Pith. sign in

REVIEW 3 major objections 5 minor 2 cited by

Reinforced Refinement with Self-Aware Expansion for End-to-End Autonomous Driving

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read R2SE: a self-aware refinement loop that turns imitation-learned end-to-end driving into a hard-case specialist without forgetting the generalist.

desk verdict Plausible refinement pipeline with strong experiments, but the paper as written inverts its own expansion gate and contains other sign errors, so the executed test-time policy is underspecified. read the letter →

arxiv 2506.09800 v1 pith:ENOVKCSF submitted 2025-06-11 cs.RO

classification cs.RO
keywords end-to-endautonomousdrivingreinforcementlearningimitationcatastrophicforgettinglow-rankadaptersuncertaintyestimationclosed-loopsimulationhard-caserefinement
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper proposes a learning pipeline, R2SE, that refines an already-trained end-to-end driving policy on hard, failure-prone cases while preserving the general driving skill learned by imitation. The motivation is that pure imitation learning generalizes poorly to corner cases, while pure reinforcement learning overfits and catastrophically forgets general knowledge. R2SE's central claim is that a residual correction trained by cost-constrained reinforcement on a small set of hard clips, combined with a test-time uncertainty gate, can improve hard-case performance and keep the generalist intact. The authors report closed-loop gains on the NAVSIM navtest leaderboard (PDMS from 87.0 for the Hydra-MDP baseline to 91.6) and on CARLA Bench2Drive, with lower forgetting rates than standard fine-tuning, EWC, or dynamic networks.

What carries the argument

The load-bearing machinery is a three-part closed loop: (1) a difficulty score $F_X$ that fuses the closed-loop PDM Score, normalized perception loss, and SOTIF Shannon entropy to select the hard-case subset $\mathcal{D}_{\text{hard}}$; (2) a residual reinforced specialist composed of $K$ low-rank adapters (LoRA) attached to detached generalist weights, trained by GRPO with a constrained-MDP penalty where rewards and costs derive from non-reactive log simulation; and (3) a self-aware expansion policy that models the specialist's ensemble uncertainty with a Generalized Pareto Distribution (GPD) and activates the specialist only when the GPD cumulative probability exceeds a threshold $\sigma$. The GPD tail model is the piece that lets the system decide when a specialist is confident enough to overrule the generalist, and the adapter ensemble provides the uncertainty estimate for free.

What would settle it

Run the same R2SE-refined policy in a reactive closed-loop simulator (or on a real vehicle) on the navtest scenarios and compare the score gap against the non-reactive PDMS gap: if the refined policy no longer outperforms the baseline, or if the improvement reverses, then the claim that non-reactive PDMS is a faithful training and evaluation signal for hard-case refinement is falsified.

Watch

Extended reading notes

Core claim

The key discovery is that hard-case refinement can be decoupled from general-knowledge retention by freezing the pretrained policy and attaching a small ensemble of low-rank adapters that learn only residual corrections. These adapters are trained with a GRPO-based reinforcement learning objective that includes cost constraints, using non-reactive log simulation to provide dense per-step process rewards and costs. At test time, the system fits a Generalized Pareto Distribution to the ensemble uncertainty observed on the hard-case training set and expands the specialist adapter only when the test input falls within that tail; otherwise it reverts to the generalist. This mechanism yields a PDMS of 91.6 on the NAVSIM navtest split (up from 87.0 for the Hydra-MDP baseline), the highest reported driving score and success rate on CARLA Bench2Drive among compared end-to-end systems, and a lower forgetting rate than full fine-tuning, EWC, and dynamic networks.

Load-bearing premise

The whole refinement loop—selecting hard cases, computing rewards and costs, and evaluating success—uses non-reactive log simulation with the PDM Score as a proxy for real closed-loop driving quality, so if that proxy diverges from true on-road safety, the reported gains may not transfer to real vehicles.

Editorial extensions

If this is right

  • Any pretrained end-to-end planner, whether categorical, Gaussian-mixture, or diffusion-based, can be refined on hard cases by attaching LoRA ensembles and running the same GRPO-plus-cost loop, since the RL objective reduces to a negative log-likelihood for categorical/GMM policies and to conditional-guidance optimization for diffusion policies.
  • The hard-case allocation step means only a small percentage of training data (top 1% by difficulty) drives refinement; the paper shows that roughly 500 hard clips (0.45% of the training set) already produce most of the PDMS gain, so the pipeline is substantially more data-efficient than full RL fine-tuning.
  • Forgetting is structurally bounded: generalist weights are detached and adapters are low-rank, so refinement cannot overwrite the pretrained policy, and the GPD gate reverts to the generalist on out-of-tail inputs.
  • The same adapter ensemble yields uncertainty estimates that plug into online adaptation schemes, as demonstrated by the paper's R2SE-TTT and R2SE-RS variants, which further improve PDMS beyond the base refinement.
  • If the claims hold, the standard practice of fine-tuning an end-to-end driving policy on all available data can be replaced by targeted, logged-only refinement on the few hard cases, which would make continual improvement of deployed systems more practical.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Since non-reactive PDMS is both the training reward and the evaluation metric, the reported gains could partly reflect optimization toward that specific closed-loop proxy; a reactive simulator or real-vehicle test would be needed to confirm that the refinement generalizes to true closed-loop safety.
  • The GPD gate depends on a hard-case uncertainty buffer collected offline; under distribution shift beyond that buffer, the gate may silently revert to the generalist, so continual expansion of the hard-case buffer would be needed for long-term deployment.
  • The data-efficiency numbers suggest that logged-data-only RL refinement of large end-to-end models is feasible in practice; extending R2SE with generative scene or behavior augmentation, as the paper lists as future work, could remove the residual dependence on having hard clips already present in the log.
  • Because the method is architecture-agnostic, it could plausibly be applied to vision-language-model planners or world-model planners that currently address data rarity but not the hard-case underfitting that behavioral cloning leaves unsolved.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes R2SE, a model-agnostic pipeline for refining end-to-end autonomous driving policies. A generalist IL policy is first pretrained; hard cases are identified from closed-loop log simulation using PDMS and an entropy-based uncertainty term; LoRA adapter ensembles are fine-tuned on these cases with GRPO under cost constraints; and at test time a GPD-based uncertainty estimate is used to decide whether to "expand" to the specialist adapters or fall back to the generalist. Experiments are reported on the NAVSIM nuPlan leaderboard and on CARLA Bench2Drive, with ablations over refinement components, adapter ranks, hard-case thresholds, expansion distributions, and expansion confidence. The main reported results are a PDMS gain from 87.0 (Hydra-MDP baseline) to 91.6 for R2SE-MDP, and improved driving score / success rate on Bench2Drive relative to several end-to-end baselines.

