Recent end-to-end driving systems demonstrate strong performance on closed-loop benchmarks, yet are still predominantly trained on fixed expert-collected data using open-loop imitation learning. This training-inference mismatch leaves the policy vulnerable in policy-induced states, where accumulated errors can lead to safety-critical failures. A promising post-training approach to overcome this issue is Dataset Aggregation (DAgger), which gathers expert demonstrations in policy-induced states and subsequently fine-tunes the policy on the resulting aggregated dataset. Existing driving DAgger pipelines, however, face three challenges: i) the expert is restricted to a limited trajectory-and-speed solution space, ii) takeover may occur too early or too late relative to impending failures, and iii) privileged expert decisions may rely on information unavailable to the student. To address this, we introduce RoG-DAgger, a post-training framework that uses short-horizon kinematic rollouts to construct high-quality expert demonstrations in safety-critical states. Specifically, RoG-DAgger expands the expert's trajectory-and-speed solution space and evaluates candidate plans through rollout to construct preventive supervision. Moreover, it uses rollout solvability to time the takeover near the estimated point of no return. Lastly, it aligns the expert's field of view with that of the student to provide student-compatible supervision. Across in-distribution (including long-horizon) and out-of-distribution evaluations, RoG-DAgger improves the end-to-end model SimLingo by 5.3 driving-score points and 6.2 percentage points in success rate on Bench2Drive, doubles its driving score from 22 to 44 on Longest6 v2, and improves out-of-distribution success rate from 55% to 66% on Fail2Drive.
Figures & tables
Figure 1: SimLingo pre-training data (orange) lacks safety-critical interactions as measured by time-to-collision (TTC), limiting the model’s ability to learn how to prevent or recover from such situations. As interactions, however, occur frequently in closed-loop Bench2Drive inference (blue), this constitutes a substantial distribution shift. Samples with TTC > 2s are omitted for clarity.
Figure 2: RoG-DAgger leverages kinematic rollout and introduces three improvements to existing DAgger methods. Panel A shows how rollouts are used to construct trajectory-and-speed supervision and to evaluate candidate expert plans. Panels B1-B3 highlight three modifications: (1) an expanded trajectory-and-speed solution space (red: collision-prone nominal plan, green: improved plan), (2) solvability-aware takeover near the estimated point of no return (yellow star: PONR, red region: no feasible corrective plan remains), and (3) expert-student field-of-view alignment (yellow triangle) for student-learnable supervision.
Figure 3: Overview of the RoG-DAgger training pipeline. Starting from a pretrained E2E policy, we collect post-training data in two independent closed-loop runs (see zoom-in panel) and fine-tune the policy using a mixture of pre- and post-training data. The infraction-based run captures policy-induced failures and expert recovery demonstrations, while the solvability-aware run intervenes before failures become unavoidable to collect corrective behavior (gray hatched bar indicates avoided infraction). A restart trigger is used in both runs but omitted for clarity.
Figure 4: Comparison between experts with fixed and expanded sets of geometric trajectories. In (a), the expert follows a fixed pre-computed route and attempts to avoid the predicted collision through emergency braking. However, due to vehicle dynamics, the ego vehicle cannot decelerate sufficiently before reaching the collision point. In (b), the expert additionally considers alternative geometric trajectories, which are evaluated through rollouts for collision-freeness before selection. Geometric trajectories (red dots in the images) correspond to the routes executed in simulation. TTC and distance are estimated using our rollout model.
Method
Bench2Drive
Longest6 v2
DS ↑
SR ↑
DS ↑
RC ↑
Privileged methods
PDM-Lite ( Zimmerlin et al. 2024 )
97.02
92.27
73
100
RoG-Expert (ours)
93.41
81.81
55
100
Pre-trained-only methods
UniAD ( Hu et al. 2023 )
45.81
16.36
–
–
Table 1: Closed-loop performance on Bench2Drive and Longest6 v2. DS, SR, and RS denote Driving Score, Success Rate, and Route Completion, respectively. RoG-DAgger achieves strong performance across both benchmarks. Dash indicates an unreported result.
