Single-Hop Question Answering on TriviaQA (out-of-domain)
68AccuracyGEPO
Evaluation Results
| Method | Links | |
|---|---|---|
| GEPOType=RL Training, Model=Qwen2.5-7B-Instruct2025.10 | 68 | |
| GEPOType=RL Training, Model=Qwen2.5-3B-Instruct2025.10 | 65.2 | |
| GiGPOType=RL Training, Model=Qwen2.5-7B-Instruct2025.10 | 64.7 | |
| EvolveRBackbone=Qwen2.5-7B-Instruct2026.02 | 63.4 | |
| SKILLRLBackbone=Qwen2.5-7B-Instruct2026.02 | 63.3 | |
| ZeroSearchBackbone=Qwen2.5-7B-Instruct2026.02 | 61.8 | |
| ZeroSearchType=RL Training, Model=Qwen2.5-7B-Instruct2025.10 | 61.8 | |
| Search-R1Backbone=Qwen2.5-7B-Instruct2026.02 | 61 | |
| Search-R1Type=RL Training, Model=Qwen2.5-7B-Instruct2025.10 | 61 | |
| GiGPOType=RL Training, Model=Qwen2.5-3B-Instruct2025.10 | 59.5 | |
| RAGBackbone=Qwen2.5-7B-Instruct2026.02 | 58.2 | |
| ZeroSearchType=RL Training, Model=Qwen2.5-3B-Instruct2025.10 | 57.4 | |
| Search-R1Type=RL Training, Model=Qwen2.5-3B-Instruct2025.10 | 54.5 | |
| R1-InstructType=RL Training, Model=Qwen2.5-3B-Instruct2025.10 | 53.7 | |
| R1-InstructBackbone=Qwen2.5-7B-Instruct2026.02 | 44.9 | |
| R1-InstructType=RL Training, Model=Qwen2.5-7B-Instruct2025.10 | 44.9 | |
| Search-o1Backbone=Qwen2.5-7B-Instruct2026.02 | 40.6 | |
| Qwen2.5Backbone=Qwen2.5-7B-Instruct2026.02 | 35.6 | |
| CoTBackbone=Qwen2.5-7B-Instruct2026.02 | 35.6 |