Significance. If the mechanism were as described in the prose, R2SE would be a useful contribution: the pipeline is model-agnostic, combines IL pretraining with cost-constrained RL fine-tuning, and explicitly targets anti-forgetting via low-rank adapters and uncertainty-gated expansion. The empirical breadth is a strength: two closed-loop benchmarks, multiple base architectures (categorical and diffusion), test-time adaptation variants, and ablations covering the main design choices. However, the formal specification of the core mechanism is internally inconsistent. The expansion gate in Eq. (15) and Algorithm 2 is inverted relative to the prose and to the ablation trend in Table VIII, and Eqs. (1), (3), and (5) contain sign/inequality errors that affect the stated pretraining objective, the difficulty score, and the hard-case selection rule. Because the expansion step is credited with a +1.9 PDMS gain in Table III and with the anti-forgetting behavior, the manuscript does not currently pin down which policy is actually executed at test time. These issues are fixable, but they are load-bearing for the central claim.

major comments (3)
  1. [§III-C, Eq. (15), Algorithm 2; §IV-C5, Table VIII] The adapter expansion gate is specified with opposite semantics in different places. Eq. (15) uses the specialist when P_GPD(U_test) > σ and the generalist when P_GPD(U_test) ≤ σ, and Algorithm 2 lines 9–12 repeat this. The prose immediately above Eq. (15) states that OOD/high-uncertainty cases should revert to the generalist, with the specialist used only for confident cases. Table VIII reports Expand Rate increasing from 2.09 at σ=0.05 to 82.8 at σ=0.95. Under Eq. (15), increasing σ makes the specialist condition harder to satisfy, so Expand Rate should fall; the observed rise is consistent only with the opposite gate (specialist when P_GPD ≤ σ). Since the expansion component is credited with a +1.9 PDMS gain in Table III (ID-4 vs. ID-0) and with the paper's anti-forgetting claim, the manuscript does not currently specify which policy actually runs at test time. The authors must correct Eq. (15)/Algorithm 2 to match the implemented policy and confirm that the ablation trend in Table VIII is interpreted consistently (a rising Expand Rate cannot simultaneously mean more frequent fallback to the generalist).
  2. [§III-A1, Eq. (1)] The behavior-cloning term is written as max KL(π_exp∥πθ) = max Σ_M π_exp(X) log(π_exp(X)/πθ(X)). This maximizes the divergence from the expert distribution rather than matching it; the standard BC objective is to minimize KL(π_exp∥πθ), which is equivalent to maximizing E_{π_exp}[log πθ(X)] when the entropy of π_exp is constant. As written, Eq. (1) also has the wrong sign for the stated pretraining objective, since subtracting a positive KL term would reward divergence from the expert. This needs correction because the generalist policy is the foundation on which the specialist fine-tuning and expansion are built.
  3. [§III-A2, Eqs. (3)–(5)] The difficulty score has sign problems that contradict the stated monotonic relationship with case difficulty. Eq. (3) defines F_Ent(Ŷ) as Σ_m πθ(ŷ_m|X) log πθ(ŷ_m|X), which is the negative of Shannon entropy; with this sign, high planning uncertainty decreases F_X in Eq. (4) and can even make it negative, exactly opposite to the intended "difficulty increases with uncertainty" behavior. In addition, Eq. (5) selects D_hard = {d | F_X ≤ Q_ϵ(F)}, which is the set of lowest-difficulty cases, while the text and implementation (Sec. IV-A3, "99th percentile (ϵ=1)") describe selecting the hardest cases. The authors should either define F as a loss-like quantity consistently or change the inequality in Eq. (5) so that hard-case allocation matches the reported experiments.
minor comments (5)
  1. [§III-B2, Eq. (10)] The sentence "Following GRPO, we emit the clip term for simplicity, and KL(·||·) term as it already presents in L_pretrain" is unclear: the final objective in Eq. (10) does not show the clip, the KL, or the relative coefficient, so the exact GRPO surrogate used in the experiments is not fully specified.
  2. [Appendix C, Theorem 1] Theorem 1 states a PAC-Bayes bound for the adapter ensemble, but the proof is not actually supplied; the sentence "This follows directly from the PAC-Bayes bound" is not a derivation. If this theorem is meant to support a formal guarantee, the proof should be included.
  3. [§IV-A3, Algorithm 1] Algorithm 1 line 22 writes the combined loss as L_RL + α_Pretrain L_Pretrain, but α_Pretrain is not defined in the main text or in Table IX. Please define this coefficient and state its value for each experimental setup.
  4. [Fig. 5b and §IV-B4] The text states that EWC has overhead "exceeding ×4.1 MFLOPs" while Fig. 5b reports FLOPs values in units of M; please clarify whether the values are mega-FLOPs or relative multipliers, and make the text consistent with the figure.
  5. [§IV-C5, Table VIII] The column ER is introduced as "Expand Rate" but the surrounding text describes it as a measure of fallback to the generalist. Please define ER explicitly and align the terminology with the gate in Eq. (15).

Circularity Check

1 steps flagged · score 4.0 of 10

PDMS closes the loop from hard-case selection to RL reward to headline metric; the reported nuPlan gains are partly the training objective restated as evaluation, though CARLA provides independent support.

  1. fitted input called prediction [Section III-A2 (Eqs. 2, 4), Section III-B2 (Eq. 10), Section IV-A2 (Eq. 17)]
    ""While as the final output, planning feedback FPlan( ˆY,X) requires strict quantification using PDMScore (PDMS) [2]"; "Reward and cost terms follow the standardized process supervisions [74] (Fig. 3b) leveraging planning feedback: Rt =1(t=T) γ^{T−t} Fplan( ˆY,X)"; "the planning trajectories from the E2E model are comprehensively evaluated by PDMS"."

    Eq. 4 defines case difficulty as F_X = (1 − F_Plan(Ŷ,X)) + ...; Eq. 5 selects D_Hard as hard cases with low F_X, i.e., low PDMS. Eq. 10 then fits the specialist adapters by maximizing R_t = γ^{T−t} F_plan(Ŷ,X) over exactly those D_Hard clips. Section IV-A2 reports success with PDMS, the same scalar. Thus the nuPlan 'hard-case refinement' result is the training reward (and the selection criterion) reported back as the evaluation metric; the specialist is not independently tested on a metric it was not trained to maximize. The reduction is partial, not total: navtest is held out, and CARLA DS/SR and the VAD/TCP transfer tables use different metrics, so the central claim retains independent content.

full rationale

The main circularity concern is the single-scalar loop: PDMS is used to identify hard cases (Eqs. 2 and 4), to select D_Hard (Eq. 5), to define the GRPO reward/cost for fitting specialists (Eq. 10), and to evaluate the headline nuPlan result (Eq. 17). Consequently, the +4.6 PDMS gain over the Hydra-MDP baseline is partly the optimization objective restated as the outcome, rather than an independent confirmation of hard-case generalization. The paper does provide genuine external grounding: Bench2Drive uses the different Driving Score and success rate metrics, and the VAD/TCP transfer results are measured on a different base architecture and scenario suite. These independent results prevent the score from rising to 6 or above. The gate-polarity inconsistency between Eq. 15/Algorithm 2 (specialist when P_GPD > σ) and the prose/Fig. 2 (revert to generalist when uncertain), together with the Expand Rate trend in Table VIII, is a serious internal coherence defect that affects whether the expansion mechanism behaves as claimed; however, an inconsistency is not a circular reduction of a prediction to an input, so it does not add to the circularity score. No load-bearing self-citation chain was found: the NAVSIM PDMS benchmark and the Pickands-Balkema-De Haan theorem are external, machine-independent support, and the cited prior works do not supply the central derivation.

