Multi-hop Question Answering on 2WikiMultiHopQA online Google Search API (test val)
63.5Exact MatchESearch-base
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ESearch-baseModel Scale=7B2025.10 | 63.5 | 73 | |
| ESearch-baseModel Scale=3B2025.10 | 58.1 | 68.4 | |
| StepSearch-baseModel Scale=7B2025.10 | 56.1 | 63.8 | |
| ESearch-instructModel Scale=7B2025.10 | 55 | 65.4 | |
| Research-instructModel Scale=7B2025.10 | 53.9 | 62.8 | |
| Search-R1-baseModel Scale=7B2025.10 | 53.3 | 60.7 | |
| ZeroSearch-baseModel Scale=7B2025.10 | 52.5 | 59.3 | |
| ESearch-instructModel Scale=3B2025.10 | 52.1 | 64.4 | |
| R-Search-instruct-PPOModel Scale=3B2025.10 | 49.6 | 55.8 | |
| StepSearch-instructModel Scale=7B2025.10 | 48.5 | 57 | |
| Search-R1-instructModel Scale=7B2025.10 | 47.5 | 56.1 | |
| R-Search-instruct-GRPOModel Scale=7B2025.10 | 47 | 53.5 | |
| R-Search-instruct-PPOModel Scale=7B2025.10 | 46.2 | 52.7 | |
| R-Search-instruct-GRPOModel Scale=3B2025.10 | 45.7 | 51.9 | |
| Research-baseModel Scale=7B2025.10 | 45.7 | 53.4 | |
| StepSearch-baseModel Scale=3B2025.10 | 43.4 | 54.2 | |
| ZeroSearch-instructModel Scale=7B2025.10 | 43.1 | 53.4 | |
| StepSearch-instructModel Scale=3B2025.10 | 40.2 | 49.6 | |
| ZeroSearch-baseModel Scale=3B2025.10 | 39.2 | 47.3 | |
| Search-R1-baseModel Scale=3B2025.10 | 38.1 | 44.5 | |
| Search-R1-instructModel Scale=3B2025.10 | 37.1 | 45.2 | |
| SSRL-instructModel Scale=3B2025.10 | 36.5 | 46.1 | |
| SSRL-instructModel Scale=7B2025.10 | 35.8 | 44.2 | |
| ZeroSearch-instructModel Scale=3B2025.10 | 35.5 | 44.1 |