Question Answering on MultiExpertQA-P Base and Fine-tuned (FT)
85AccuracyLlama 2
Evaluation Results
| Method | Links | |
|---|---|---|
| Llama 2#Exp=2, Expert Variant=Base, Model Family=Llama 2, Model Scale (Parameter Count)=70B2025.11 | 85 | |
| Comp-LLM#Exp=2, Expert Variant=Base, Model Family=Llama 2, Model Scale (Parameter Count)=35B2025.11 | 83 | |
| Comp-LLM#Exp=2, Expert Variant=FT, Model Family=Llama 2, Model Scale (Parameter Count)=35B2025.11 | 83 | |
| Llama 2#Exp=3, Expert Variant=Base, Model Family=Llama 2, Model Scale (Parameter Count)=70B2025.11 | 82 | |
| Comp-LLM#Exp=3, Expert Variant=Base, Model Family=Llama 2, Model Scale (Parameter Count)=42B2025.11 | 81 | |
| Comp-LLM#Exp=3, Expert Variant=FT, Model Family=Llama 2, Model Scale (Parameter Count)=42B2025.11 | 81 | |
| Llama 2#Exp=2, Expert Variant=FT, Model Family=Llama 2, Model Scale (Parameter Count)=34B2025.11 | 79 | |
| Llama 2#Exp=2, Expert Variant=Base, Model Family=Llama 2, Model Scale (Parameter Count)=34B2025.11 | 75 | |
| Llama 2#Exp=3, Expert Variant=FT, Model Family=Llama 2, Model Scale (Parameter Count)=34B2025.11 | 74 | |
| Llama 2#Exp=3, Expert Variant=Base, Model Family=Llama 2, Model Scale (Parameter Count)=34B2025.11 | 70 | |
| Llama 2#Exp=2, Expert Variant=FT, Model Family=Llama 2, Model Scale (Parameter Count)=13B2025.11 | 69 | |
| Llama 2#Exp=2, Expert Variant=Base, Model Family=Llama 2, Model Scale (Parameter Count)=13B2025.11 | 67 | |
| Llama 2#Exp=3, Expert Variant=FT, Model Family=Llama 2, Model Scale (Parameter Count)=13B2025.11 | 67 | |
| Comp-LLM#Exp=2, Expert Variant=Base, Model Family=OPT, Model Scale (Parameter Count)=3.65B2025.11 | 65 | |
| Comp-LLM#Exp=2, Expert Variant=FT, Model Family=OPT, Model Scale (Parameter Count)=3.65B2025.11 | 65 | |
| Llama 2#Exp=2, Expert Variant=FT, Model Family=Llama 2, Model Scale (Parameter Count)=7B2025.11 | 64 | |
| Llama 2#Exp=3, Expert Variant=Base, Model Family=Llama 2, Model Scale (Parameter Count)=13B2025.11 | 64 | |
| OPT#Exp=3, Expert Variant=FT, Model Family=OPT, Model Scale (Parameter Count)=13B2025.11 | 63 | |
| OPT#Exp=2, Expert Variant=FT, Model Family=OPT, Model Scale (Parameter Count)=13B2025.11 | 62 | |
| Comp-LLM#Exp=3, Expert Variant=Base, Model Family=OPT, Model Scale (Parameter Count)=4B2025.11 | 61 | |
| Comp-LLM#Exp=3, Expert Variant=FT, Model Family=OPT, Model Scale (Parameter Count)=4B2025.11 | 61 | |
| OPT#Exp=2, Expert Variant=Base, Model Family=OPT, Model Scale (Parameter Count)=13B2025.11 | 59 | |
| OPT#Exp=3, Expert Variant=Base, Model Family=OPT, Model Scale (Parameter Count)=13B2025.11 | 59 | |
| OPT#Exp=2, Expert Variant=FT, Model Family=OPT, Model Scale (Parameter Count)=6.7B2025.11 | 58 | |
| OPT#Exp=3, Expert Variant=FT, Model Family=OPT, Model Scale (Parameter Count)=6.7B2025.11 | 58 | |
| Llama 2#Exp=2, Expert Variant=Base, Model Family=Llama 2, Model Scale (Parameter Count)=7B2025.11 | 56 | |
| OPT#Exp=2, Expert Variant=Base, Model Family=OPT, Model Scale (Parameter Count)=6.7B2025.11 | 55 | |
| Llama 2#Exp=3, Expert Variant=FT, Model Family=Llama 2, Model Scale (Parameter Count)=7B2025.11 | 53 | |
| OPT#Exp=3, Expert Variant=Base, Model Family=OPT, Model Scale (Parameter Count)=6.7B2025.11 | 51 | |
| Llama 2#Exp=3, Expert Variant=Base, Model Family=Llama 2, Model Scale (Parameter Count)=7B2025.11 | 49 | |
| OPT#Exp=2, Expert Variant=FT, Model Family=OPT, Model Scale (Parameter Count)=1.3B2025.11 | 44 | |
| OPT#Exp=2, Expert Variant=FT, Model Family=OPT, Model Scale (Parameter Count)=350M2025.11 | 41 | |
| OPT#Exp=2, Expert Variant=FT, Model Family=OPT, Model Scale (Parameter Count)=125M2025.11 | 34 | |
| OPT#Exp=2, Expert Variant=Base, Model Family=OPT, Model Scale (Parameter Count)=1.3B2025.11 | 32 | |
| OPT#Exp=3, Expert Variant=FT, Model Family=OPT, Model Scale (Parameter Count)=1.3B2025.11 | 32 | |
| OPT#Exp=3, Expert Variant=FT, Model Family=OPT, Model Scale (Parameter Count)=350M2025.11 | 29 | |
| OPT#Exp=3, Expert Variant=Base, Model Family=OPT, Model Scale (Parameter Count)=1.3B2025.11 | 28 | |
| OPT#Exp=2, Expert Variant=Base, Model Family=OPT, Model Scale (Parameter Count)=350M2025.11 | 26 | |
| OPT#Exp=3, Expert Variant=FT, Model Family=OPT, Model Scale (Parameter Count)=125M2025.11 | 26 | |
| OPT#Exp=3, Expert Variant=Base, Model Family=OPT, Model Scale (Parameter Count)=350M2025.11 | 24 | |
| OPT#Exp=2, Expert Variant=Base, Model Family=OPT, Model Scale (Parameter Count)=125M2025.11 | 21 | |
| OPT#Exp=3, Expert Variant=Base, Model Family=OPT, Model Scale (Parameter Count)=125M2025.11 | 19 |