Long-Context Question Answering on SQuAD 512
0.859Answer F1IN-CONTEXT
Evaluation Results
| Method | Links | |
|---|---|---|
| IN-CONTEXT2026.02 | 0.859 | |
| SHINETraining Data Scale=Full, Training Data Source=MQA + 1QA2026.02 | 0.534 | |
| SHINETraining Data Scale=1/2, Training Data Source=1QA2026.02 | 0.509 | |
| SHINETraining Data Scale=1/2, Training Data Source=MQA2026.02 | 0.492 | |
| GEN ADAPTER2026.02 | 0.488 | |
| NAIVE2026.02 | 0.22 |