Question Answering on PopQA out-of-domain
46.9Exact MatchSearch-R1-7B-Base-PPO w/ Format
Evaluation Results
| Method | Links | |
|---|---|---|
| Search-R1-7B-Base-PPO w/ FormatModel Scale=7B, Training Algorithm=PPO, Input Formatting=Included2025.10 | 46.9 | |
| EM+Info-Think+Think-AnsFramework=VERITAS-R1, Reward Components=EM+Info-Think+Think-Ans2025.10 | 46.6 | |
| EM+Think-AnsFramework=VERITAS-R1, Reward Components=EM+Think-Ans2025.10 | 46.4 | |
| EM+Info-ThinkFramework=VERITAS-R1, Reward Components=EM+Info-Think2025.10 | 46.3 | |
| Search-R1-7B-Base-PPOModel Scale=7B, Training Algorithm=PPO2025.10 | 45.7 | |
| ReSearch-7B-InstructModel Scale=7B, Model Variant=Instruct2025.10 | 45 | |
| ReSearch-7B-BaseModel Scale=7B, Model Variant=Base2025.10 | 44.2 | |
| DeSA-7B-Instrct-GRPOModel Scale=7B, Training Algorithm=GRPO2025.10 | 44 | |
| DeSA-3B-Instruct-GRPOModel Scale=3B, Training Algorithm=GRPO2025.10 | 39.7 | |
| RAGReasoning Strategy=Retrieval-Augmented Generation2025.10 | 39.2 | |
| Rejection SamplingReasoning Strategy=Rejection Sampling2025.10 | 38 | |
| IRCoTReasoning Strategy=Iterative Retrieval Chain-of-Thought2025.10 | 30.1 | |
| R1-baseModel Variant=Base2025.10 | 20.2 | |
| R1-instructModel Variant=Instruct2025.10 | 19.9 | |
| Direct InferenceReasoning Strategy=Direct Inference2025.10 | 14 | |
| Search-o1Reasoning Strategy=Search-o12025.10 | 13.1 | |
| SFTTraining Algorithm=Supervised Fine-Tuning2025.10 | 12.1 | |
| CoTReasoning Strategy=Chain-of-Thought2025.10 | 5.4 |