Multi-turn Chat Evaluation on MT-Bench German
73.9WinrateLlama-3.1-8B-TFree-HAT-DPO
Evaluation Results
| Method | Links | |
|---|---|---|
| Llama-3.1-8B-TFree-HAT-DPOComparison Baseline=meta-llama/Llama-3.1-8B-Instruct (German)2026.03 | 73.9 | |
| Llama-3.1-8B-TFree-HAT-DPOComparison Baseline=allenai/Llama-3.1-Tulu-3-8B (German)2026.03 | 71 |