Method
Base
Generalization
DS ↑
SR ↑
HM ↑
DS ↑
SR ↑
HM ↑
UniAD
47.5
36.3
41.2
44.0
27.6
33.9
TF ++
83.3
78.5
80.8
75.4
61.1
67.5
ORION
53.0
52.0
52.5
51.2
46.0
48.5
HiP-AD
74.1
70.7
72.4
67.1
56.7
61.5
SimLingo
82.6
79.3
80.9
71.7
55.0
62.2
Table 2: Closed-loop performance on Fail2Drive. Base denotes in-distribution evaluation, while Generalization refers to distribution-shifted scenarios. DS, SR, and HM denote Driving Score, Success Rate, and their harmonic mean, respectively. RoG-DAgger remains competitive on base scenarios while showing improved OOD generalization over SimLingo.
Ablation
Bench2Drive
Traj.- speed
Solv.- aware
FoV alignment
DS ↑
SR ↑
Component
−
+
+
88.71
68.94
+
−
+
88.65
69.86
+
+
−
88.96
69.86
Takeover data
Post-takeover only
88.05
68.04
PDM-Lite DAgger baseline
87.85
66.66
Table 3: Ablation study on Bench2Drive. We evaluate the effects of trajectory-and-speed space expansion, solvability-aware triggering, FoV alignment, takeover data selection and expert model.
Figure 5: Qualitative comparison of RoG-DAgger with SimLingo on a Bench2Drive scene. At t1 , both policies approach a junction while a cross-traffic fire truck is already visible. Both policies continue to accelerate, likely because the pre-training data contains primarily successful junction traversals and lacks corrective supervision for such hazardous interactions. As a result, SimLingo keeps accelerating at t2 and collides with the fire truck at t3 . In contrast, after RoG-DAgger post-training, the policy slows down at t2 and shifts to the left, avoiding the collision at t3 . Red dots denote the predicted geometric trajectory, while green dots denote the predicted temporal trajectory. “thr”: throttle, “brk”: brake.
Appendix figures & tables11 assets
Supplementary material from the paper’s appendix.
Appendix
Model
DS
SR (%)
SimLingo ( Renz et al. 2025 )
85.07
67.27
SimLingo-Qwen3-VL-2B ( ours )
86.59
66.36
Appendix
Table 4: Comparison between the original SimLingo and our Qwen3-VL-2B-based version. Both variants achieve comparable driving performance despite the change in VLM backbone. DS and SR denote driving score and success rate, respectively.
Failure Category
Count
Proportion (%)
Total recorded failures
73
–
Ignored cases
5
–
Analyzed failures
68
100.0
Collision-related failures
53
77.9
Collision with vehicle
49
72.1
Collision with static
4
5.9
Appendix
Table 5: Breakdown of the recorded failures. After excluding five Yield To Emergency Vehicle cases, we analyze 68 failures, of which 53 are collision-related. The collision subcategories are not mutually exclusive, as one failure involves both a vehicle and a static object.
Split
Window
Samples
Collision
Rate
Pre-takeover
20
5568
2404
43.18%
Pre-takeover
40
9565
2829
29.58%
Post-takeover
40
20845
99
0.47%
Appendix
Table 6: Collision statistics for pre- and post-takeover samples. For pre-takeover samples, we compare temporal windows of 20 and 40 frames preceding the trigger, while the post-takeover window is 40. Collision rate denotes the fraction of retained samples for which our rollout model predicts a collision within the corresponding window when executing the trajectory-and-speed plan provided by PDM-Lite.
Trigger
Trigger Condition
Infraction
A standard CARLA infraction, including a collision, red-light violation, or out-of-lane violation.
Solvability-Aware Takeover
Replan Success: R-TTC<1sAND replanning succeeds. Repeated Replan Success: 1s<R-TTC<1.5sAND replanning succeeds in two of four consecutive frames. Replan Failure: R-TTC<1sAND replanning fails.
Restart
A persistent discrepancy is detected between the ego vehicle and the expert model: the ego speed remains below 0.1m/s for all of the previous 200 frames, while both the mean expert target speed over these frames and the current expert target speed exceed 0.4m/s .
Appendix
Table 7: Details on the three types of trigger conditions used for post-training data collection. R-TTC denotes the time-to-collision estimated by our rollout model .
