Long-form deep-research answering on ResearchQA Mini
79.1ScoreOpenAI Deep Research
Evaluation Results
| Method | Links | |
|---|---|---|
| OpenAI Deep ResearchFamily=DR agent2026.05 | 79.1 | |
| GPT-5 + SearchFamily=LLM+search2026.05 | 78.2 | |
| Perplexity Deep ResearchFamily=DR agent2026.05 | 75.1 | |
| WebThinker-32B-DPOFamily=Open DR2026.05 | 74.1 | |
| WebThinker QwQ-32BFamily=Open DR2026.05 | 72.3 | |
| DecomposeR-8BFamily=Ours, Training Stage=SFT + RL2026.05 | 71.4 | |
| Gemini Deep ResearchFamily=DR agent2026.05 | 68.8 | |
| Tongyi DeepResearch-30B-A3BFamily=Open DR2026.05 | 66.7 | |
| WebExplorer-8BFamily=Open search2026.05 | 65.5 | |
| DecomposeR-8BFamily=Ours, Training Stage=SFT only2026.05 | 63.8 | |
| Qwen3-8B + SearchFamily=LLM+search2026.05 | 57 | |
| Search-R1-7BFamily=Open search2026.05 | 28.4 | |
| ASearcher-Web-7BFamily=Open Search2026.05 | 22.5 |