Multiple Choice
Transformers
Safetensors
Korean
qwen3
text-generation
korean
classification
natural-language-inference
function-calling
tool-use
calibrated
Eval Results (legacy)
text-generation-inference
Instructions to use coderecode95/KJH-Judge-4B-ko with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use coderecode95/KJH-Judge-4B-ko with Transformers:
# pip install -U transformers accelerate # Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("coderecode95/KJH-Judge-4B-ko") model = AutoModelForCausalLM.from_pretrained("coderecode95/KJH-Judge-4B-ko", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Download eval/chart.html from coderecode95/KJH-Judge-4B-ko: direct link, hf CLI and curl.
- Browser
- Download file 6.36 kB
-
https://huggingface.co/coderecode95/KJH-Judge-4B-ko/resolve/main/eval/chart.html
- Command line
-
hf download hf://coderecode95/KJH-Judge-4B-ko/eval/chart.html
-
curl -L -o chart.html https://huggingface.co/coderecode95/KJH-Judge-4B-ko/resolve/main/eval/chart.html
6.36 kB
| <html><head><meta charset="utf-8"> | |
| <link rel="stylesheet" href="https://fonts.googleapis.com/css2?family=IBM+Plex+Sans+KR:wght@400;500;600;700&family=IBM+Plex+Mono:wght@500;600&display=swap"> | |
| <style> | |
| *{box-sizing:border-box}body{margin:0;background:#ffffff;font-family:'IBM Plex Sans KR','Malgun Gothic',sans-serif;color:#121823} | |
| .wrap{width:1200px;padding:40px 48px 36px} | |
| h1{margin:0;font-size:30px;font-weight:700;letter-spacing:-.02em}h1 span{color:#1d4ed8} | |
| .sub{margin:8px 0 28px;color:#5a6476;font-size:16px} | |
| .grid{display:grid;grid-template-columns:1fr 270px;gap:36px} | |
| .row{display:grid;grid-template-columns:210px 1fr 96px;gap:16px;align-items:center;height:62px} | |
| .nm b{display:block;font-size:17px}.nm span{font-size:13px;color:#5a6476} | |
| .track{position:relative;height:26px;background:#eef1f5;border-radius:5px} | |
| .bar{position:absolute;inset:0 auto 0 0;background:#1d4ed8;border-radius:5px} | |
| .know .bar{background:#94a3b8} | |
| .ci{position:absolute;top:11px;height:4px;background:rgba(255,255,255,.75);border-radius:2px} | |
| .chance{position:absolute;top:-6px;bottom:-6px;border-left:2px dashed #121823;opacity:.55} | |
| .chance i{position:absolute;top:-17px;left:4px;font-style:normal;font-size:11px;white-space:nowrap;color:#121823} | |
| .val{font:600 26px 'IBM Plex Mono',monospace;text-align:right}.val small{font-size:15px;color:#5a6476} | |
| .sep{margin:6px 0 4px;font-size:12.5px;color:#5a6476;padding-left:226px} | |
| .side{display:flex;flex-direction:column;gap:12px} | |
| .box{border:1px solid #d8dde5;border-radius:10px;padding:14px 16px} | |
| .box b{display:block;font:600 28px 'IBM Plex Mono',monospace;letter-spacing:-.02em}.box span{font-size:13px;color:#5a6476;line-height:1.45;display:block} | |
| .foot{margin-top:22px;font-size:12.5px;color:#5a6476} | |
| .legend{display:flex;flex-wrap:wrap;gap:6px 18px;font-size:13px;color:#5a6476;margin-bottom:10px;padding-left:226px}.legend i{display:inline-block;width:12px;height:12px;border-radius:3px;margin-right:6px;vertical-align:-1px} | |
| .row.two{height:78px}.trk{display:flex;flex-direction:column;gap:5px} | |
| .cmp{position:relative;height:10px;background:#cbd5e1;border-radius:3px;max-width:100%} | |
| .cmp i{position:absolute;left:calc(100% + 6px);top:-4px;font:500 11.5px 'IBM Plex Mono',monospace;color:#5a6476;font-style:normal} | |
| </style></head><body><div class="wrap"> | |
| <h1><span>KJH-Judge-4B-ko</span> ยท ํ์ตํ์ง ์์ ํ๊ตญ์ด ๋ฒค์น๋งํฌ</h1> | |
| <div class="sub">ํ์ต ๋ฐ์ดํฐ์ ๋ค๋ฅธ ์ถ์ฒ์ ๊ณต๊ฐ ๋ฐ์ดํฐ, 0-shot, ๋ณด๊ธฐ ๊ธฐํธ ์ ์ ๋ฐฉ์. ๋ง๋ ์ ํฐ ์ ์ 95% ์ ๋ขฐ๊ตฌ๊ฐ.</div> | |
