Multi-hop Question Answering on 2WikiMultiHopQA offline Wiki-18 (test val)
43.6Exact MatchESearch-base
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ESearch-baseModel Scale=7B2025.10 | 43.6 | 51.3 | |
| ESearch-baseModel Scale=3B2025.10 | 42.8 | 49.9 | |
| ESearch-instructModel Scale=7B2025.10 | 41.9 | 49.4 | |
| ESearch-instructModel Scale=3B2025.10 | 41.5 | 50 | |
| StepSearch-baseModel Scale=7B2025.10 | 38.5 | 45 | |
| StepSearch-instructModel Scale=7B2025.10 | 36.6 | 43.1 | |
| ZeroSearch-instructModel Scale=7B2025.10 | 36 | 42.2 | |
| Research-instructModel Scale=7B2025.10 | 35.4 | 41.6 | |
| Search-R1-baseModel Scale=7B2025.10 | 35 | 41.1 | |
| R-Search-instruct-GRPOModel Scale=7B2025.10 | 34.6 | 40.1 | |
| StepSearch-baseModel Scale=3B2025.10 | 33.9 | 39.5 | |
| SSRL-instructModel Scale=7B2025.10 | 33.2 | 39.9 | |
| StepSearch-instructModel Scale=3B2025.10 | 32 | 38.5 | |
| Search-R1-instructModel Scale=7B2025.10 | 31.2 | 37.6 | |
| R-Search-instruct-GRPOModel Scale=3B2025.10 | 30.7 | 35.1 | |
| ZeroSearch-baseModel Scale=7B2025.10 | 29.7 | 35.6 | |
| Search-R1-instructModel Scale=3B2025.10 | 29.3 | 35.2 | |
| SSRL-instructModel Scale=3B2025.10 | 29 | 34.8 | |
| R-Search-instruct-PPOModel Scale=3B2025.10 | 27.7 | 32.8 | |
| R-Search-instruct-PPOModel Scale=7B2025.10 | 27.4 | 33.9 | |
| Research-baseModel Scale=7B2025.10 | 26.4 | 31.3 | |
| ZeroSearch-baseModel Scale=3B2025.10 | 25.3 | 31.1 | |
| Search-R1-baseModel Scale=3B2025.10 | 24.8 | 29.6 | |
| ZeroSearch-instructModel Scale=3B2025.10 | 23.9 | 28.8 |