Pith. sign in

REVIEW 3 major objections 6 minor 29 references

PrismAD: Decoupled Planning via Semantic Mixture-of-Planners for End-to-End Autonomous Driving

T0 review · 3 major / 6 minor · reviewed 2026-07-14 · grok-4.5

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read PrismAD shows that splitting scene tokens into interaction, geometry, and intent experts improves end-to-end driving plans.

desk verdict Solid engineering letter: semantic full-branch planning experts + dual motion/planning routing give consistent gains over SparseDrive/DiffusionDrive; novelty is real but bounded, and the independence premise is only partially stress-tested. read the letter →

arxiv 2607.10336 v1 pith:ZSBVNI62 submitted 2026-07-11 cs.RO cs.CV

Kang Ding , Zhigui Lin , Hongsong Wang , Jie Gui , Qi Liu , Zhe Wang , Luqi Tang , Lei He This is my paper
classification cs.ROcs.CV
keywords end-to-endautonomousdrivingmixtureofexpertssemantictokenpartitionmotionplanningroutingnuScenesclosed-loopsafety
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

End-to-end autonomous driving planners usually dump agent, map, and command tokens into one shared representation and force a single branch to reason about all of them at once. The authors argue this coupling blurs which cues matter for a given decision and weakens specialization. PrismAD instead partitions the tokens into three semantic groups—interaction, geometry, and intent—and gives each group its own complete planning expert with separate parameters. A semantics-aware router then produces separate fusion weights for motion prediction and for ego trajectory planning, activating only the experts that matter for the sample. On open-loop nuScenes, closed-loop NeuroNCAP, and a turning-focused subset, the same design lifts two strong baselines in trajectory accuracy and collision rate. The practical claim is that planning-level expert specialization, not just generic mixture-of-experts layers, yields safer and more interpretable trajectories without large speed loss.

What carries the argument

Semantic Mixture-of-Planners: three architecture-identical but parameter-separate experts (interaction always on; geometry and intent sparsely gated) plus a semantics-aware router that outputs independent top-K weights for motion prediction and for ego planning.

What would settle it

On the same base planners and NeuroNCAP closed-loop protocol, a version that allows free cross-expert attention during representation learning (or a single expert receiving the full mixed token set) matching or beating PrismAD's average score and collision rate would falsify the necessity of the strict semantic partition.

Watch

Extended reading notes

Core claim

PrismAD establishes that explicitly partitioning perception tokens into interaction, geometry, and intent groups, training independent full motion-planning experts on those groups, and fusing their outputs with separate motion versus planning router weights produces better open-loop accuracy, closed-loop safety, and turning robustness than the coupled single-branch baselines it builds on.

Load-bearing premise

The claim rests on the idea that safe plans can be built by keeping interaction, geometry, and intent largely separate inside their own experts, with only late weighted fusion and the interaction expert forced on; if critical decisions need joint multi-factor features inside one branch, the specialization premise fails.

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 / 6 minor

Summary. PrismAD proposes a Semantic Mixture-of-Planners for end-to-end autonomous driving: scene tokens from a sparse perception backbone are hard-partitioned into interaction, geometry, and intent groups (Eqs. 1–2), each concatenated with an ego token and processed by an independent full motion-planning expert with shared architecture but unshared parameters and no cross-expert attention (Eqs. 3–4). A semantics-aware router produces separate top-K routing weights for motion prediction and ego planning (Eqs. 5–10), with noisy gating and a forced-on interaction expert. Training is three-stage (base planner → independent experts → frozen-expert router). The method is instantiated on SparseDrive and DiffusionDrive and evaluated on nuScenes open-loop L2/collision, NeuroNCAP closed-loop score/collision, and Turning-nuScenes, with ablations on expert composition, fusion, and top-K.

Significance. If the reported gains hold under stronger controls, the paper offers a clear, implementable design pattern for planning-level expert specialization in sparse end-to-end AD: full planning branches rather than intermediate MoE layers, explicit semantic token groups, and separate motion vs. planning routing. Improvements over two strong baselines—especially collision reductions on NeuroNCAP and Turning-nuScenes—are practically relevant, and the ablations on gating vs. uniform fusion and top-K efficiency are useful. The work is letter-scale systems research rather than a foundational theoretical result; its main value is the architectural recipe and empirical compatibility with SparseDrive/DiffusionDrive. Planned code release would strengthen reproducibility.