Figure 6: Expert without expanded trajectory-and-speed solution space fails to prevent collision. An expert such as PDM-Lite follows a precomputed route and primarily responds to predicted collisions with full braking along the nominal path. However, in safety-critical states, especially those induced by the policy, braking alone does not necessarily prevent an impending collision. At t1 , the ego vehicle controlled by expert follows the nominal route and prepares for lane change. At t2 , the expert predicts a collision and initiates full braking, but the intervention is insufficient due to the vehicle dynamics, and a collision occurs at t4 . The geometric trajectories shown as red dots in the lower row correspond to the trajectories illustrated in the upper row.
Figure 8: SimLingo with solvability-aware takeover. At t1 , the policy controls the ego vehicle while preparing for a left turn. As oncoming traffic approaches, our kinematic rollout model predicts at t2 a time-to-collision (TTC) of 0.05s , indicating an imminent collision. This activates the solvability-aware takeover, which shifts the geometric trajectory 0.5m to the right while setting the target speed to 50% of the current speed. At t3 , the RoG-Expert applies a second, larger correction of 0.9m to the right, again setting the target speed to 50% of the current speed, successfully avoiding the collision. At t4 , the ego vehicle continues safely under expert control during the post-takeover period.
(a) Takeover Phase
Pre-takeover
27.3%
Post-takeover
72.7%
(b) Collection Trigger / Event Type
Infraction
55.3%
Collision
47.7%
Red-Light Infraction
2.9%
Appendix
Table 8: Composition of the post-training dataset from two different perspectives: (a) takeover phase and (b) collection trigger and event type. Percentages are normalized separately within each panel. Bold rows in (b) denote high-level trigger classes, while indented rows show their event-type breakdown .
Scenario Class
SimLingo
RoG-DAgger
Δ
Visual-longitudinal
71.1
81.2
+10.1
Visual-lateral
45.9
51.3
+5.4
Behavior
31.2
62.0
+30.8
Robustness
86.8
80.7
-6.1
Mean
58.8
68.8
+10.0
Appendix
Table 9: Per-class harmonic mean of DS and SR on generalization scenarios of Fail2Drive. RoG-DAgger improves over SimLingo on three out of four scenario classes, with the largest gain on behavior generalization (+30.8 HM), while SimLingo performs better on robustness scenarios.
Method
Efficiency
Comfort
Multi-Capability (%)
Merging
Overtaking
Emergency Brake
Give Way
Traffic Sign
Mean
Privileged methods
PDM-Lite ( Zimmerlin et al. 2024 )
213.47
22.33
82.05
91.11
81.67
80.00
70.97
81.16
RoG-Expert (ours)
196.89
14.13
75.00
84.44
93.33
40.00
88.95
76.35
Pre-trained-only methods
UniAD ( Hu et al. 2023 )
129.21
43.58
12.16
20.00
23.64
10.00
13.89
15.89
Appendix
Table 10: Detailed Bench2Drive evaluation of efficiency, comfort, and multi-ability performance. RoG-DAgger improves over SimLingo in merging, overtaking, and traffic-sign handling, and achieves a higher overall mean. It performs comparably to TakeVLA overall, with stronger merging but lower emergency-braking performance. Beyond these scenario-specific capabilities, we further assess whether the observed gains preserve desirable overall driving behavior in terms of efficiency and comfort. RoG-DAgger maintains competitive efficiency while improving comfort over both SimLingo and TakeVLA, suggesting that its closed-loop gains do not come at the cost of less comfortable driving.
Figure 9: Head-to-head comparison of RoG-DAgger (bottom) and SimLingo (top) in the Construction Pedestrian scenario of Fail2Drive. SimLingo tends to steer left despite oncoming traffic approaching from the left while the right side of the construction site is completely free. This behavior is consistent with a training bias in which construction zones predominantly appear on the right side of the road. As a result, SimLingo may resort to a late evasive maneuver and collide with the traffic warning sign while attempting to avoid the oncoming vehicles. In contrast, RoG-DAgger learns to steer right in the same situation. This behavior can be likely attributed to the RoG-Expert, whose expanded trajectory-and-speed solution space exposes the student to multiple feasible recovery strategies and thereby encourages greater trajectory multimodality.
Figure 11: Qualitative examples of RoG-DAgger on Bench2Drive. In row (A), RoG-DAgger successfully merges into traffic while executing a left turn, using emergency braking to avoid a collision with the lead vehicle. In row (B), the predicted geometric trajectory adapts laterally from left to right as the pedestrian moves across the route. This contrasts with PDM-Lite/SimLingo-style behavior, where the geometric path remains largely fixed and collision avoidance is handled primarily through speed adjustment. Row (C) shows a failure case: RoG-DAgger slows down and attempts a lane change to pass through the construction area, however, due to the limited field of view, a vehicle approaching from the rear-left remains unobserved and causes a side collision .
