Multi-Agent Collaboration Role Overstepping on SWE total full set (dev)
0.2Overstepping Rate (<INFO>)ChatDev
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| ChatDevCEO Configuration=FT, CPO Configuration=FT, Backbone=Llama, System=ChatDev, Role Clarity Regularization=enabled2026.04 | 0.2 | 0.2 | -43.2 | -44.4 | |
| ChatDevCEO Configuration=FT, CPO Configuration=Base, Backbone=Llama, System=ChatDev, Role Clarity Regularization=enabled2026.04 | 0.65 | 0.65 | -42.75 | -43.95 | |
| ChatDevCEO Configuration=Base, CPO Configuration=FT, Backbone=Llama, System=ChatDev, Role Clarity Regularization=enabled2026.04 | 15.6 | 16.2 | -27.8 | -28.4 | |
| ChatDevCEO Configuration=Base, CPO Configuration=Base, Backbone=Llama, System=ChatDev, Role Clarity Regularization=enabled2026.04 | 43.4 | 44.6 | — | — |