coderecode95 commited on
Commit
af72c54
ยท
verified ยท
1 Parent(s): cca00cc

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +95 -51
README.md CHANGED
@@ -1,53 +1,97 @@
1
- ---
2
- license: apache-2.0
3
- license_link: https://huggingface.co/Qwen/Qwen3-4B/blob/main/LICENSE
4
- base_model: Qwen/Qwen3-4B
5
- base_model_relation: finetune
6
- language:
7
- - ko
8
- library_name: transformers
9
- tags:
10
- - korean
11
- - classification
12
- - natural-language-inference
13
- - function-calling
14
- - tool-use
15
- - calibrated
16
- - multiple-choice
17
- datasets:
18
- - klue/klue
19
- - skt/kobest_v1
20
- - kocohub/korean-hate-speech
21
- - jason9693/APEACH
22
- - Davlan/sib200
23
- model-index:
24
- - name: KJH-Judge-4B-ko
25
- results:
26
- - task: {type: text-classification, name: Sentiment classification}
27
- dataset: {name: NSMC, type: e9t/nsmc, split: test}
28
- metrics:
29
- - {type: accuracy, value: 85.7, name: accuracy (0-shot, n=1000)}
30
- - task: {type: text-classification, name: Natural language inference}
31
- dataset: {name: KorNLI (XNLI), type: kakaobrain/kor_nli, config: xnli, split: test}
32
- metrics:
33
- - {type: accuracy, value: 74.5, name: accuracy (0-shot, n=1000)}
34
- - task: {type: text-classification, name: Paraphrase detection (binary STS)}
35
- dataset: {name: KLUE-STS, type: klue/klue, config: sts, split: validation}
36
- metrics:
37
- - {type: accuracy, value: 85.9, name: accuracy (0-shot, n=519)}
38
- - task: {type: text-classification, name: Hate speech detection}
39
- dataset: {name: Korean UnSmile, type: smilegate-ai/kor_unsmile, split: valid}
40
- metrics:
41
- - {type: accuracy, value: 82.1, name: accuracy (0-shot, n=1000)}
42
- - task: {type: multiple-choice, name: Knowledge QA}
43
- dataset: {name: KMMLU, type: HAERAE-HUB/KMMLU, split: test}
44
- metrics:
45
- - {type: accuracy, value: 42.6, name: accuracy (0-shot, n=990)}
46
- - task: {type: multiple-choice, name: Korean knowledge QA}
47
- dataset: {name: HAE-RAE Bench 1.0, type: HAERAE-HUB/HAE_RAE_BENCH_1.0, split: test}
48
- metrics:
49
- - {type: accuracy, value: 56.1, name: accuracy (0-shot, n=983)}
50
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
51
 
52
  # KJH-Judge-4B-ko
53
 
@@ -495,4 +539,4 @@ B. {๋ณด๊ธฐ 2}
495
 
496
  - ๋ชจ๋ธ ๊ฐ€์ค‘์น˜๋Š” ์›๋ณธ Qwen3-4B์™€ ๊ฐ™์€ **Apache 2.0**์ž…๋‹ˆ๋‹ค ([LICENSE](LICENSE)). ๊ณต๊ฐœ ํ•™์Šต ๋ฐ์ดํ„ฐ๋Š” ๋ชจ๋‘ CC BY-SA 4.0์ž…๋‹ˆ๋‹ค.
497
  - ํ˜์˜ค ํ‘œํ˜„ ํŒ์ •, ์ •์ฑ… ํŒ๋‹จ ๋“ฑ ์‚ฌ๋žŒ์—๊ฒŒ ์˜ํ–ฅ์„ ์ฃผ๋Š” ๊ฒฐ์ •์— ์“ธ ๋•Œ๋Š” ํ™•์‹ ์ด ๋‚ฎ์€ ๊ฒฐ๊ณผ๋ฅผ ์‚ฌ๋žŒ์ด ๊ฒ€ํ† ํ•˜๋„๋ก ํ•˜์„ธ์š”.
498
- - ๋ฒ•๋ฅ ยท๊ธˆ์œต ํŒ๋‹จ์˜ ๊ทผ๊ฑฐ๋กœ ์“ฐ์ง€ ๋งˆ์„ธ์š”. ์ฃผ์–ด์ง„ ๊ธ€ ์•ˆ์—์„œ ๋‚ด์šฉ์„ ์ฐพ๊ณ  ๋Œ€์กฐํ•˜๋Š” ๋„๊ตฌ์ž…๋‹ˆ๋‹ค.
 
