.63896848137536,
        'Tatabahasa 3-shots': 75.64469914040114,
    },
    {
        'model': 'gpt-3.5-turbo-0613',
        'BM-PT3 0-shot': 36.53846153846153,
        'BM-PT3 1-shot': 28.846153846153843,
        'BM-PT3 3-shots': 24.528301886792452,
        'Tatabahasa 0-shot': 59.530791788856305,
        'Tatabahasa 1-shot': 60.80691642651297,
        'Tatabahasa 3-shots': 63.03724928366762,
    },
    {
        'model': 'Antrophic Claude 2',
        'Tatabahasa 0-shot': 61,
        'Tatabahasa 3-shots': 57.8,
    },
    {
        'model': 'Antrophic Claude 1',
        'Tatabahasa 3-shots': 67,
    },
]

open_source = [
    {
        'model': '[llama2-7b](https://huggingface.co/meta-llama/Llama-2-7b-hf)',
        'Tatabahasa 0-shot': 24.355300859598856,
        'Tatabahasa 1-shot': 28.08022922636103,
        'Tatabahasa 3-shots': 24.641833810888254,
    },
    {
        'model': '[malaysian-llama2-7b-32k](https://huggingface.co/mesolitica/llama-7b-hf-32768-fpf)',
        'BM-PT3 0-shot': 20.37037037037037,
        'BM-PT3 1-shot': 20.37037037037037,
        'BM-PT3 3-shots': 29.629629629629626,
        'Tatabahasa 0-shot': 17.765042979942695,
        'Tatabahasa 1-shot': 24.068767908309454,
        'Tatabahasa 3-shots': 27.507163323782237,
    },
    {
        'model': '[malaysian-llama2-7b-32k-instructions](https://huggingface.co/mesolitica/malaysian-llama2-7b-32k-instructions)',
        'BM-PT3 0-shot': 35.294117647058826,
        'BM-PT3 1-shot': 21.153846153846153,
        'BM-PT3 3-shots': 28.30188679245283,
    },
    {
        'model': '[malaysian-llama2-13b-32k](https://huggingface.co/mesolitica/llama-13b-hf-32768-fpf)',
        'BM-PT3 0-shot': 33.33333333333333,
        'BM-PT3 1-shot': 20.37037037037037,
        'BM-PT3 3-shots': 31.48148148148148,
        'Tatabahasa 0-shot': 26.07449856733524,
        'Tatabahasa 1-shot': 25.214899713467048,
        'Tatabahasa 3-shots': 24.355300859598856,
    },
    {
        'model': '[malaysian-llama2-13b-32k-instructions](https://huggingface.co/mesolitica/malaysian-llama2-13b-32k-instructions)',
        'BM-PT3 0-shot': 28.57142857142857,
        'BM-PT3 1-shot': 12.244897959183673,
        'BM-PT3 3-shots': 17.307692307692307,
    },
    {
        'model': '[mistral-7b](https://huggingface.co/mistralai/Mistral-7B-v0.1)',
        'Tatabahasa 0-shot': 28.939828080229223,
        'Tatabahasa 1-shot': 34.38395415472779,
        'Tatabahasa 3-shots': 32.95128939828081,
    },
    {
        'model': '[malaysian-mistral-7b-4k](https://huggingface.co/mesolitica/mistral-7b-4096-fpf)',
        'BM-PT3 0-shot': 20.37037037037037,
        'BM-PT3 1-shot': 22.22222222222222,
        'BM-PT3 3-shots': 33.33333333333333,
        'Tatabahasa 0-shot': 21.48997134670487,
        'Tatabahasa 1-shot': 28.939828080229223,
        'Tatabahasa 3-shots': 24.641833810888254,
    },
    {
        'model': '[malaysian-mistral-7b-32k](https://huggingface.co/mesolitica/mistral-7b-32768-fpf)',
        'BM-PT3 0-shot': 16.666666666666664,
        'BM-PT3 1-shot': 16.666666666666664,
        'BM-PT3 3-shots': 25.925925925925924,
        'Tatabahasa 0-shot': 18.624641833810887,
        'Tatabahasa 1-shot': 24.355300859598856,
        'Tatabahasa 3-shots': 28.653295128939828,
    },
    {
        'model': '[malaysian-mistral-7b-32k-instructions](https://huggingface.co/mesolitica/malaysian-mistral-7b-32k-instructions)',
        'BM-PT3 0-shot': 35.18518518518518,
        'BM-PT3 1-shot': 33.33333333333333,
        'BM-PT3 3-shots': 37.03703703703704,
        'Tatabahasa 0-shot': 55.014326647564474,
        'Tatabahasa 1-shot': 42.693409742120345,
        'Tatabahasa 3-shots': 33.33333333333333,
    },
    {
        'model': '[aisingapore/sealion3b](https://huggingface.co/aisingapore/sealion3b)',
        'BM-PT3 0-shot': 20.37037037037037,
        'BM-PT3 1-shot': 25.925925925925924,
        'BM-PT3 3-shots': 31.48148148148148,
        'Tatabahasa 0-shot': 21.776504297994272,
        'Tatabahasa 1-shot': 21.776504297994272,
        'Tatabahasa 3-shots': 24.641833810888254,
    },
    {
        'model': '[aisingapore/sealion7b](https://huggingface.co/aisingapore/sealion7b)',
        'BM-PT3 0-shot': 20.37037037037037,
        'BM-PT3 1-shot': 24.074074074074073,
        'BM-PT3 3-shots': 33.33333333333333,
        'Tatabahasa 0-shot': 25.787965616045845,
        'Tatabahasa 1-shot': 27.507163323782237,
        'Tatabahasa 3-shots': 26.07449856733524,
    }
]

data = pd.DataFrame(close_source + open_source)

demo = gr.Blocks(css=custom_css)
with demo:
    gr.HTML(TITLE)
    gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text")
    gr.DataFrame(data, datatype = 'markdown')

demo.launch()