Question Answering on HotpotQA single-predictor (test)
60.4Exact MatchTree contrastive
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Tree contrastiveValidation slice=500-example2026.06 | 60.4 | 9 | |
| MIPROv2-lightMode=instruction-only2026.06 | 59.4 | 8 | |
| GEPA-lightOptimizer validation slice=100-example2026.06 | 57 | 5.6 | |
| Baseline promptSettings=instruction-only2026.06 | 51.4 | — |