Conversational versatility on Arena-Hard
61.16Win RateTeacher
Evaluation Results
| Method | Links | |
|---|---|---|
| TeacherModel Family=GEMMA-2, Backbone Size=9B2025.02 | 61.16 | |
| PAD w/ LPPDModel Family=GEMMA-2, Backbone Size=2B2025.02 | 60 | |
| PAD w/ LVPDModel Family=GEMMA-2, Backbone Size=2B2025.02 | 59.07 | |
| PROModel Family=GEMMA-2, Backbone Size=2B2025.02 | 58.95 | |
| SimPOModel Family=GEMMA-2, Backbone Size=2B2025.02 | 58.64 | |
| DPOModel Family=GEMMA-2, Backbone Size=2B2025.02 | 57.43 | |
| MiniLLMModel Family=GEMMA-2, Backbone Size=2B2025.02 | 55.75 | |
| SeqKDModel Family=GEMMA-2, Backbone Size=2B2025.02 | 54.87 | |
| TeacherModel Family=LLAMA-3, Backbone Size=8B2025.02 | 52.66 | |
| Standard KDModel Family=GEMMA-2, Backbone Size=2B2025.02 | 52.36 | |
| PAD w/ LPPDModel Family=LLAMA-3, Backbone Size=3B2025.02 | 46.73 | |
| PAD w/ LVPDModel Family=LLAMA-3, Backbone Size=3B2025.02 | 45.23 | |
| PROModel Family=LLAMA-3, Backbone Size=3B2025.02 | 45.09 | |
| SimPOModel Family=LLAMA-3, Backbone Size=3B2025.02 | 44.85 | |
| DPOModel Family=LLAMA-3, Backbone Size=3B2025.02 | 44.71 | |
| SeqKDModel Family=LLAMA-3, Backbone Size=3B2025.02 | 42.52 | |
| MiniLLMModel Family=LLAMA-3, Backbone Size=3B2025.02 | 42.21 | |
| Standard KDModel Family=LLAMA-3, Backbone Size=3B2025.02 | 41.68 | |
| StudentModel Family=GEMMA-2, Backbone Size=2B2025.02 | 37.55 | |
| StudentModel Family=LLAMA-3, Backbone Size=3B2025.02 | 31.7 |