Web Agent Task Completion on HLE
37.6AccuracySlimSearcher (SFT+RL)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| SlimSearcher (SFT+RL)Backbone Policy=Tongyi-DeepResearch, Optimization Strategy=SFT+RL2026.06 | 37.6 | 19.82 | 11,418 | |
| SlimSearcher (SFT)Backbone Policy=Tongyi-DeepResearch, Optimization Strategy=SFT2026.06 | 36.6 | 27.16 | 12,936 | |
| Tongyi-DeepResearch2026.06 | 35.8 | 23.92 | 13,664 | |
| Prompt ControlBackbone Policy=Tongyi-DeepResearch, Optimization Strategy=Prompt Control2026.06 | 34.9 | 23.91 | 14,107 | |
| WebLeaperType=Open-weights Agent2026.06 | 29.8 | 24.86 | 11,924 | |
| SlimSearcher (SFT+RL)Backbone Policy=Qwen3-30B-A3B-Instruct-2507, Optimization Strategy=SFT+RL2026.06 | 27.8 | 19.51 | 11,497 | |
| OpenAI DeepResearchType=Proprietary System2026.06 | 26.6 | — | — | |
| SlimSearcher (SFT)Backbone Policy=Qwen3-30B-A3B-Instruct-2507, Optimization Strategy=SFT2026.06 | 25.9 | 27.86 | 13,169 | |
| OpenAI o3Type=Proprietary System2026.06 | 24.9 | — | — | |
| Claude-4-SonnetType=Proprietary System2026.06 | 20.3 | — | — | |
| Qwen3-235B-A22B-InstructType=Open-weights Agent2026.06 | 19.9 | 7.45 | 2,960 | |
| Kimi-K2-Instruct-0905Type=Open-weights Agent2026.06 | 14.6 | 5.17 | 2,349 | |
| WebExplorerType=Open-weights Agent2026.06 | 11.6 | 15.52 | 6,579 | |
| DeepSeek-V3Type=Open-weights Agent2026.06 | 6 | 13 | 2,713 | |
| Qwen3-30B-A3B-Instruct-2507Type=Open-weights Agent, Unified Environment=true2026.06 | 0.9 | 25.77 | 2,570 |