Assumptions & free parameters 8 free parameters · 6 assumptions · 0 invented entities

The central claim rests on PDMS as the joint selection, reward, and evaluation signal, on simulation fidelity, and on several hyperparameters tuned on the same benchmark. No new physical entities are introduced.

free parameters (8)
  • beta_Per = 0.1
    Perception difficulty weight in Eq. 4; fixed in Table IX without data-driven justification.
  • beta_Ent = 0.01
    Entropy difficulty weight in Eq. 4; fixed in Table IX.
  • epsilon = 1.00 (top 1%)
    Hard-case allocation percentile in Eq. 5; selected by ablation in Table VII.
  • sigma = 0.75
    Expansion confidence threshold in Eq. 12 and Eq. 15; selected by ablation in Table VIII.
  • K = 6
    Number of LoRA adapters in the ensemble; set in Table IX.
  • r = 16
    LoRA rank; set in Table IX and discussed in Fig. 8.
  • lambda = 1
    Cost penalty multiplier in Eq. 9 and Eq. 10; fixed in Table IX.
  • gamma = 0.99
    Discount factor in Eq. 8; fixed in Table IX.
assumptions (6)
  • domain assumption Non-reactive NAVSIM log simulation is a faithful proxy for real closed-loop driving quality.
    PDMS from non-reactive simulation is used as the RL reward and as the headline evaluation metric; if this proxy diverges from real driving, the claimed safety gains do not transfer.
  • domain assumption The LoRA ensemble variance in Eq. 16 is a calibrated measure of out-of-distribution uncertainty at test time.
    The self-aware expansion gate relies on this variance and on GPD tail fitting to D_Hard; no calibration against true OOD labels is shown.
  • standard math The Pickands-Balkema-de Haan theorem applies to the uncertainty tail at the chosen threshold.
    Appendix C Theorem 2 is a known result; its applicability to the finite-sample uncertainty distribution is assumed.
  • standard math The PAC-Bayes bound from Langford and Shawe-Taylor applies to the LoRA ensemble.
    Appendix C Theorem 1 is a direct restatement of a known bound; it does not independently justify the experimental method.
  • domain assumption The importance sampling ratio IS(X) in Eq. 10 is bounded and valid for the policy ratio.
    Without clipping or a KL penalty, the objective's stability depends on the ratio pi_{theta+Delta theta}/pi_theta remaining well-behaved.
  • ad hoc to paper Increasing the hard-case threshold improves expansion-boundary estimation via the Central Limit Theorem.
    Section IV-C3 invokes CLT for expansion-boundary accuracy, but no formal connection is established.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Reinforced Refinement with Self-Aware Expansion for End-to-End Autonomous Driving." pith.science (2026). https://pith.science/paper/ENOVKCSF

@misc{pith2026250609800,
  author       = {Pith},
  title        = {Pith review of: Reinforced Refinement with Self-Aware Expansion for End-to-End Autonomous Driving},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ENOVKCSF}},
  note         = {Machine review of arXiv:2506.09800}
}
read the original abstract

End-to-end autonomous driving has emerged as a promising paradigm for directly mapping sensor inputs to planning maneuvers using learning-based modular integrations. However, existing imitation learning (IL)-based models suffer from generalization to hard cases, and a lack of corrective feedback loop under post-deployment. While reinforcement learning (RL) offers a potential solution to tackle hard cases with optimality, it is often hindered by overfitting to specific driving cases, resulting in catastrophic forgetting of generalizable knowledge and sample inefficiency. To overcome these challenges, we propose Reinforced Refinement with Self-aware Expansion (R2SE), a novel learning pipeline that constantly refines hard domain while keeping generalizable driving policy for model-agnostic end-to-end driving systems. Through reinforcement fine-tuning and policy expansion that facilitates continuous improvement, R2SE features three key components: 1) Generalist Pretraining with hard-case allocation trains a generalist imitation learning (IL) driving system while dynamically identifying failure-prone cases for targeted refinement; 2) Residual Reinforced Specialist Fine-tuning optimizes residual corrections using reinforcement learning (RL) to improve performance in hard case domain while preserving global driving knowledge; 3) Self-aware Adapter Expansion dynamically integrates specialist policies back into the generalist model, enhancing continuous performance improvement. Experimental results in closed-loop simulation and real-world datasets demonstrate improvements in generalization, safety, and long-horizon policy robustness over state-of-the-art E2E systems, highlighting the effectiveness of reinforce refinement for scalable autonomous driving.

Figures

Figures reproduced from arXiv: 2506.09800 by the authors.

Figure 1
Figure 1. General Learning Framework in E2E ADS. a) IL and RL pipelines directly train a generalist ADS from expert demonstrations or environmental rewards. b) Post-training approaches aim to improve E2E performance through online adaptation or continual learning in novel domains. c)–e) R2SE is established upon c) self-aware hard cases by pretrained ADS generalist. Then, d) a reinforced refinement paradigm in R2SE enhances pe… view at source ↗
Figure 2
Figure 2. System overview of R2SE. R2SE is a reinforced refinement framework designed to improve model-agnostic E2E autonomous driving through adapter expansion. a) It begins by pretraining a generalist policy πθ using sensor inputs X and multi-head outputs Y for perception and planning. b) Hard cases are then identified based on a difficulty metric Fx, capturing uncertain or failed behaviors, and sampled for specialization. … view at source ↗
Figure 3
Figure 3. Residual Reinforced Specialist Learning in R2SE. a) Retrieval-based exploration for DRL; b) GRPO-enabled reinforced finetuning. Set of process reward R and cost C functions evaluate each candidate planning policy yˆ ∈ Yˆ from batched non-reactive simulations using exploration clips of DRL. systems often exhibit fragile failures in scenarios where ADS exhibits increasing planning uncertainty [71], which can be quanti… view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: The baseline E2E ADS that serves as the generalist policy [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Quantitative comparison across refinement frameworks. R2SE is evaluated against state-of-the-art refinement paradigms on navtest in terms of (a) general performance metrics, (b) computational cost, and (c) forgetting robustness. (a) R2SE ranks highest in PDMS, achievin…
Figure 6
Figure 6. Figure 6: R2SE could be seamlessly integrated with online adaptation tech [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: Ablation studies for R2SE trends. a) Convergence effect of increasing corner case threshold ϵ for loosing difficulty criteria. b) A reversed converging trend with stricter expansion confidence σ. c) R2SE consistently improves E2E driving under incremental refinement by…
Figure 8
Figure 8. Figure 8: Effect of Specialist Adapter Rank.R2SE reports fluctuated conver￾gence by increasing adapter ranks. shown in [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]
Figure 9
Figure 9. Figure 9: Qualitative results on navsafe test splits. Model-agnostic R2SE is evaluated against SOTA E2E ADS on nuPlan under NHTSA Typologies. a)-b): Compared to HydraMDP [75], the categorical form of R2SE-MDP can a) more accurately identify under blurred sensory conditions, and …
Figure 10
Figure 10. Figure 10: Qualitative Results on Bench2Drive test scenarios. a) R2SE-refined ADS identifies appropriate nudging windows to overtake cyclist flows. b) At night, the ADS performs an overtaking maneuver by temporarily entering the opposite lane upon detecting construction barriers…
Figure 11
Figure 11. Figure 11: Qualitative Results on Hard Case Improvement by R2SE. t-SNE visualizations colored by corresponding PDMS are presented for navtrain hard cases and navtest test cases. a) R2SE largely improves PDMS on training hard cases with minimal displacement in feature space. b) T…
Figure 12
Figure 12. Figure 12: Challenge and frontier upon R2SE. a) Combat with catastrophic forgetting of RL finetuning (RL-FT); b) Cost / Safe RL explorations are necessary in E2E ADS; c) Data efficiency matters in R2SE compared with scalable RL [60], and could be further boosted by E2E generativ…