As scaling pre-training data alone yields diminishing returns, post-training is becoming increasingly important across physical AI domains such as autonomous driving. End-to-end driving policies are pre-trained in open loop with behavior cloning on human demonstrations. However, compounding errors during closed-loop deployment can take the vehicle outside the training data distribution, increasing the risk of safety-critical incidents. Closed-loop post-training can mitigate this risk but requires costly simulation for sensor-based policies. We propose OPTED (on-policy fine-tuning for end-to-end driving) which decouples reinforcement learning from the post-training of the end-to-end policy: a privileged teacher is trained using RL on vectorized inputs (HD-map and bounding boxes). This teacher then provides supervision to the pre-trained student during closed-loop post-training. We apply OPTED to two camera-based models, TransFuser and VaVAM, and fine-tune them in AlpaSim, using neural reconstructions (3DGS) of real driving logs. Driving scores increase by factors of 1.6× and 9.5×, respectively. In controlled experiments OPTED matches closed-loop performance with approximately three orders of magnitude fewer simulator interactions than direct RL post-training, while staying closer to the human prior. Project page: https://01dami23.github.io/opted/
Safe autonomous driving requires more than accurate planning at a single frame: it demands reliable closed-loop execution in dynamic environments. Although end-to-end driving policies have advanced rapidly, the accumulated closed-loop errors, especially in unseen scenarios, can push vehicles into states from which the same policy struggles to recover. We address the challenge by treating closed-loop driving safety as a test-time execution problem. We introduce AutoAgent0, an agentic runtime and harness framework that augments existing driving experts with complementary detection skills, rule-based action verification, and feedback-driven recovery. To address inaccurate scene perception, an independent detection module constructs scene context using complementary perception skills, and a rule-based verifier checks proposed trajectories against this context before execution. When the nominal trajectories are unreliable, the orchestrator of the runtime activates Recovery Planning, where an agentic loop constructs alternative action proposals from action primitives and iteratively refines them using verification feedback. This design combines independent scene assessment with feasibility checks and online recovery, addressing both perception errors and the limited recovery capabilities of individual experts. Evaluations on Fail2Drive and NavSafe-∞ demonstrate improved closed-loop robustness over single-expert execution in long-tail and safety-critical scenarios. These results demonstrate that runtime supervision can improve closed-loop robustness across multiple driving experts without retraining the underlying policies.
End-to-end autonomous driving policies are commonly trained via imitation learning on logged demonstrations without observing the consequences of their own actions, leading to causal confusion in closed-loop real-world deployment. To address this issue, reinforcement learning (RL) post-training offers a promising alternative by leveraging world models as interactive training environments to enable future scene generation for policy improvement. Nevertheless, existing approaches either rely on reconstruction-based simulators, offering limited counterfactual interaction, or adopt synthetic simulators to enable long-horizon closed-loop interaction at the cost of a substantial sim-to-real gap. Recently, video world models have exhibited the ability to generate realistic multi-step future rollouts but may not faithfully reflect action conditions, resulting in action-vision mismatch. In this paper, we introduce RoXDrive, a plug-and-play closed-loop RL framework that enables reliable policy optimization by identifying action-faithful world-model rollouts, consisting of two stages: 1) Model pre-training: In addition to imitation-based policy pre-training, we devise an Action-Vision Faithfulness Evaluator for inverse dynamics estimation with our geometry-aware auxiliary trajectory supervision, enabling long-horizon assessment of whether visual dynamics faithfully reflect the conditioning ego actions. 2) Action-faithful RL post-training: Agents iteratively interact with world models to form long-horizon scene rollouts, retaining only action-faithful ones for dense safety-aware scoring and scene-level closed-loop RL post-training. Extensive experiments on nuScenes and an in-house dataset with over 130K training scenarios demonstrate consistent gains across planners, reducing safety violations by 27.6% with DiffusionDrive on nuScenes and 33.7% with Qwen3-VL on the internal data.
Hongbin Lin, Chaoda Zheng, Yiming Yang +11
The Chinese University of Hong Kong, Shenzhen · XPeng Motors · Xidian University