| title: KAIdol Thinking Experiment | |
| emoji: π€ | |
| colorFrom: purple | |
| colorTo: pink | |
| sdk: docker | |
| pinned: false | |
| license: apache-2.0 | |
| tags: | |
| - roleplay | |
| - korean | |
| - llm-evaluation | |
| - a-b-testing | |
| # KAIdol A/B Test Arena | |
| K-pop μμ΄λ λ‘€νλ μ΄ μ±λ΄ λͺ¨λΈ A/B λΉκ΅ νκ° νλ«νΌ | |
| ## Features | |
| - **A/B Arena**: λ λͺ¨λΈμ μλ΅μ λλν λΉκ΅ | |
| - **Blind Mode**: λͺ¨λΈλͺ μ¨κΈ°κ³ μμ νμ§ νκ° | |
| - **ELO Ranking**: ν¬ν κΈ°λ° λͺ¨λΈ μμ | |
| - **5 Characters**: κ°μ¨, μμ΄μ, μ΄μ§ν, μ°¨λν, μ΅λ―Ό | |
| ## Models (19κ° μν Student λͺ¨λΈ) | |
| ### DPO v5 (7-14B) | |
| - qwen2.5-7b/14b-dpo-v5 | |
| - exaone-7.8b-dpo-v5 | |
| - qwen3-8b-dpo-v5 | |
| - solar-10.7b-dpo-v5 | |
| ### SFT Thinking (7-14B) | |
| - qwen2.5-7b/14b-thinking | |
| - exaone-7.8b-thinking | |
| ### Phase 7 Kimi Students | |
| - qwen2.5-7b/14b-kimi | |
| - exaone-7.8b-kimi | |
| ### V7 Students | |
| - qwen2.5-7b/14b-v7 | |
| - exaone-7.8b-v7 | |
| - qwen3-8b-v7 | |
| - varco-8b-v7 | |
| ## Usage | |
| 1. μΊλ¦ν°μ μλλ¦¬μ€ μ ν | |
| 2. λ©μμ§ μ λ ₯ λλ λλ€ μλλ¦¬μ€ μ¬μ© | |
| 3. λ λͺ¨λΈμ μλ΅ λΉκ΅ | |
| 4. ν¬νλ‘ λ λμ μλ΅ μ ν | |
| ## Tech Stack | |
| - Gradio 4.x | |
| - Python 3.11 | |