Simulation deck generation on GEOS v1 (held-out-eval)
78.9TreeSim ScoreSelf-Evolve
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Self-EvolveRetrieval=false, Refine loop=true, Validator=true, Memory=true2026.06 | 78.9 | 0.069 | |
| S+X+MRetrieval=false, Refine loop=true, Validator=true, Memory=true2026.06 | 78.3 | 0.063 | |
| S+XRetrieval=false, Refine loop=true, Validator=true, Memory=false2026.06 | 78.1 | 0.061 | |
| Self-Evolve-proseRetrieval=false, Refine loop=true, Validator=true, Memory=true2026.06 | 77.5 | 0.055 | |
| X+MRetrieval=false, Refine loop=false, Validator=true, Memory=true2026.06 | 76.8 | 0.048 | |
| VanillaRetrieval=false, Refine loop=false, Validator=false, Memory=false2026.06 | 72 | — |