Discussion (0). Sign in to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. SpaceDrive: Infusing Spatial Awareness into VLM-based Autonomous Driving

    cs.CV 2025-12 conditional novelty 6.0 of 10

    SpaceDrive replaces textual coordinate tokens with shared 3D positional encodings in a VLM driving planner, achieving state-of-the-art open-loop planning on nuScenes and 78.02 Driving Score on Bench2Drive.

  2. MindDrive: A Vision-Language-Action Model for Autonomous Driving via Online Reinforcement Learning

    cs.CV 2025-12 conditional novelty 5.0 of 10

    An autonomous-driving vision-language model that uses online RL over discrete language actions, with a separate action expert mapping decisions to trajectories, reports DS 78.04 and SR 55.09% on Bench2Drive.

Reference graph

Works this paper leans on

104 extracted references · 30 canonical work pages · cited by 2 Pith papers

  1. [1]

    DECODE: Domain-aware Continual Domain Expansion for Motion Prediction

    B. Li, H. Zhu, and H. X. Liu, “Decode: Domain-aware con- tinual domain expansion for motion prediction,”arXiv preprint arXiv:2411.17917, 2024

  2. [2]

    Navsim: Data- driven non-reactive autonomous vehicle simulation and benchmarking,

    D. Dauner, M. Hallgarten, T. Li, X. Weng, Z. Huang, Z. Yang, H. Li, I. Gilitschenski, B. Ivanovic, M. Pavoneet al., “Navsim: Data- driven non-reactive autonomous vehicle simulation and benchmarking,” Advances in Neural Information Processing Systems, vol. 37, pp. 28 706–28 719, 2024

  3. [3]

    Milestones in autonomous driving and intelligent vehicles: Survey of surveys,

    L. Chen, Y . Li, C. Huang, B. Li, Y . Xing, D. Tian, L. Li, Z. Hu, X. Na, Z. Liet al., “Milestones in autonomous driving and intelligent vehicles: Survey of surveys,”IEEE Transactions on Intelligent Vehicles, 2022

  4. [4]

    End- to-end autonomous driving: Challenges and frontiers,

    L. Chen, P. Wu, K. Chitta, B. Jaeger, A. Geiger, and H. Li, “End- to-end autonomous driving: Challenges and frontiers,”arXiv preprint arXiv:2306.16927, 2023

  5. [5]

    A review of mo- tion planning techniques for automated vehicles,

    D. González, J. Pérez, V . Milanés, and F. Nashashibi, “A review of mo- tion planning techniques for automated vehicles,”IEEE Transactions on intelligent transportation systems, vol. 17, no. 4, pp. 1135–1145, 2015

  6. [6]

    Planning-oriented autonomous driving,

    Y . Hu, J. Yang, L. Chen, K. Li, C. Sima, X. Zhu, S. Chai, S. Du, T. Lin, W. Wanget al., “Planning-oriented autonomous driving,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 17 853–17 862

  7. [7]

    Vad: Vectorized scene representation for efficient autonomous driving,

    B. Jiang, S. Chen, Q. Xu, B. Liao, J. Chen, H. Zhou, Q. Zhang, W. Liu, C. Huang, and X. Wang, “Vad: Vectorized scene representation for efficient autonomous driving,”arXiv preprint arXiv:2303.12077, 2023

  8. [8]

    Hybrid- prediction integrated planning for autonomous driving,

    H. Liu, Z. Huang, W. Huang, H. Yang, X. Mo, and C. Lv, “Hybrid- prediction integrated planning for autonomous driving,”IEEE Trans- actions on Pattern Analysis and Machine Intelligence, 2025

Show all 104 references
  1. [9]

    Corner case generation and analysis for safety assessment of autonomous vehicles,

    H. Sun, S. Feng, X. Yan, and H. X. Liu, “Corner case generation and analysis for safety assessment of autonomous vehicles,”Transportation research record, vol. 2675, no. 11, pp. 587–600, 2021

  2. [10]

    Curse of rarity for autonomous vehicles,

    H. X. Liu and S. Feng, “Curse of rarity for autonomous vehicles,” nature communications, vol. 15, no. 1, p. 4808, 2024

  3. [11]

    Dense reinforcement learning for safety validation of autonomous vehicles,

    S. Feng, H. Sun, X. Yan, H. Zhu, Z. Zou, S. Shen, and H. X. Liu, “Dense reinforcement learning for safety validation of autonomous vehicles,”Nature, vol. 615, no. 7953, pp. 620–627, 2023

  4. [12]

    Dy- namically conservative self-driving planner for long-tail cases,

    W. Zhou, Z. Cao, N. Deng, X. Liu, K. Jiang, and D. Yang, “Dy- namically conservative self-driving planner for long-tail cases,”IEEE Transactions on Intelligent Transportation Systems, vol. 24, no. 3, pp. 3476–3488, 2022

  5. [13]

    Towards corner case detection for autonomous driving,

    J.-A. Bolte, A. Bar, D. Lipinski, and T. Fingscheidt, “Towards corner case detection for autonomous driving,” in2019 IEEE Intelligent vehicles symposium (IV). IEEE, 2019, pp. 438–445

  6. [14]

    Gameformer: Game-theoretic modeling and learning of transformer-based interactive prediction and planning for autonomous driving,

    Z. Huang, H. Liu, and C. Lv, “Gameformer: Game-theoretic modeling and learning of transformer-based interactive prediction and planning for autonomous driving,”arXiv preprint arXiv:2303.05760, 2023

  7. [15]

    Generalized predictive model for autonomous driving,

    J. Yang, S. Gao, Y . Qiu, L. Chen, T. Li, B. Dai, K. Chitta, P. Wu, J. Zeng, P. Luoet al., “Generalized predictive model for autonomous driving,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 14 662–14 672

  8. [16]

    Vista: A generalizable driving world model with high fidelity and versatile controllability,

    S. Gao, J. Yang, L. Chen, K. Chitta, Y . Qiu, A. Geiger, J. Zhang, and H. Li, “Vista: A generalizable driving world model with high fidelity and versatile controllability,”arXiv preprint arXiv:2405.17398, 2024

  9. [17]

    Survey of deep reinforcement learning for motion planning of autonomous vehicles,

    S. Aradi, “Survey of deep reinforcement learning for motion planning of autonomous vehicles,”IEEE Transactions on Intelligent Transporta- tion Systems, vol. 23, no. 2, pp. 740–759, 2020

  10. [18]