1
+ ---
2
+ license: apache-2.0
3
+ license_link: https://huggingface.co/Qwen/Qwen3-4B/blob/main/LICENSE
4
+ base_model: Qwen/Qwen3-4B
5
+ base_model_relation: finetune
6
+ language:
7
+ - ko
8
+ library_name: transformers
9
+ tags:
10
+ - korean
11
+ - classification
12
+ - natural-language-inference
13
+ - function-calling
14
+ - tool-use
15
+ - calibrated
16
+ - multiple-choice
17
+ model-index:
18
+ - name: KJH-Judge-4B-ko
19
+ results:
20
+ - task:
21
+ type: text-classification
22
+ name: Sentiment classification
23
+ dataset:
24
+ name: NSMC
25
+ type: e9t/nsmc
26
+ split: test
27
+ metrics:
28
+ - type: accuracy
29
+ value: 85.7
30
+ name: accuracy (0-shot
31
+ n=1000): null
32
+ - task:
33
+ type: text-classification
34
+ name: Natural language inference
35
+ dataset:
36
+ name: KorNLI (XNLI)
37
+ type: kakaobrain/kor_nli
38
+ config: xnli
39
+ split: test
40
+ metrics:
41
+ - type: accuracy
42
+ value: 74.5
43
+ name: accuracy (0-shot
44
+ n=1000): null
45
+ - task:
46
+ type: text-classification
47
+ name: Paraphrase detection (binary STS)
48
+ dataset:
49
+ name: KLUE-STS
50
+ type: klue/klue
51
+ config: sts
52
+ split: validation
53
+ metrics:
54
+ - type: accuracy
55
+ value: 85.9
56
+ name: accuracy (0-shot
57
+ n=519): null
58
+ - task:
59
+ type: text-classification
60
+ name: Hate speech detection
61
+ dataset:
62
+ name: Korean UnSmile
63
+ type: smilegate-ai/kor_unsmile
64
+ split: valid
65
+ metrics:
66
+ - type: accuracy
67
+ value: 82.1
68
+ name: accuracy (0-shot
69
+ n=1000): null
70
+ - task:
71
+ type: multiple-choice
72
+ name: Knowledge QA
73
+ dataset:
74
+ name: KMMLU
75
+ type: HAERAE-HUB/KMMLU
76
+ split: test
77
+ metrics:
78
+ - type: accuracy
79
+ value: 42.6
80
+ name: accuracy (0-shot
81
+ n=990): null
82
+ - task:
83
+ type: multiple-choice
84
+ name: Korean knowledge QA
85
+ dataset:
86
+ name: HAE-RAE Bench 1.0
87
+ type: HAERAE-HUB/HAE_RAE_BENCH_1.0
88
+ split: test
89
+ metrics:
90
+ - type: accuracy
91
+ value: 56.1
92
+ name: accuracy (0-shot
93
+ n=983): null
94
+ ---
95
 
96
  # KJH-Judge-4B-ko
97
 
 
539
 
540
  - ๋ชจ๋ธ ๊ฐ€์ค‘์น˜๋Š” ์›๋ณธ Qwen3-4B์™€ ๊ฐ™์€ **Apache 2.0**์ž…๋‹ˆ๋‹ค ([LICENSE](LICENSE)). ๊ณต๊ฐœ ํ•™์Šต ๋ฐ์ดํ„ฐ๋Š” ๋ชจ๋‘ CC BY-SA 4.0์ž…๋‹ˆ๋‹ค.
541
  - ํ˜์˜ค ํ‘œํ˜„ ํŒ์ •, ์ •์ฑ… ํŒ๋‹จ ๋“ฑ ์‚ฌ๋žŒ์—๊ฒŒ ์˜ํ–ฅ์„ ์ฃผ๋Š” ๊ฒฐ์ •์— ์“ธ ๋•Œ๋Š” ํ™•์‹ ์ด ๋‚ฎ์€ ๊ฒฐ๊ณผ๋ฅผ ์‚ฌ๋žŒ์ด ๊ฒ€ํ† ํ•˜๋„๋ก ํ•˜์„ธ์š”.
542
+ - ๋ฒ•๋ฅ ยท๊ธˆ์œต ํŒ๋‹จ์˜ ๊ทผ๊ฑฐ๋กœ ์“ฐ์ง€ ๋งˆ์„ธ์š”. ์ฃผ์–ด์ง„ ๊ธ€ ์•ˆ์—์„œ ๋‚ด์šฉ์„ ์ฐพ๊ณ  ๋Œ€์กฐํ•˜๋Š” ๋„๊ตฌ์ž…๋‹ˆ๋‹ค.