Multi-turn dialogue on MT-bench
5.25Kendall's TauCORAL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CORALmodel=Llama2-13B-Chat, temperature=0, LM head router=false2025.02 | 5.25 | 3.45 | |
| HASSmodel=Llama2-13B-Chat, temperature=0, LM head router=false2025.02 | 5.2 | 3.42 | |
| CORALmodel=Llama2-13B-Chat, temperature=1, LM head router=false2025.02 | 5.1 | 3.32 | |
| CORALmodel=Llama2-7B-Chat, temperature=0, LM head router=false2025.02 | 5.09 | 3.13 | |
| HASSmodel=Llama2-13B-Chat, temperature=1, LM head router=false2025.02 | 5.02 | 3.26 | |
| HASSmodel=Llama2-7B-Chat, temperature=0, LM head router=false2025.02 | 5.02 | 3.09 | |
| CORALmodel=Llama2-7B-Chat, temperature=1, LM head router=false2025.02 | 4.86 | 2.94 | |
| EAGLE-2model=Llama2-13B-Chat, temperature=0, LM head router=false2025.02 | 4.8 | 3.16 | |
| HASSmodel=Llama2-7B-Chat, temperature=1, LM head router=false2025.02 | 4.77 | 2.88 | |
| EAGLE-2model=Llama2-13B-Chat, temperature=1, LM head router=false2025.02 | 4.68 | 3.06 | |
| EAGLE-2model=Llama2-7B-Chat, temperature=0, LM head router=false2025.02 | 4.68 | 2.89 | |
| CORALmodel=Llama3-8B-Instruct, temperature=0, LM head router=false2025.02 | 4.57 | 2.5 | |
| CORALmodel=Qwen2.5-14B-Instruct, temperature=0, LM head router=false2025.02 | 4.56 | 2.62 | |
| HASSmodel=Qwen2.5-14B-Instruct, temperature=0, LM head router=false2025.02 | 4.52 | 2.59 | |
| HASSmodel=Llama3-8B-Instruct, temperature=0, LM head router=false2025.02 | 4.48 | 2.45 | |
| EAGLE-2model=Llama2-7B-Chat, temperature=1, LM head router=false2025.02 | 4.45 | 2.7 | |
| CORALmodel=Llama3-8B-Instruct, temperature=0, LM head router=true2025.02 | 4.26 | 2.63 | |
| CORALmodel=Qwen2.5-14B-Instruct, temperature=0, LM head router=true2025.02 | 4.26 | 2.74 | |
| CORALmodel=Llama3-70B-Instruct, temperature=0, LM head router=false2025.02 | 4.23 | 2.79 | |
| CORALmodel=Qwen2.5-7B-Instruct, temperature=0, LM head router=false2025.02 | 4.22 | 2.3 | |
| EAGLE-2model=Llama3-8B-Instruct, temperature=0, LM head router=false2025.02 | 4.16 | 2.28 | |
| CORALmodel=Llama3-8B-Instruct, temperature=1, LM head router=false2025.02 | 4.15 | 2.24 | |
| HASSmodel=Qwen2.5-7B-Instruct, temperature=0, LM head router=false2025.02 | 4.15 | 2.26 | |
| CORALmodel=Llama3-70B-Instruct, temperature=1, LM head router=false2025.02 | 4.13 | 2.72 | |
| CORALmodel=Qwen2.5-14B-Instruct, temperature=1, LM head router=false2025.02 | 4.13 | 2.35 | |
| HASSmodel=Llama3-8B-Instruct, temperature=1, LM head router=false2025.02 | 4.12 | 2.21 | |
| HASSmodel=Qwen2.5-14B-Instruct, temperature=1, LM head router=false2025.02 | 4.12 | 2.35 | |
| HASSmodel=Llama3-70B-Instruct, temperature=0, LM head router=false2025.02 | 4.1 | 2.71 | |
| EAGLE-2model=Llama3-70B-Instruct, temperature=0, LM head router=false2025.02 | 4.08 | 2.7 | |
| EAGLE-2model=Qwen2.5-14B-Instruct, temperature=0, LM head router=false2025.02 | 4.08 | 2.36 | |
| CORALmodel=Qwen2.5-7B-Instruct, temperature=0, LM head router=true2025.02 | 4.02 | 2.5 | |
| HASSmodel=Llama3-70B-Instruct, temperature=1, LM head router=false2025.02 | 4 | 2.65 | |
| EAGLEmodel=Llama2-13B-Chat, temperature=0, LM head router=false2025.02 | 3.93 | 3.04 | |
| CORALmodel=Llama3-8B-Instruct, temperature=1, LM head router=true2025.02 | 3.92 | 2.39 | |
| EAGLE-2model=Llama3-70B-Instruct, temperature=1, LM head router=false2025.02 | 3.91 | 2.61 | |
| EAGLE-2model=Qwen2.5-7B-Instruct, temperature=0, LM head router=false2025.02 | 3.91 | 2.13 | |
| CORALmodel=Qwen2.5-14B-Instruct, temperature=1, LM head router=true2025.02 | 3.88 | 2.46 | |
| EAGLE-2model=Llama3-8B-Instruct, temperature=1, LM head router=false2025.02 | 3.84 | 2.08 | |
| CORALmodel=Qwen2.5-7B-Instruct, temperature=1, LM head router=false2025.02 | 3.83 | 2.05 | |
| EAGLEmodel=Llama2-7B-Chat, temperature=0, LM head router=false2025.02 | 3.8 | 2.67 | |
| EAGLE-2model=Qwen2.5-14B-Instruct, temperature=1, LM head router=false2025.02 | 3.76 | 2.15 | |
| HASSmodel=Qwen2.5-7B-Instruct, temperature=1, LM head router=false2025.02 | 3.65 | 1.96 | |
| EAGLEmodel=Llama2-13B-Chat, temperature=1, LM head router=false2025.02 | 3.64 | 2.65 | |
| EAGLEmodel=Llama2-7B-Chat, temperature=1, LM head router=false2025.02 | 3.62 | 2.37 | |
| CORALmodel=Qwen2.5-7B-Instruct, temperature=1, LM head router=true2025.02 | 3.62 | 2.21 | |
| EAGLE-2model=Qwen2.5-7B-Instruct, temperature=1, LM head router=false2025.02 | 3.45 | 1.86 | |
| EAGLEmodel=Llama3-70B-Instruct, temperature=0, LM head router=false2025.02 | 2.87 | 2.24 | |
| EAGLEmodel=Llama3-70B-Instruct, temperature=1, LM head router=false2025.02 | 2.67 | 2.06 | |
| EAGLEmodel=Llama3-8B-Instruct, temperature=0, LM head router=false2025.02 | 2.63 | 1.65 | |
| EAGLEmodel=Qwen2.5-14B-Instruct, temperature=0, LM head router=false2025.02 | 2.63 | 1.83 | |
| EAGLEmodel=Qwen2.5-7B-Instruct, temperature=0, LM head router=false2025.02 | 2.53 | 1.56 | |
| EAGLEmodel=Qwen2.5-14B-Instruct, temperature=1, LM head router=false2025.02 | 2.44 | 1.62 | |
| EAGLEmodel=Llama3-8B-Instruct, temperature=1, LM head router=false2025.02 | 2.4 | 1.41 | |
| EAGLEmodel=Qwen2.5-7B-Instruct, temperature=1, LM head router=false2025.02 | 2.25 | 1.27 |