    Think2drive: Efficient reinforce- ment learning by thinking with latent world model for autonomous driving (in carla-v2),

    Q. Li, X. Jia, S. Wang, and J. Yan, “Think2drive: Efficient reinforce- ment learning by thinking with latent world model for autonomous driving (in carla-v2),” inEuropean Conference on Computer Vision. Springer, 2024, pp. 142–158

  11. [19]

    Improved deep reinforcement learning with expert demonstrations for urban autonomous driving,

    H. Liu, Z. Huang, J. Wu, and C. Lv, “Improved deep reinforcement learning with expert demonstrations for urban autonomous driving,” in2022 IEEE intelligent vehicles symposium (IV). IEEE, 2022, pp. 921–928

  12. [20]

    Efficient deep reinforcement learning with imitative expert priors for autonomous driving,

    Z. Huang, J. Wu, and C. Lv, “Efficient deep reinforcement learning with imitative expert priors for autonomous driving,”IEEE Transactions on Neural Networks and Learning Systems, vol. 34, no. 10, pp. 7391– 7403, 2022

  13. [21]

    Model-free deep reinforcement learning for urban autonomous driving,

    J. Chen, B. Yuan, and M. Tomizuka, “Model-free deep reinforcement learning for urban autonomous driving,” in2019 IEEE intelligent transportation systems conference (ITSC). IEEE, 2019, pp. 2765– 2771

  14. [22]

    Rad: Training an end-to-end driving policy via large-scale 3dgs-based reinforcement learning,

    H. Gao, S. Chen, B. Jiang, B. Liao, Y . Shi, X. Guo, Y . Pu, H. Yin, X. Li, X. Zhanget al., “Rad: Training an end-to-end driving policy via large-scale 3dgs-based reinforcement learning,”arXiv preprint arXiv:2502.13144, 2025

  15. [23]

    Dtpp: Differentiable joint conditional prediction and cost evalua- tion for tree policy planning in autonomous driving,

    Z. Huang, P. Karkus, B. Ivanovic, Y . Chen, M. Pavone, and C. Lv, “Dtpp: Differentiable joint conditional prediction and cost evalua- tion for tree policy planning in autonomous driving,”arXiv preprint arXiv:2310.05885, 2023

  16. [24]

    Diffstack: A differentiable and modular control stack for autonomous vehicles,

    P. Karkus, B. Ivanovic, S. Mannor, and M. Pavone, “Diffstack: A differentiable and modular control stack for autonomous vehicles,” in Conference on Robot Learning. PMLR, 2023, pp. 2170–2180

  17. [25]

    Centaur: Robust end-to-end autonomous driving with test- time training,

    C. Sima, K. Chitta, Z. Yu, S. Lan, P. Luo, A. Geiger, H. Li, and J. M. Alvarez, “Centaur: Robust end-to-end autonomous driving with test- time training,”arXiv preprint arXiv:2503.11650, 2025

  18. [26]

    Differentiable integrated motion prediction and planning with learnable cost function for autonomous driving,

    Z. Huang, H. Liu, J. Wu, and C. Lv, “Differentiable integrated motion prediction and planning with learnable cost function for autonomous driving,”arXiv preprint arXiv:2207.10422, 2022

  19. [27]

    Continuous improvement of self-driving cars using dynamic confidence-aware reinforcement learning,

    Z. Cao, K. Jiang, W. Zhou, S. Xu, H. Peng, and D. Yang, “Continuous improvement of self-driving cars using dynamic confidence-aware reinforcement learning,”Nature Machine Intelligence, vol. 5, no. 2, pp. 145–158, 2023

  20. [28]

    Continual pedestrian trajectory learning with social generative replay,

    Y . Wu, A. Bighashdel, G. Chen, G. Dubbelman, and P. Jancura, “Continual pedestrian trajectory learning with social generative replay,” IEEE Robotics and Automation Letters, vol. 8, no. 2, pp. 848–855, 2022

  21. [29]

    Human-guided continual learning for personalized decision-making of autonomous driving,

    H. Yang, Y . Zhou, J. Wu, H. Liu, L. Yang, and C. Lv, “Human-guided continual learning for personalized decision-making of autonomous driving,”IEEE Transactions on Intelligent Transportation Systems, 2025

  22. [30]

    Sustainable adaptation for autonomous driving with the mix- ture of progressive experts network,

    Y . Cui, S. Yang, C. Wan, X. Li, J. Xing, Y . Zhang, Y . Huang, and H. Chen, “Sustainable adaptation for autonomous driving with the mix- ture of progressive experts network,”arXiv preprint arXiv:2502.05943, 2025

  23. [31]

    Lora: Low-rank adaptation of large language models

    E. J. Hu, Y . Shen, P. Wallis, Z. Allen-Zhu, Y . Li, S. Wang, L. Wang, W. Chenet al., “Lora: Low-rank adaptation of large language models.” ICLR, vol. 1, no. 2, p. 3, 2022

  24. [32]

    Bevformer: Learning bird’s-eye-view representation from multi-camera images via spatiotemporal transformers,

    Z. Li, W. Wang, H. Li, E. Xie, C. Sima, T. Lu, Y . Qiao, and J. Dai, “Bevformer: Learning bird’s-eye-view representation from multi-camera images via spatiotemporal transformers,” inEuropean conference on computer vision. Springer, 2022, pp. 1–18

  25. [33]

    Beverse: Unified perception and prediction in birds-eye-view for vision-centric autonomous driving,

    Y . Zhang, Z. Zhu, W. Zheng, J. Huang, G. Huang, J. Zhou, and J. Lu, “Beverse: Unified perception and prediction in birds-eye-view for vision-centric autonomous driving,”arXiv preprint arXiv:2205.09743, 2022

  26. [34]

    Stretchbev: Stretching future instance pre- diction spatially and temporally,

    A. K. Akan and F. Güney, “Stretchbev: Stretching future instance pre- diction spatially and temporally,” inEuropean Conference on Computer Vision. Springer, 2022, pp. 444–460

  27. [35]

    Delving into the devils of bird’s-eye-view perception: A review, evaluation and recipe,

    H. Li, C. Sima, J. Dai, W. Wang, L. Lu, H. Wang, J. Zeng, Z. Li, J. Yang, H. Denget al., “Delving into the devils of bird’s-eye-view perception: A review, evaluation and recipe,”IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023

  28. [36]

    Fusionad: Multi-modality fusion for pre- diction and planning tasks of autonomous driving,

    T. Ye, W. Jing, C. Hu, S. Huang, L. Gao, F. Li, J. Wang, K. Guo, W. Xiao, W. Maoet al., “Fusionad: Multi-modality fusion for pre- diction and planning tasks of autonomous driving,”arXiv preprint arXiv:2308.01006, 2023. 16

  29. [37]

    Driveadapter: Breaking the coupling barrier of perception and planning in end-to-end autonomous driving,

    X. Jia, Y . Gao, L. Chen, J. Yan, P. L. Liu, and H. Li, “Driveadapter: Breaking the coupling barrier of perception and planning in end-to-end autonomous driving,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 7953–7963

  30. [38]

