Question Answering on SQuAD (EM, Tokens(k))
35.93EMPIR
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PIRzero-shot-CoT=true, pass@1=true, mode=interactive stronger US (gpt-4o-mini)2026.01 | 35.93 | 0.83 | |
| GPT-4o-minizero-shot-CoT=true, pass@1=true2026.01 | 24.66 | 0.61 | |
| PIRzero-shot-CoT=true, pass@1=true, mode=interactive2026.01 | 22.94 | 0.81 | |
| PIRzero-shot-CoT=true, pass@1=true, mode=self-play2026.01 | 6.98 | 0.99 | |
| Reasoning Basezero-shot-CoT=true, pass@1=true2026.01 | 6.24 | 1.18 |