Step 2의 held-out 9 episode split을 그대로 쓰고, Exp11이 실제로 예측 가능한 공통 valid window subset 50개에서 직접 비교했습니다.
| Path | Exp11 | Step 2 | Delta |
|---|---|---|---|
| center_straight | 0/2 (0.0%) | 0/2 (0.0%) | +0.0% |
| center_left | 0/6 (0.0%) | 0/6 (0.0%) | +0.0% |
| center_right | 6/6 (100.0%) | 4/6 (66.7%) | -33.3% |
| left_straight | 5/6 (83.3%) | 5/6 (83.3%) | +0.0% |
| left_left | 2/7 (28.6%) | 1/7 (14.3%) | -14.3% |
| left_right | 4/7 (57.1%) | 3/7 (42.9%) | -14.3% |
| right_straight | 5/6 (83.3%) | 4/6 (66.7%) | -16.7% |
| right_left | 1/6 (16.7%) | 0/6 (0.0%) | -16.7% |
| right_right | 2/4 (50.0%) | 0/4 (0.0%) | -50.0% |