    Deepemplanner: An em motion planner with iterative interactions,

    Z. Chen, M. Ye, S. Xu, T. Cao, and Q. Chen, “Deepemplanner: An em motion planner with iterative interactions,”arXiv preprint arXiv:2311.08100, 2023

  31. [39]

    Vadv2: End-to-end vectorized autonomous driving via probabilistic planning,

    S. Chen, B. Jiang, H. Gao, B. Liao, Q. Xu, Q. Zhang, C. Huang, W. Liu, and X. Wang, “Vadv2: End-to-end vectorized autonomous driving via probabilistic planning,”arXiv preprint arXiv:2402.13243, 2024

  32. [40]

    Diffusiondrive: Truncated diffusion model for end-to-end autonomous driving,

    B. Liao, S. Chen, H. Yin, B. Jiang, C. Wang, S. Yan, X. Zhang, X. Li, Y . Zhang, Q. Zhanget al., “Diffusiondrive: Truncated diffusion model for end-to-end autonomous driving,”arXiv preprint arXiv:2411.15139, 2024

  33. [41]

    Gpt-driver: Learning to drive with gpt,

    J. Mao, Y . Qian, H. Zhao, and Y . Wang, “Gpt-driver: Learning to drive with gpt,”arXiv preprint arXiv:2310.01415, 2023

  34. [42]

    Drivevlm: The convergence of autonomous driving and large vision-language models,

    X. Tian, J. Gu, B. Li, Y . Liu, C. Hu, Y . Wang, K. Zhan, P. Jia, X. Lang, and H. Zhao, “Drivevlm: The convergence of autonomous driving and large vision-language models,”arXiv preprint arXiv:2402.12289, 2024

  35. [43]

    A language agent for autonomous driving,

    J. Mao, J. Ye, Y . Qian, M. Pavone, and Y . Wang, “A language agent for autonomous driving,”arXiv preprint arXiv:2311.10813, 2023

  36. [44]

    Drivelm: Driving with graph visual question answering,

    C. Sima, K. Renz, K. Chitta, L. Chen, H. Zhang, C. Xie, P. Luo, A. Geiger, and H. Li, “Drivelm: Driving with graph visual question answering,”arXiv preprint arXiv:2312.14150, 2023

  37. [45]

    Unleashing generalization of end- to-end autonomous driving with controllable long video generation,

    E. Ma, L. Zhou, T. Tang, Z. Zhang, D. Han, J. Jiang, K. Zhan, P. Jia, X. Lang, H. Sunet al., “Unleashing generalization of end- to-end autonomous driving with controllable long video generation,” arXiv preprint arXiv:2406.01349, 2024

  38. [46]

    Driving into the future: Multiview visual forecasting and planning with world model for autonomous driving,

    Y . Wang, J. He, L. Fan, H. Li, Y . Chen, and Z. Zhang, “Driving into the future: Multiview visual forecasting and planning with world model for autonomous driving,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 14 749–14 759

  39. [47]

    Deep reinforcement learning for autonomous driving: A survey,

    B. R. Kiran, I. Sobh, V . Talpaert, P. Mannion, A. A. Al Sallab, S. Yo- gamani, and P. Pérez, “Deep reinforcement learning for autonomous driving: A survey,”IEEE transactions on intelligent transportation systems, vol. 23, no. 6, pp. 4909–4926, 2021

  40. [48]

    Human-guided re- inforcement learning with sim-to-real transfer for autonomous naviga- tion,

    J. Wu, Y . Zhou, H. Yang, Z. Huang, and C. Lv, “Human-guided re- inforcement learning with sim-to-real transfer for autonomous naviga- tion,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 45, no. 12, pp. 14 745–14 759, 2023

  41. [49]

    Augmenting reinforcement learn- ing with transformer-based scene representation learning for decision- making of autonomous driving,

    H. Liu, Z. Huang, X. Mo, and C. Lv, “Augmenting reinforcement learn- ing with transformer-based scene representation learning for decision- making of autonomous driving,”IEEE Transactions on Intelligent Vehicles, 2024

  42. [50]

    Learning to drive from a world on rails,

    D. Chen, V . Koltun, and P. Krähenbühl, “Learning to drive from a world on rails,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 15 590–15 599

  43. [51]

    Feedback-guided autonomous driving,

    J. Zhang, Z. Huang, A. Ray, and E. Ohn-Bar, “Feedback-guided autonomous driving,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 15 000–15 011

  44. [52]

    End-to- end urban driving by imitating a reinforcement learning coach,

    Z. Zhang, A. Liniger, D. Dai, F. Yu, and L. Van Gool, “End-to- end urban driving by imitating a reinforcement learning coach,” in Proceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 15 222–15 232

  45. [53]

    Think twice before driving: Towards scalable decoders for end-to-end autonomous driving,

    X. Jia, P. Wu, L. Chen, J. Xie, C. He, J. Yan, and H. Li, “Think twice before driving: Towards scalable decoders for end-to-end autonomous driving,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 21 983–21 994

  46. [54]

    Adawm: Adaptive world model based planning for autonomous driving,

    H. Wang, X. Ye, F. Tao, C. Pan, A. Mallik, B. Yaman, L. Ren, and J. Zhang, “Adawm: Adaptive world model based planning for autonomous driving,”arXiv preprint arXiv:2501.13072, 2025

  47. [55]

    Tail: Task-specific adapters for imitation learning with large pretrained models,

    Z. Liu, J. Zhang, K. Asadi, Y . Liu, D. Zhao, S. Sabach, and R. Fakoor, “Tail: Task-specific adapters for imitation learning with large pretrained models,”arXiv preprint arXiv:2310.05905, 2023

  48. [56]

    Preventing catas- trophic forgetting and distribution mismatch in knowledge distillation via synthetic data,

    K. Binici, N. T. Pham, T. Mitra, and K. Leman, “Preventing catas- trophic forgetting and distribution mismatch in knowledge distillation via synthetic data,” inProceedings of the IEEE/CVF winter conference on applications of computer vision, 2022, pp. 663–671

  49. [57]

    Self- correcting llm-controlled diffusion models,

    T.-H. Wu, L. Lian, J. E. Gonzalez, B. Li, and T. Darrell, “Self- correcting llm-controlled diffusion models,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 6327–6336

  50. [58]

    Active finetuning: Exploiting annotation budget in the pretraining-finetuning paradigm,

    Y . Xie, H. Lu, J. Yan, X. Yang, M. Tomizuka, and W. Zhan, “Active finetuning: Exploiting annotation budget in the pretraining-finetuning paradigm,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 23 715–23 724

  51. [59]

    Activead: Planning-oriented active learning for end-to-end autonomous driving,

    H. Lu, X. Jia, Y . Xie, W. Liao, X. Yang, and J. Yan, “Activead: Planning-oriented active learning for end-to-end autonomous driving,” arXiv preprint arXiv:2403.02877, 2024

  52. [60]

    Robust autonomy emerges from self-play,

    M. Cusumano-Towner, D. Hafner, A. Hertzberg, B. Huval, A. Petrenko, E. Vinitsky, E. Wijmans, T. Killian, S. Bowers, O. Seneret al., “Robust autonomy emerges from self-play,”arXiv preprint arXiv:2502.03349, 2025

  53. [61]

