Scientific Question Answering on Science & QA Domain Out-of-Domain
3.19SampleQA ScoreReinforce++
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Reinforce++training_domain=Math Domain2026.02 | 3.19 | 4.53 | 3.71 | 3.81 | 38.1 | |
| Basetraining_domain=Math Domain2026.02 | 3.12 | 3.8 | 2.89 | 3.27 | 37.18 | |
| Dr. GRPOtraining_domain=Math Domain2026.02 | 3.12 | 4.35 | 3.48 | 3.65 | 38.23 | |
| BAPOtraining_domain=Math Domain2026.02 | 3.12 | 4.35 | 3.06 | 3.51 | 34.48 | |
| GRPOtraining_domain=Math Domain2026.02 | 2.91 | 4.17 | 3.57 | 3.55 | 35.39 | |
| KTAEtraining_domain=Math Domain2026.02 | 2.91 | 3.99 | 3.38 | 3.43 | 34.62 | |
| lambda-GRPOtraining_domain=Math Domain2026.02 | 2.91 | 4.35 | 3.71 | 3.66 | 37.51 | |
| FlowRLtraining_domain=Math Domain2026.02 | 2.84 | 3.8 | 3.56 | 3.4 | 38.78 | |
| DFPOtraining_domain=Math Domain2026.02 | 2.84 | 4.53 | 4.22 | 3.86 | 39.34 | |
| PPOtraining_domain=Math Domain2026.02 | 2.7 | 3.99 | 3.34 | 3.34 | 34.8 | |
| Robust Bellmantraining_domain=Math Domain2026.02 | 2.61 | 3.62 | 3.43 | 3.22 | 35.98 |