Task-oriented Agent Skill Evolution on SkillFlow
100Cross-Format Data Reconciliation ScoreSelf-Evolve
Evaluation Results
| Method | Links | ||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Self-EvolveHeterogeneity Setting=Backbone, Backbone=Qwen3.6-Plus, Harness=Claude Code2026.06 | 100 | 62.5 | 87.5 | 50 | 62.5 | 22.2 | 62.5 | 50 | 0 | 50 | 25 | 66.7 | 37.5 | 66.7 | 0 | 36.75 | |
| FederatedSkillHeterogeneity Setting=Backbone, Backbone=Qwen3.6-Plus, Harness=Claude Code2026.06 | 100 | 75 | 100 | 50 | 87.5 | 33.3 | 87.5 | 50 | 11.1 | 50 | 37.5 | 77.8 | 37.5 | 77.8 | 0 | 43.37 | |
| FederatedSkillHeterogeneity Setting=Backbone, Backbone=Kimi K2.5, Harness=Claude Code2026.06 | 100 | 50 | 75 | 62.5 | 87.5 | 22.2 | 100 | 50 | 11.1 | 75 | 37.5 | 44.4 | 25 | 88.9 | 0 | 40.96 | |
| Self-EvolveHeterogeneity Setting=Backbone & Harness, Backbone=Qwen3.6-Plus, Harness=Qwen Coder2026.06 | 100 | 62.5 | 100 | 50 | 75 | 66.7 | 100 | 37.5 | 22.2 | 75 | 50 | 77.8 | 25 | 77.8 | 0 | 45.78 | |
| FederatedSkillHeterogeneity Setting=Backbone & Harness, Backbone=Qwen3.6-Plus, Harness=Qwen Coder2026.06 | 100 | 75 | 100 | 50 | 100 | 55.6 | 100 | 62.5 | 22.2 | 62.5 | 37.5 | 66.7 | 50 | 88.9 | 12.5 | 48.8 | |
| FederatedSkillHeterogeneity Setting=Backbone & Harness, Backbone=GLM-5, Harness=Claude Code2026.06 | 100 | 50 | 100 | 50 | 87.5 | 11.1 | 100 | 50 | 11.1 | 75 | 37.5 | 66.7 | 37.5 | 100 | 0 | 43.37 | |
| Self-EvolveHeterogeneity Setting=Backbone, Backbone=GLM-5, Harness=Claude Code2026.06 | 87.5 | 62.5 | 87.5 | 50 | 87.5 | 0 | 100 | 37.5 | 0 | 62.5 | 25 | 33.3 | 25 | 100 | 0 | 37.35 | |
| FederatedSkillHeterogeneity Setting=Backbone, Backbone=GLM-5, Harness=Claude Code2026.06 | 87.5 | 62.5 | 87.5 | 50 | 75 | 22.2 | 100 | 62.5 | 0 | 75 | 37.5 | 66.7 | 50 | 100 | 0 | 43.37 | |
| Self-EvolveHeterogeneity Setting=Backbone, Backbone=Kimi K2.5, Harness=Claude Code2026.06 | 87.5 | 75 | 75 | 37.5 | 62.5 | 11.1 | 87.5 | 62.5 | 22.2 | 62.5 | 25 | 11.1 | 12.5 | 100 | 0 | 36.14 | |
| Self-EvolveHeterogeneity Setting=Backbone & Harness, Backbone=GLM-5, Harness=Claude Code2026.06 | 87.5 | 62.5 | 87.5 | 50 | 87.5 | 0 | 100 | 37.5 | 0 | 62.5 | 25 | 33.3 | 25 | 100 | 0 | 37.35 | |
| Self-EvolveHeterogeneity Setting=Backbone & Harness, Backbone=Kimi K2.5, Harness=Kimi CLI2026.06 | 87.5 | 37.5 | 87.5 | 37.5 | 87.5 | 33.3 | 100 | 62.5 | 0 | 62.5 | 37.5 | 44.4 | 12.5 | 88.9 | 0 | 38.55 | |
| FederatedSkillHeterogeneity Setting=Backbone & Harness, Backbone=Kimi K2.5, Harness=Kimi CLI2026.06 | 75 | 62.5 | 100 | 37.5 | 87.5 | 66.7 | 100 | 50 | 11.1 | 62.5 | 25 | 88.9 | 12.5 | 100 | 0 | 43.98 |