    Learning to drive via asymmetric self-play,

    C. Zhang, S. Biswas, K. Wong, K. Fallah, L. Zhang, D. Chen, S. Casas, and R. Urtasun, “Learning to drive via asymmetric self-play,” in European Conference on Computer Vision. Springer, 2024, pp. 149– 168

  54. [62]

    Finetuning generative trajectory model with reinforcement learning from human feedback,

    D. Li, J. Ren, Y . Wang, X. Wen, P. Li, L. Xu, K. Zhan, Z. Xia, P. Jia, X. Langet al., “Finetuning generative trajectory model with reinforcement learning from human feedback,”arXiv preprint arXiv:2503.10434, 2025

  55. [63]

    Confidence-aware reinforcement learning for self-driving cars,

    Z. Cao, S. Xu, H. Peng, D. Yang, and R. Zidek, “Confidence-aware reinforcement learning for self-driving cars,”IEEE Transactions on Intelligent Transportation Systems, vol. 23, no. 7, pp. 7419–7430, 2021

  56. [64]

    Monte carlo motion plan- ning for robot trajectory optimization under uncertainty,

    L. Janson, E. Schmerling, and M. Pavone, “Monte carlo motion plan- ning for robot trajectory optimization under uncertainty,” inRobotics Research: Volume 2. Springer, 2017, pp. 343–361

  57. [65]

    Deep eviden- tial regression,

    A. Amini, W. Schwarting, A. Soleimany, and D. Rus, “Deep eviden- tial regression,”Advances in neural information processing systems, vol. 33, pp. 14 927–14 937, 2020

  58. [66]

    Improving expert predictions with conformal prediction,

    E. Straitouri, L. Wang, N. Okati, and M. G. Rodriguez, “Improving expert predictions with conformal prediction,” inInternational Confer- ence on Machine Learning. PMLR, 2023, pp. 32 633–32 653

  59. [67]

    Forecast- peft: Parameter-efficient fine-tuning for pre-trained motion forecasting models,

    J. Wang, K. Messaoud, Y . Liu, J. Gall, and A. Alahi, “Forecast- peft: Parameter-efficient fine-tuning for pre-trained motion forecasting models,”arXiv preprint arXiv:2407.19564, 2024

  60. [68]

    Motion style transfer: Modular low-rank adaptation for deep motion forecasting,

    P. Kothari, D. Li, Y . Liu, and A. Alahi, “Motion style transfer: Modular low-rank adaptation for deep motion forecasting,” inConference on Robot Learning. PMLR, 2023, pp. 774–784

  61. [69]

    Lord: Adapting differentiable driving policies to distribution shifts,

    C. Diehl, P. Karkus, S. Veer, M. Pavone, and T. Bertram, “Lord: Adapting differentiable driving policies to distribution shifts,” inICRA, 2025

  62. [70]

    Parting with misconceptions about learning-based vehicle motion planning,

    D. Dauner, M. Hallgarten, A. Geiger, and K. Chitta, “Parting with misconceptions about learning-based vehicle motion planning,” in Conference on Robot Learning. PMLR, 2023, pp. 1268–1281

  63. [71]

    A survey on an emerging safety challenge for autonomous vehicles: Safety of the intended functionality,

    H. Wang, W. Shao, C. Sun, K. Yang, D. Cao, and J. Li, “A survey on an emerging safety challenge for autonomous vehicles: Safety of the intended functionality,”Engineering, vol. 33, pp. 17–34, 2024

  64. [72]

    Sotif entropy: Online sotif risk quantification and mitigation for autonomous driving,

    L. Peng, B. Li, W. Yu, K. Yang, W. Shao, and H. Wang, “Sotif entropy: Online sotif risk quantification and mitigation for autonomous driving,” IEEE Transactions on Intelligent Transportation Systems, vol. 25, no. 2, pp. 1530–1546, 2023

  65. [73]

    Safe reinforcement learning in constrained markov decision processes,

    A. Wachi and Y . Sui, “Safe reinforcement learning in constrained markov decision processes,” inInternational Conference on Machine Learning. PMLR, 2020, pp. 9797–9806

  66. [74]

    Deepseekmath: Pushing the limits of mathematical reasoning in open language models,

    Z. Shao, P. Wang, Q. Zhu, R. Xu, J. Song, X. Bi, H. Zhang, M. Zhang, Y . Li, Y . Wuet al., “Deepseekmath: Pushing the limits of mathematical reasoning in open language models,”arXiv preprint arXiv:2402.03300, 2024

  67. [75]

    Hydra-mdp: End-to-end multimodal planning with multi- target hydra-distillation,

    Z. Li, K. Li, S. Wang, S. Lan, Z. Yu, Y . Ji, Z. Li, Z. Zhu, J. Kautz, Z. Wuet al., “Hydra-mdp: End-to-end multimodal planning with multi- target hydra-distillation,”arXiv preprint arXiv:2406.06978, 2024

  68. [76]

    Rehearsal revealed: The limits and merits of revisiting samples in continual learning,

    E. Verwimp, M. De Lange, and T. Tuytelaars, “Rehearsal revealed: The limits and merits of revisiting samples in continual learning,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 9385–9394

  69. [77]

    Posterior network: Uncertainty estimation without ood samples via density-based pseudo- counts,

    B. Charpentier, D. Zügner, and S. Günnemann, “Posterior network: Uncertainty estimation without ood samples via density-based pseudo- counts,”Advances in neural information processing systems, vol. 33, pp. 1356–1367, 2020

  70. [78]

    Coles, J

    S. Coles, J. Bawa, L. Trenner, and P. Dorazio,An introduction to statistical modeling of extreme values. Springer, 2001, vol. 208

  71. [79]

    Statistical inference using extreme order statistics,

    J. Pickands III, “Statistical inference using extreme order statistics,” the Annals of Statistics, pp. 119–131, 1975

  72. [80]

    Parameter and quantile estimation for the generalized pareto distribution,

    J. R. Hosking and J. R. Wallis, “Parameter and quantile estimation for the generalized pareto distribution,”Technometrics, vol. 29, no. 3, pp. 339–349, 1987. 17

  73. [81]

    Navsim leaderboard,

    nuPlan, “Navsim leaderboard,” https://huggingface.co/spaces/ AGC2024-P/e2e-driving-navsim/, 2024

  74. [82]

    Carla: An open urban driving simulator,

    A. Dosovitskiy, G. Ros, F. Codevilla, A. Lopez, and V . Koltun, “Carla: An open urban driving simulator,” inConference on robot learning. PMLR, 2017, pp. 1–16

  75. [83]

    Bench2drive: Towards multi-ability benchmarking of closed-loop end-to-end autonomous driving,

    X. Jia, Z. Yang, Q. Li, Z. Zhang, and J. Yan, “Bench2drive: Towards multi-ability benchmarking of closed-loop end-to-end autonomous driving,”Advances in neural information processing systems, 2024

  76. [84]

    Pre- crash scenario typology for crash avoidance research,

    W. G. Najm, J. D. Smith, M. Yanagisawa, A. Johnet al., “Pre- crash scenario typology for crash avoidance research,” United States. Department of Transportation. National Highway Traffic Safety . . . , Tech. Rep., 2007

  77. [85]

