Information Loss Question Answering Generation on INFOLOSSQA 1.0 (evaluation)
89.7Answerable (Source Context Yes)NLI Pipeline
Evaluation Results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| NLI Pipeline2024.01 | 89.7 | 7.8 | 2.5 | 49.5 | 25 | 25.5 | 77.5 | 4.4 | 18.1 | 77.9 | 20.6 | 1.5 | |
| GPT-42024.01 | 87.5 | 12.5 | 0 | 58.5 | 33 | 8.5 | 98 | 2 | 0 | 94 | 6 | 0 | |
| Llama2-70B2024.01 | 83 | 9.7 | 7.3 | 21.4 | 32.5 | 46.1 | 77.7 | 11.7 | 10.7 | 85.9 | 10.7 | 3.4 | |
| Mistral-7B2024.01 | 72.5 | 19.6 | 7.8 | 32.4 | 30.4 | 37.3 | 74.5 | 12.7 | 12.7 | 84.8 | 12.3 | 2.9 |