Deep Search on Browsecomp
52AccuracyGLM-4.7
Evaluation Results
| Method | Links | |
|---|---|---|
| GLM-4.7Context Budget=-2026.05 | 52 | |
| OpenAI-o3Context Budget=-2026.05 | 49.7 | |
| GLM-4.6Context Budget=-2026.05 | 49.5 | |
| Qwen3-30B-A3B-thinking-SFT + SAPOContext Budget=128K2026.05 | 42.8 | |
| Tongyi DeepResearchContext Budget=128K2026.05 | 40.7 | |
| Qwen3-30B-A3B-thinking-SFT + GRPOContext Budget=128K2026.05 | 14.9 | |
| Kimi K2Context Budget=-2026.05 | 14.1 | |
| Qwen3-30B-A3B-thinking-SFTContext Budget=128K2026.05 | 13.9 | |
| Web-30B-E-GRPOContext Budget=-2026.05 | 12.9 | |
| WebSailor-32BContext Budget=32K2026.05 | 5.5 | |
| Qwen3-8B-SFT + SAPOContext Budget=64K2026.05 | 4.9 | |
| Qwen3-8B-SFT + GRPOContext Budget=64K2026.05 | 4.3 | |
| Qwen3-8B-SFT + ARPOContext Budget=64K2026.05 | 4.1 | |
| WebExplorer-8BContext Budget=64K2026.05 | 2.9 | |
| Qwen3-8B-SFTContext Budget=64K2026.05 | 2.7 | |
| SAGEBackbone=QWEN-7B, Training data=SAGE (Ours), Search Engine=Google Search, Max queries=202026.01 | 2.6 | |
| MiroThinker-v1.0-8BContext Budget=64K2026.05 | 2.2 | |
| MusiqueBackbone=QWEN-7B, Training data=Musique, Search Engine=Google Search, Max queries=202026.01 | 2.1 | |
| NQ + HotpotQABackbone=QWEN-7B, Training data=NQ + HotpotQA, Search Engine=Google Search, Max queries=202026.01 | 1.6 | |
| NQ + HotpotQABackbone=QWEN-3B, Training data=NQ + HotpotQA, Search Engine=Google Search, Max queries=202026.01 | 1 | |
| MusiqueBackbone=QWEN-3B, Training data=Musique, Search Engine=Google Search, Max queries=202026.01 | 1 | |
| SAGEBackbone=QWEN-3B, Training data=SAGE (Ours), Search Engine=Google Search, Max queries=202026.01 | 1 | |
| WebSailor-7BContext Budget=32K2026.05 | 1 | |
| MiroThinker-v1.5-30BContext Budget=128K2026.05 | 0.7 |