    Trans- fuser: Imitation with transformer-based sensor fusion for autonomous driving,

    K. Chitta, A. Prakash, B. Jaeger, Z. Yu, K. Renz, and A. Geiger, “Trans- fuser: Imitation with transformer-based sensor fusion for autonomous driving,”IEEE Transactions on Pattern Analysis and Machine Intelli- gence, 2022

  78. [86]

    Drama: An efficient end-to-end motion planner for autonomous driving with mamba,

    C. Yuan, Z. Zhang, J. Sun, S. Sun, Z. Huang, C. D. W. Lee, D. Li, Y . Han, A. Wong, K. P. Teeet al., “Drama: An efficient end-to-end motion planner for autonomous driving with mamba,”arXiv preprint arXiv:2408.03601, 2024

  79. [87]

    Goalflow: Goal-driven flow matching for multimodal trajec- tories generation in end-to-end autonomous driving,

    Z. Xing, X. Zhang, Y . Hu, B. Jiang, T. He, Q. Zhang, X. Long, and W. Yin, “Goalflow: Goal-driven flow matching for multimodal trajec- tories generation in end-to-end autonomous driving,”arXiv preprint arXiv:2503.05689, 2025

  80. [88]

    Hydra-next: Robust closed-loop driving with open-loop training,

    Z. Li, S. Wang, S. Lan, Z. Yu, Z. Wu, and J. M. Alvarez, “Hydra-next: Robust closed-loop driving with open-loop training,”arXiv preprint arXiv:2503.12030, 2025

  81. [89]

    Para- drive: Parallelized architecture for real-time autonomous driving,

    X. Weng, B. Ivanovic, Y . Wang, Y . Wang, and M. Pavone, “Para- drive: Parallelized architecture for real-time autonomous driving,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 15 449–15 458

  82. [90]

    Hidden biases of end-to-end driving datasets,

    J. Zimmerlin, J. Beißwenger, B. Jaeger, A. Geiger, and K. Chitta, “Hidden biases of end-to-end driving datasets,”arXiv preprint arXiv:2412.09602, 2024

  83. [91]

    Rllib: Abstractions for distributed rein- forcement learning,

    E. Liang, R. Liaw, R. Nishihara, P. Moritz, R. Fox, K. Goldberg, J. Gon- zalez, M. Jordan, and I. Stoica, “Rllib: Abstractions for distributed rein- forcement learning,” inInternational conference on machine learning. PMLR, 2018, pp. 3053–3062

  84. [92]

    Trajectory-guided control prediction for end-to-end autonomous driving: A simple yet strong baseline,

    P. Wu, X. Jia, L. Chen, J. Yan, H. Li, and Y . Qiao, “Trajectory-guided control prediction for end-to-end autonomous driving: A simple yet strong baseline,”Advances in Neural Information Processing Systems, vol. 35, pp. 6119–6132, 2022

  85. [93]

    Rethinking the open-loop evaluation of end-to-end autonomous driving in nuscenes,

    J.-T. Zhai, Z. Feng, J. Du, Y . Mao, J.-J. Liu, Z. Tan, Y . Zhang, X. Ye, and J. Wang, “Rethinking the open-loop evaluation of end-to-end autonomous driving in nuscenes,”arXiv preprint arXiv:2305.10430, 2023

  86. [94]

    Drivetransformer: Unified trans- former for scalable end-to-end autonomous driving,

    X. Jia, J. You, Z. Zhang, and J. Yan, “Drivetransformer: Unified trans- former for scalable end-to-end autonomous driving,” inInternational Conference on Learning Representations (ICLR), 2025

  87. [95]

    Transdiffuser: End-to-end trajectory generation with decorrelated multi-modal representation for autonomous driving,

    X. Jiang, Y . Ma, P. Li, L. Xu, X. Wen, K. Zhan, Z. Xia, P. Jia, X. Lang, and S. Sun, “Transdiffuser: End-to-end trajectory generation with decorrelated multi-modal representation for autonomous driving,” arXiv preprint arXiv:2505.09315, 2025

  88. [96]

    Carplanner: Consistent auto-regressive trajectory planning for large-scale reinforcement learning in autonomous driving,

    D. Zhang, J. Liang, K. Guo, S. Lu, Q. Wang, R. Xiong, Z. Miao, and Y . Wang, “Carplanner: Consistent auto-regressive trajectory planning for large-scale reinforcement learning in autonomous driving,”arXiv preprint arXiv:2502.19908, 2025

  89. [97]

    Gen-drive: Enhancing diffusion generative driving policies with reward modeling and reinforcement learning fine-tuning,

    Z. Huang, X. Weng, M. Igl, Y . Chen, Y . Cao, B. Ivanovic, M. Pavone, and C. Lv, “Gen-drive: Enhancing diffusion generative driving policies with reward modeling and reinforcement learning fine-tuning,”arXiv preprint arXiv:2410.05582, 2024

  90. [98]

    Lora learns less and forgets less,

    D. Biderman, J. Portes, J. J. G. Ortiz, M. Paul, P. Greengard, C. Jen- nings, D. King, S. Havens, V . Chiley, J. Frankleet al., “Lora learns less and forgets less,”arXiv preprint arXiv:2405.09673, 2024

  91. [99]

    Mtgs: Multi-traversal gaussian splatting,

    T. Li, Y . Qiu, Z. Wu, C. Lindström, P. Su, M. Nießner, and H. Li, “Mtgs: Multi-traversal gaussian splatting,”arXiv preprint arXiv:2503.12552, 2025

  92. [100]

    Decoupled diffusion sparks adaptive scene generation,

    Y . Zhou, N. Ye, W. Ljungbergh, T. Li, J. Yang, Z. Yang, H. Zhu, C. Petersson, and H. Li, “Decoupled diffusion sparks adaptive scene generation,”arXiv preprint arXiv:2504.10485, 2025

  93. [101]

    Welcome to the era of experience,

    D. Silver and R. S. Sutton, “Welcome to the era of experience,”Preprint of a chapter to appear in Designing an Intelligence, edited by George Konidaris, MIT Press (forthcoming), 2025

  94. [102]

    Does reinforcement learning really incentivize reasoning capacity in llms beyond the base model?

    Y . Yue, Z. Chen, R. Lu, A. Zhao, Z. Wang, S. Song, and G. Huang, “Does reinforcement learning really incentivize reasoning capacity in llms beyond the base model?”arXiv preprint arXiv:2504.13837, 2025

  95. [103]

    Pac-bayes & margins,

    J. Langford and J. Shawe-Taylor, “Pac-bayes & margins,”Advances in neural information processing systems, vol. 15, 2002. APPENDIX A. Additional Testing Results We supplement detailed Per-case Testing Results of R2SE refined E2E systems on NA VSIM [81] (nuPlan) and Bench2Drive ...

  96. [2021]

    degree with the School of Mechanical and Aerospace Engineer- ing, Nanyang Technological University, Singapore

    He is currently pursuing the Ph.D. degree with the School of Mechanical and Aerospace Engineer- ing, Nanyang Technological University, Singapore. He won Frist Place with Waymo Open Dataset Challenge 2024. His current research interests in- clude deep learning-enabled motion pr...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.