Downstream skill execution on SkillsBench (test)
48.9RewardGOSKILLS
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GOSKILLSModel=GPT-5.4, Skill Access=Yes, Bounded Retrieval=Yes, Group Context=Yes2026.05 | 48.9 | 694,825 | 352.9 | |
| GOSKILLSModel=Claude Sonnet 4.5, Skill Access=Yes, Bounded Retrieval=Yes, Group Context=Yes2026.05 | 46.8 | 714,360 | 347.6 | |
| GOSKILLSModel=Kimi K2.5, Skill Access=Yes, Bounded Retrieval=Yes, Group Context=Yes2026.05 | 41.7 | 705,812 | 338.4 | |
| Graph of SkillsModel=Claude Sonnet 4.5, Skill Access=Yes, Bounded Retrieval=Yes, Group Context=dependency-graph2026.05 | 38.7 | 421,884 | 571.9 | |
| GOSKILLSModel=Gemini 3 Pro, Skill Access=Yes, Bounded Retrieval=Yes, Group Context=Yes2026.05 | 38.6 | 881,492 | 327.9 | |
| Graph of SkillsModel=GPT-5.4, Skill Access=Yes, Bounded Retrieval=Yes, Group Context=dependency-graph2026.05 | 36.4 | 380,199 | 603.7 | |
| Graph of SkillsModel=Kimi K2.5, Skill Access=Yes, Bounded Retrieval=Yes, Group Context=dependency-graph2026.05 | 33.8 | 455,906 | 545.1 | |
| GOSKILLSModel=Qwen 3.5, Skill Access=Yes, Bounded Retrieval=Yes, Group Context=Yes2026.05 | 33.5 | 756,930 | 421.7 | |
| Graph of SkillsModel=Gemini 3 Pro, Skill Access=Yes, Bounded Retrieval=Yes, Group Context=dependency-graph2026.05 | 31 | 864,577 | 366.2 | |
| Vanilla SkillsModel=Claude Sonnet 4.5, Skill Access=Yes, Bounded Retrieval=No, Group Context=No2026.05 | 29.6 | 905,420 | 641.5 | |
| Vanilla SkillsModel=GPT-5.4, Skill Access=Yes, Bounded Retrieval=No, Group Context=No2026.05 | 28.4 | 832,786 | 686.8 | |
| Graph of SkillsModel=Qwen 3.5, Skill Access=Yes, Bounded Retrieval=Yes, Group Context=dependency-graph2026.05 | 27.6 | 505,220 | 589.3 | |
| Vector SkillsModel=Claude Sonnet 4.5, Skill Access=Yes, Bounded Retrieval=Yes, Group Context=No2026.05 | 26.2 | 610,730 | 707.2 | |
| Vanilla SkillsModel=Kimi K2.5, Skill Access=Yes, Bounded Retrieval=No, Group Context=No2026.05 | 25.6 | 980,115 | 590.7 | |
| Vanilla SkillsModel=Gemini 3 Pro, Skill Access=Yes, Bounded Retrieval=No, Group Context=No2026.05 | 25 | 967,791 | 465.8 | |
| Vector SkillsModel=GPT-5.4, Skill Access=Yes, Bounded Retrieval=Yes, Group Context=No2026.05 | 25 | 569,353 | 742.1 | |
| GOSKILLSModel=MiniMax M2.7, Skill Access=Yes, Bounded Retrieval=Yes, Group Context=Yes2026.05 | 24.3 | 867,452 | 402.5 | |
| Vector SkillsModel=Kimi K2.5, Skill Access=Yes, Bounded Retrieval=Yes, Group Context=No2026.05 | 22.4 | 642,288 | 662.5 | |
| Vanilla SkillsModel=Qwen 3.5, Skill Access=Yes, Bounded Retrieval=No, Group Context=No2026.05 | 20.9 | 1,026,300 | 612.6 | |
| No SkillsModel=Claude Sonnet 4.5, Skill Access=No, Bounded Retrieval=No, Group Context=No2026.05 | 20.8 | 712,460 | 605.4 | |
| Graph of SkillsModel=MiniMax M2.7, Skill Access=Yes, Bounded Retrieval=Yes, Group Context=dependency-graph2026.05 | 19.9 | 560,442 | 518.4 | |
| Vector SkillsModel=Gemini 3 Pro, Skill Access=Yes, Bounded Retrieval=Yes, Group Context=No2026.05 | 19.3 | 894,640 | 357.3 | |
| No SkillsModel=GPT-5.4, Skill Access=No, Bounded Retrieval=No, Group Context=No2026.05 | 18.6 | 612,870 | 748.4 | |
| Vector SkillsModel=Qwen 3.5, Skill Access=Yes, Bounded Retrieval=Yes, Group Context=No2026.05 | 18.1 | 721,480 | 681.7 | |
| Vanilla SkillsModel=MiniMax M2.7, Skill Access=Yes, Bounded Retrieval=No, Group Context=No2026.05 | 17.2 | 942,113 | 580.7 | |
| No SkillsModel=Kimi K2.5, Skill Access=No, Bounded Retrieval=No, Group Context=No2026.05 | 16.2 | 698,440 | 603.8 | |
| No SkillsModel=Gemini 3 Pro, Skill Access=No, Bounded Retrieval=No, Group Context=No2026.05 | 14.8 | 724,306 | 431.6 | |
| No SkillsModel=Qwen 3.5, Skill Access=No, Bounded Retrieval=No, Group Context=No2026.05 | 11.5 | 754,900 | 640.2 | |
| Vector SkillsModel=MiniMax M2.7, Skill Access=Yes, Bounded Retrieval=Yes, Group Context=No2026.05 | 10.4 | 852,881 | 552.9 | |
| No SkillsModel=MiniMax M2.7, Skill Access=No, Bounded Retrieval=No, Group Context=No2026.05 | 8.9 | 808,420 | 621 |