Correlation analysis of validation error versus rollout success on Simulation Environment
0.67Pearson's rRaw MSE
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Raw MSEModel variant family=Data scale2026.06 | 0.67 | 0.9 | |
| Raw MSEModel variant family=Architecture2026.06 | -0.24 | -0.5 | |
| Critical Interval MSEModel variant family=Action head2026.06 | -0.6 | -0.7 | |
| Critical Interval MSEModel variant family=Architecture2026.06 | -0.74 | -1 | |
| Raw MSEModel variant family=Action head2026.06 | -0.75 | -0.7 | |
| Raw MSEModel variant family=VLM backbone2026.06 | -0.83 | -0.4 | |
| Critical Interval MSEModel variant family=VLM backbone2026.06 | -0.96 | -1 | |
| Critical Interval MSEModel variant family=Data scale2026.06 | -0.97 | -0.7 | |
| Critical Interval MSEModel variant family=PEFT2026.06 | -0.99 | -0.94 | |
| Raw MSEModel variant family=PEFT2026.06 | -0.99 | -0.77 | |
| Critical Interval MSEModel variant family=Training steps2026.06 | -1 | -1 | |
| Raw MSEModel variant family=Training steps2026.06 | -1 | -1 |