Multi-hop Question Answering on Bamboogle online Google Search API (test val)
68.7Exact MatchESearch-instruct
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ESearch-instructModel Scale=7B2025.10 | 68.7 | 82.3 | |
| ESearch-instructModel Scale=3B2025.10 | 67.4 | 81.3 | |
| ESearch-baseModel Scale=3B2025.10 | 63.3 | 79.7 | |
| ESearch-baseModel Scale=7B2025.10 | 62.2 | 79.9 | |
| StepSearch-instructModel Scale=7B2025.10 | 60 | 71.8 | |
| Search-R1-baseModel Scale=7B2025.10 | 57.7 | 69.2 | |
| Research-instructModel Scale=7B2025.10 | 54.4 | 66.6 | |
| StepSearch-baseModel Scale=7B2025.10 | 54.4 | 69.8 | |
| Search-R1-instructModel Scale=7B2025.10 | 53.6 | 66 | |
| StepSearch-instructModel Scale=3B2025.10 | 52 | 62.6 | |
| ZeroSearch-baseModel Scale=7B2025.10 | 50.5 | 63.4 | |
| R-Search-instruct-GRPOModel Scale=3B2025.10 | 50.4 | 64.4 | |
| StepSearch-baseModel Scale=3B2025.10 | 50.2 | 63.1 | |
| R-Search-instruct-PPOModel Scale=3B2025.10 | 49.6 | 65.6 | |
| Research-baseModel Scale=7B2025.10 | 48.8 | 58.2 | |
| R-Search-instruct-GRPOModel Scale=7B2025.10 | 46.4 | 60.1 | |
| R-Search-instruct-PPOModel Scale=7B2025.10 | 46.4 | 59.3 | |
| ZeroSearch-instructModel Scale=3B2025.10 | 42.1 | 54.3 | |
| Search-R1-instructModel Scale=3B2025.10 | 39.2 | 51.3 | |
| ZeroSearch-instructModel Scale=7B2025.10 | 36.8 | 49.2 | |
| ZeroSearch-baseModel Scale=3B2025.10 | 36.1 | 52.2 | |
| SSRL-instructModel Scale=3B2025.10 | 34.4 | 45.3 | |
| SSRL-instructModel Scale=7B2025.10 | 33.6 | 43.8 | |
| Search-R1-baseModel Scale=3B2025.10 | 28 | 40 |