Question Answering on HotpotQA (Accuracy and Latency)
32AccuracyToT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ToTBackbone=LLAMA2-13B2024.06 | 32 | 1,271 | |
| TS-SFTBackbone=LLAMA2-13B2024.06 | 30.3 | 65.5 | |
| CPOBackbone=LLAMA2-13B2024.06 | 30.3 | 63.8 | |
| ToTBackbone=Mistral-7B2024.06 | 30 | 4,698 | |
| CPOBackbone=Mistral-7B2024.06 | 29.4 | 56.9 | |
| CoTBackbone=LLAMA2-13B2024.06 | 29 | 65.2 | |
| TS-SFTBackbone=Mistral-7B2024.06 | 28.6 | 56.2 | |
| CoTBackbone=Mistral-7B2024.06 | 28 | 58.4 | |
| CPOBackbone=LLAMA2-7B2024.06 | 24 | 41.1 | |
| ToTBackbone=LLAMA2-7B2024.06 | 23 | 1,100.7 | |
| TS-SFTBackbone=LLAMA2-7B2024.06 | 22.7 | 44.8 | |
| CoTBackbone=LLAMA2-7B2024.06 | 21 | 45.5 |