| <div class="grid"><div> | |
| <div class="legend"><span><i style="background:#1d4ed8"></i>KJH-Judge-4B-ko ยท ํ์ ๊ณผ์ </span><span><i style="background:#94a3b8"></i>KJH-Judge-4B-ko ยท ์ง์ ์ํ</span><span><i style="background:#cbd5e1;border:1px solid #94a3b8"></i>Gemma 4 26B-A4B-it (0-shot)</span></div> | |
| <div class="row two"><div class="nm"><b>NSMC</b><span>์ํ ๋ฆฌ๋ทฐ ๊ฐ์ฑ</span></div> | |
| <div class="trk"><div class="track"><div class="bar" style="width:85.7%"></div><div class="ci" style="left:83.39%;width:4.339999999999989%"></div> | |
| <div class="chance" style="left:50.0%"><i>๋ฌด์์ 50%</i></div></div><div class="cmp" style="width:88.7%"><i>88.7</i></div></div><div class="val">85.7<small>%</small></div></div><div class="row two"><div class="nm"><b>KLUE-STS</b><span>๋ ๋ฌธ์ฅ์ด ๊ฐ์ ๋ป์ธ๊ฐ</span></div> | |
| <div class="trk"><div class="track"><div class="bar" style="width:85.92999999999999%"></div><div class="ci" style="left:82.67999999999999%;width:5.980000000000018%"></div> | |
| <div class="chance" style="left:50.0%"><i>๋ฌด์์ 50%</i></div></div><div class="cmp" style="width:88.44%"><i>88.4</i></div></div><div class="val">85.9<small>%</small></div></div><div class="row two"><div class="nm"><b>Korean UnSmile</b><span>์ ํยทํ์ค ํํ ํ๋ณ</span></div> | |
| <div class="trk"><div class="track"><div class="bar" style="width:82.1%"></div><div class="ci" style="left:79.60000000000001%;width:4.75%"></div> | |
| <div class="chance" style="left:50.0%"><i>๋ฌด์์ 50%</i></div></div><div class="cmp" style="width:81.3%"><i>81.3</i></div></div><div class="val">82.1<small>%</small></div></div><div class="row two"><div class="nm"><b>KorNLI</b><span>๋ฌธ์ฅ ๊ด๊ณ 3๋ถ๋ฅ</span></div> | |
| <div class="trk"><div class="track"><div class="bar" style="width:74.5%"></div><div class="ci" style="left:71.71%;width:5.390000000000015%"></div> | |
| <div class="chance" style="left:33.33%"><i>๋ฌด์์ 33%</i></div></div><div class="cmp" style="width:77.10000000000001%"><i>77.1</i></div></div><div class="val">74.5<small>%</small></div></div><div class="row know two"><div class="nm"><b>HAE-RAE 1.0</b><span>ํ๊ตญ ๋ฌธํยท์ดํ ์ง์</span></div> | |
| <div class="trk"><div class="track"><div class="bar" style="width:56.15%"></div><div class="ci" style="left:53.03%;width:6.200000000000003%"></div> | |
| <div class="chance" style="left:20.84%"><i>๋ฌด์์ 21%</i></div></div><div class="cmp" style="width:79.65%"><i>79.7</i></div></div><div class="val">56.1<small>%</small></div></div><div class="row know two"><div class="nm"><b>KMMLU</b><span>์ ๋ฌธ ์ง์ 4์ง์ ๋ค</span></div> | |
| <div class="trk"><div class="track"><div class="bar" style="width:42.63%"></div><div class="ci" style="left:39.58%;width:6.149999999999999%"></div> | |
| <div class="chance" style="left:25.0%"><i>๋ฌด์์ 25%</i></div></div><div class="cmp" style="width:64.03999999999999%"><i>64.0</i></div></div><div class="val">42.6<small>%</small></div></div></div> | |
| <div class="side"> | |
| <div class="box"><b>4.02B</b><span>ํ๋ผ๋ฏธํฐ ยท Qwen3-4B์์ ๋ฏธ์ธ์กฐ์ </span></div> | |
| <div class="box"><b>250ms</b><span>ํ์ ํ ๋ฒ (์ค์๊ฐ, ์๋ณต ํฌํจ)<br>Gemma 4 26B-A4B 621ms ยท ์๊ฐ ์ผ๊ณ ๋ตํ๋ฉด 6.3์ด</span></div> | |
| <div class="box"><b>0.027</b><span>๊ธฐ๋ ๋ณด์ ์ค์ฐจ(ECE)<br>ํ๋ฅ ์ ๊ทธ๋๋ก ๊ธฐ์ค์ผ๋ก ์ธ ์ ์์</span></div> | |
| <div class="box"><b>97.9%</b><span>๋ณด๊ธฐ๋ฅผ ๊ฑฐ๊พธ๋ก ๋์๋ ๊ฐ์ ๋ต</span></div> | |
| </div></div> | |
| <div class="foot">๋ฌธํญ ์: NSMCยทKorNLIยทUnSmile ๊ฐ 1,000 (๊ณ ์ ์๋ ํ๋ณธ), KLUE-STS 519 (์ ์ฒด), KMMLU 990 (45๊ณผ๋ชฉ ๊ณ ๋ฅด๊ฒ), HAE-RAE 983 (6์์ญ ๊ณ ๋ฅด๊ฒ). ์ฌํ: eval/bench_heldout.py</div> | |
| </div></body></html> |