Question Answering on HotpotQA (test) (Cost and Coverage Metrics)
0Cost (cents)LLaMA-2 EXIT
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| LLaMA-2 EXITBase model=LLaMA-2-7B, alpha=0.12025.11 | 0 | 65.3 | 2.55 | 2.55 | |
| LLaMA-2-7BBase model=LLaMA-2-7B, alpha=0.12025.11 | 4.551 | 67.5 | 1.03 | 1 | |
| CPOBase model=LLaMA-2-7B, alpha=0.12025.11 | 4.704 | 83.2 | 1.004 | 1 | |
| CCPOlambda=0, Base model=LLaMA-2-7B, alpha=0.12025.11 | 6.552 | 90.2 | 1.34 | 2.35 | |
| RandomBase model=LLaMA-2-7B, alpha=0.12025.11 | 6.811 | 57.8 | 1.44 | 1 | |
| CCPOlambda=2e-4, Base model=LLaMA-2-7B, alpha=0.12025.11 | 7.026 | 90.3 | 1.48 | 2.22 | |
| CPO onlineBase model=LLaMA-2-7B, alpha=0.12025.11 | 7.484 | 89.7 | 1.58 | 2.38 | |
| CPO batchBase model=LLaMA-2-7B, alpha=0.12025.11 | 7.835 | 90.5 | 1.7 | 2.19 | |
| UALABase model=LLaMA-2-7B, alpha=0.12025.11 | 9.153 | 92.3 | 2 | 2 | |
| GPT-4oBase model=LLaMA-2-7B, alpha=0.12025.11 | 18.65 | 78 | 1.2 | 1 | |
| GPT-4o EXITBase model=LLaMA-2-7B, alpha=0.12025.11 | 827 | 90.8 | 2.39 | 2.39 |