README.md

---
pipeline_tag: text-classification
tags: [encoder, binary-classification, routing, ai-champs]
base_model: intfloat/multilingual-e5-base
---

# ai-champ: Question → A/B 라우팅 분류기

- 입력: `question` + `[MODEL] {model_name}`
- 라벨: score ≥ 4 → **B(1)**, else **A(0)**
- 학습 데이터: `HAERAE-HUB/ai-champs-train`의 train/test 각각 10% 샘플
- 테스트 메트릭: {'eval_loss': 0.5179986953735352, 'eval_accuracy': 0.7566917950398776, 'eval_f1': 0.7842681391068488, 'eval_roc_auc': 0.8158790589102392, 'eval_runtime': 19.1572, 'eval_samples_per_second': 955.569, 'eval_steps_per_second': 7.517, 'epoch': 3.0}