Conversational Question Answering on COQA (AIBC)
86.5AIBCAutoMix + T
Evaluation results
| Method | Links | |
|---|---|---|
| AutoMix + TSLM=LLAMA2-13B, Router=Thresholding2023.10 | 86.5 | |
| AutoMix + PSLM=LLAMA2-13B, Router=POMDP-based2023.10 | 83.1 | |
| AutoMix + PSLM=GPT-3.5, Router=POMDP-based2023.10 | 65 | |
| AutoMix + TSLM=GPT-3.5, Router=Thresholding2023.10 | 62.7 | |
| FrugalGPTSLM=GPT-3.52023.10 | 37 | |
| HybridLLMSLM=GPT-3.52023.10 | 20.1 | |
| AutoMix + PSLM=MISTRAL-7B, Router=POMDP-based2023.10 | 12.4 | |
| AutoMix + TSLM=MISTRAL-7B, Router=Thresholding2023.10 | 12.1 | |
| HybridLLMSLM=MISTRAL-7B2023.10 | 10.5 | |
| FrugalGPTSLM=LLAMA2-13B2023.10 | 2.6 | |
| HybridLLMSLM=LLAMA2-13B2023.10 | -0.5 | |
| FrugalGPTSLM=MISTRAL-7B2023.10 | -16.7 |