Deep Research on Human-Level English (HLE) (test)
17.6Mean Correct Answer RateOur best agent
Evaluation Results
| Method | Links | |
|---|---|---|
| Our best agentType=Live web2025.10 | 17.6 | |
| WebSailorType=Live web2025.10 | 12.8 | |
| ASearcherType=Live web2025.10 | 11.4 | |
| Search-R1Type=Static & simulated web2025.10 | 11.1 | |
| DeepResearcherType=Live web2025.10 | 10.22 | |
| Our base agentType=Live web2025.10 | 9.6 | |
| ZeroSearchType=Static & simulated web2025.10 | 6.96 | |
| R1-SearcherType=Static & simulated web2025.10 | 4.13 |