Question Answering on NQ (performance)
97.8PerformanceOur Best Agent
Evaluation Results
| Method | Links | |
|---|---|---|
| Our Best Agent2025.10 | 97.8 | |
| WebSailor2025.10 | 97.6 | |
| DeepResearcher2025.10 | 89.6 | |
| Base agent2025.10 | 88.2 | |
| ASearcher2025.10 | 87.2 | |
| Search-R12025.10 | 79.2 | |
| R1-Searcher2025.10 | 66.2 | |
| Llama 80Bquery placement=reverse2026.01 | 63.13 | |
| GPT-4oquery placement=reverse2026.01 | 61.24 | |
| DeepSeekquery placement=reverse2026.01 | 59.69 | |
| DeepSeekquery placement=vanilla2026.01 | 56.91 | |
| GPT-4oquery placement=vanilla2026.01 | 56.31 | |
| Llama 80Bquery placement=vanilla2026.01 | 54.07 | |
| Qwen2.5 7Bquery placement=vanilla2026.01 | 53.63 | |
| Qwen2.5 7Bquery placement=reverse2026.01 | 53.47 | |
| Llama 8Bquery placement=reverse2026.01 | 52.43 | |
| Mistral 7Bquery placement=reverse2026.01 | 52.27 | |
| Mistral 7Bquery placement=vanilla2026.01 | 51.63 | |
| Llama 8Bquery placement=vanilla2026.01 | 46.7 | |
| ZeroSearch2025.10 | 30 |