Long-horizon agentic tasks on BrowseComp-ZH Full
65Pass@1DeepSeek-v3.2
Evaluation results
| Method | Links | |
|---|---|---|
| DeepSeek-v3.2Context Management=Baseline (w/o CM)2026.03 | 65 | |
| Tongyi-DR-30B-A3BContext Management=Baseline (w/o CM)2026.03 | 46.7 |
| Method | Links | |
|---|---|---|
| DeepSeek-v3.2Context Management=Baseline (w/o CM)2026.03 | 65 | |
| Tongyi-DR-30B-A3BContext Management=Baseline (w/o CM)2026.03 | 46.7 |