By predicting future states, latent world models provide rich self-supervision for end-to-end driving, but often inherit dense features from pretrained visual encoders as their world-state interface. This creates a mismatch: visual pretraining produces information-rich, spatially dense features, but their adaptation for motion-conditioned world modeling and planning is deferred to downstream training. The dense grid also makes future prediction more computationally expensive as visual resolution increases. We therefore argue that video pretraining should directly learn state representations suited to downstream world modeling and planning, rather than relying on downstream training to adapt inherited visual features. We propose MORE, a V-JEPA framework that learns Motion-Oriented REgisters during video pretraining as a shared state interface for driving world models and planning. MORE injects temporally aligned ego motion into a small set of registers and uses them as the sole context for masked patch prediction, channeling dense patch-level supervision through a compact state bottleneck. A complementary blockwise register-masking objective further shapes predictive structure in the register space. During downstream training, the same register space defines the future states predicted by the world model and provides shared scene memory for trajectory generation and scoring. On NAVSIM-v2, MORE outperforms the state of the art by 1.6 in EPDMS, while achieving superior efficiency with only 1/10 the inference latency of DriveLaW. These results support learning the state interface for world modeling and planning directly during video pretraining.
Figures & tables
Figure 1: State-interface mismatch and inference efficiency. (a) Comparison of dense state interfaces and MORE’s compact motion-oriented register interface for world modeling and planning. (b) EPDMS versus inference latency.
Figure 2: Overview of MORE . During pre-training, motion-oriented registers are learned through register-mediated masked patch prediction and blockwise register masking. During joint training, the same register space is used for world modeling and as shared scene memory for trajectory generation and scoring.
Method
Venue
NC ↑
DAC ↑
DDC ↑
TLC ↑
EP ↑
TTC ↑
LK ↑
HC ↑
EC ↑
EPDMS ∗↑
EPDMS ↑
End-to-End-based Methods
TransFuser ( Chitta et al., 2023 )
TPAMI’22
96.9
89.9
97.8
99.7
87.1
95.4
92.7
98.3
87.2
76.7
–
ARTEMIS ( Feng et al., 2026 )
RAL’25
98.3
95.1
98.6
99.8
87.6
97.4
96.5
98.3
89.1
83.1
–
Hydra-MDP++ ( Li et al., 2025a )
arXiv’25
98.8
97.8
99.1
100.0
84.0
95.3
70.1
–
96.8
84.1
–
DiffusionDrive ( Liao et al., 2025 )
CVPR’25
98.2
95.9
99.4
99.8
87.5
97.3
96.8
98.3
87.7
–
84.5
Drive-JEPA ( Wang et al., 2026b )
arXiv’26
98.4
98.6
99.1
99.8
88.4
97.8
97.6
97.9
84.8
87.8
–
Table 1: Comparison on the NAVSIM-v2 benchmark. EPDMS ∗ uses the original NAVSIM-v2 evaluator, while EPDMS uses the official implementation with corrected human-behavior filtering.
Method
RC ↑
HD-Score ↑
E
M
H
X
Avg.
E
M
H
X
Avg.
UniAD ( Hu et al., 2023 )
58.6
41.2
40.4
26.0
40.6
48.7
29.5
27.3
14.3
28.9
VAD ( Jiang et al., 2023 )
38.7
27.0
25.5
23.0
27.9
24.3
9.9
10.4
8.2
12.3
LTF ( Chitta et al., 2023 )
68.4
40.7
36.9
25.5
41.4
52.8
24.6
19.8
8.1
24.8
GTRS-Dense ( Li et al., 2025d )
64.2
50.0
20.7
22.3
38.0
55.5
39.0
11.7
14.3
28.6
DrivoR ( Kirby et al., 2026 )
85.9
36.6
34.1
30.4
43.7
77.2
19.3
16.8
13.9
28.1
Table 2: Photorealistic closed-loop evaluation on HUGSIM. All results use the corrected evaluator with fixes to trajectory-derived heading computation and related issues.
Method
Venue
Sensors
NC ↑
DAC ↑
TTC ↑
C ↑
EP ↑
PDMS ↑
End-to-End-based Methods
UniAD † ( Hu et al., 2023 )
CVPR’23
6C
97.8
91.9
92.9
100.0
78.8
83.4
Transfuser ( Chitta et al., 2023 )
TPAMI’22
3C&L
97.7
92.8
92.8
100.0
79.2
84.0
Hydra-MDP++ ( Li et al., 2025a )
arXiv’25
3C
98.6
98.6
95.1
100.0
85.7
91.0
DiffusionDrive † ( Liao et al., 2025 )
CVPR’25
3C&L
98.2
96.2
94.7
100.0
82.2
88.1
Drive-JEPA ( Wang et al., 2026b )
arXiv’26
1C
98.7
96.2
95.5
100
82.9
89.0
Table 3: Comparison on the NAVSIM-v1 benchmark. C and L denote camera and LiDAR input, respectively. † : The results are dependent on perception-based annotation.
Case
Configuration
Representation Pre-training
Joint Training
Performance
+Reg.
Motion
Patch Ctx.
Reg. Mask
State
Plan. Mem.
EPDMS
PDMS
#1
V-JEPA Baseline
✗
✗
Patch
✗
Patch
Obs.+Pred.
84.9
85.9
#2
V-JEPA + Patch Motion
✗
Patch
Patch
✗
Patch
Obs.+Pred.
89.4
89.2
#3
Joint-Initialized Reg.
✗
✗
Patch
✗
Reg. ∗
Obs.+Pred.
85.7
86.0
#4
Reg. Interaction Only
✓
✗
Patch
✗
Reg.
Obs.+Pred.
85.6
86.3
#5
Register-Mediated Pred.
✓
✗
Reg.
✗
Reg.
Obs.+Pred.
88.5
89.3
Table 4: Ablation Study on NAVSIM navtest . All variants use only the front-camera input.
Register Num.
4
8
16
32
EPDMS ↑
91.4
91.8
91.6
91.5
Table 5: Register-count ablation.
Figure 3: Qualitative trajectory comparison. Trajectories: GT, MORE (Ours), DrivoR.
Appendix figures & tables4 assets
Supplementary material from the paper’s appendix.
Appendix
Method
Img. Size
Obs. Frames
Cams
Params. ↓
Peak Mem. ↓
Latency ↓
EPDMS ↑
Epona
512×1024
10
1
2,915.4 M
14.70 GB
900.0 ms
85.1
World4Drive
384×640
4
6
70.2 M
1.98 GB
414.5 ms
84.8
DriveVLA-W0 †
Pre-tokenized
–
–
9,073.3 M
17.23 GB
1,633.5 ms
86.1
DriveLaW
768×1344
4
1
3,301.6 M
6.89 GB
1,167.0 ms
88.6
CoWorld-VLA
512×1024
1
1
5,372.9 M
10.36 GB
378.4 ms
90.0
Drive-JEPA
256×512
2
1
311.2 M
1.83 GB
73.3 ms
87.8
Appendix
Table 6: Efficiency. We compare model parameters, peak GPU memory, and inference latency with respect to NAVSIM-v2 EPDMS.
Figure 4: Additional qualitative comparison. Trajectories: GT, MORE (Ours), DrivoR.
Figure 5: Pairwise similarity of pretrained registers. Register similarities within and across temporal chunks are shown for (a) configuration #4 in Tab. 4 , which enables register–patch interaction while retaining patch-based prediction, and (b) MORE. Their mean similarities are 0.793 and 0.653, respectively.
Figure 6: Representative failure cases. Trajectories: GT, MORE (Ours).