major comments (3)
  1. [Sec. II-A/B, Eqs. 1–4; Table IV; Tables I–III] Central mechanism claim vs. capacity/ensembling: Sec. II-A/B and the introduction argue that hard semantic partition plus independent experts (no cross-expert attention) improves factor-specific reasoning over coupled single-branch planners. Tables I–III only compare against SparseDrive/DiffusionDrive baselines, not against a single planning branch that receives the union of all tokens under matched parameter count, depth, and training budget. Table IV only masks experts from one already-trained MoE checkpoint; it does not retrain a coupled control. Without that control, the reported L2/collision gains (and especially NeuroNCAP/Turning improvements) remain consistent with late fusion of extra capacity rather than proof that independent factor experts retain sufficient multi-factor coupling. A matched-capacity coupled baseline (and ideally a cross-attention multi-expert variant) is load-b
  2. [Sec. II-C, II-D; Table IV] Three-stage freeze and forced interaction expert: Sec. II-D freezes all expert parameters before router training, and Sec. II-C always activates the interaction expert while sparsely selecting geometry/intent. This design isolates routing but prevents experts from learning compensatory joint features and makes the “decoupled yet safe” premise depend on late weighted fusion of a permanently active interaction branch. There is no ablation that (i) jointly fine-tunes experts with the router, or (ii) allows the interaction expert to be gated off. Given that Table IV shows interaction-only already strong on L2 and that safety gains appear mainly when geometry/intent are added at fusion time, these choices need either stronger justification or additional experiments; otherwise the sufficiency of independent factor representations remains under-tested.
  3. [Tables I–III; Sec. III-D] Open-loop gains are small and closed-loop baselines uneven: On nuScenes (Table I), average L2 moves 0.61→0.59 (SparseDrive) and 0.57→0.55 (DiffusionDrive); collision 0.08→0.07 and 0.08→0.04. NeuroNCAP (Table II) shows larger relative gains, but SparseDrive’s reported average score 0.92 / 93.9% collision is very weak, so the 1.83 / 82.0% PrismAD result is a large relative lift from a low base. No multi-seed variance or significance is reported. For a letter whose central claim is consistent improvement via semantic MoP, either stronger open-loop effect sizes under matched controls or explicit multi-run statistics on NeuroNCAP/Turning would better support the claim.
minor comments (6)
  1. [Sec. III-E, Qualitative Results] Qualitative text (near Fig. 5) says “geography and intent experts”; should be “geometry.”
  2. [Fig. 2; Sec. II-A] Fig. 2 t-SNE of ego tokens is presented as evidence of specialization; distinguishable clusters after partition are expected and do not by themselves show that specialization improves planning. Soften the claim or add a quantitative specialization metric (e.g., expert weight vs. scenario type).
  3. [Sec. III-C; Eqs. 6–9] Implementation details (Sec. III-C) omit router MLP sizes, exact top-K used in main tables, noise-scale parameterization of σ(r), and whether SparseDrive and DiffusionDrive PrismAD variants share the same expert architecture depth. These affect reproducibility before code release.
  4. [Table II] Table II marks DiffusionDrive with “* official weights” but PrismAD DiffusionDrive is not similarly clarified regarding whether the base was re-trained under the three-stage schedule; state this explicitly.
  5. [References] Several bibliography entries carry 2025–2026 dates and one “Under review / ICLR 2026” anonymous entry; verify citation status and anonymization consistency for the target venue.
  6. [Abstract; Sec. IV] Abstract and conclusion say “competitive performance”; Tables I–III show clear gains on the chosen baselines but not uniform SOTA across all listed methods (e.g., some VLA/world-model L2 numbers). Prefer precise wording tied to the two base planners.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: empirical architecture paper whose claims rest on external-benchmark gains, not definitional reductions or self-forced predictions.

full rationale

PrismAD is a standard end-to-end systems paper. It hard-partitions scene tokens (Eq. 1–2), trains independent full planning experts without cross-attention (Eq. 3–4), and fuses them via a learned router with separate motion/planning weights and sparse top-K (Eq. 5–10). These are design choices whose value is measured by L2, collision rate, and NeuroNCAP score improvements over SparseDrive/DiffusionDrive on the public nuScenes, Turning-nuScenes, and NeuroNCAP benchmarks. No quantity is algebraically forced to equal its input by construction; no parameter fitted on a subset is renamed a “prediction” of a closely related target; no uniqueness theorem or load-bearing premise is imported solely via overlapping-author citation; and the three-stage training schedule (base planner → independent experts → frozen-expert router) is a methodological stabilization step, not a circular derivation. Ablations (Tables IV–VI) simply mask or re-weight already-trained experts and report external metrics. The paper is therefore self-contained against external data and exhibits zero circular steps under the stated criteria.

Assumptions & free parameters 5 free parameters · 4 assumptions · 3 invented entities

The central empirical claim rests on a modeling decomposition (three semantic factors), staged optimization choices, and several discrete design knobs (always-on interaction expert, top-K, noisy gating, three-stage freeze schedule). No physical constants are fitted; free parameters are architectural/training choices that affect reported gains. Invented entities are architectural modules, not new physical objects.

free parameters (5)
  • top-K expert activation (K∈{1,2,3})
    Inference activation budget chosen by ablation; Top-2/3 materially change L2, collision, and FPS (Table VI).
  • number and identity of semantic experts (interaction, geometry, intent)
    Fixed to three hand-defined groups; composition changes collision rate in Table IV.
  • router training schedule (24 expert epochs + 5 router-only epochs; experts frozen in stage 3)
    Staged freeze schedule is a hand-chosen optimization protocol that isolates router learning; not derived.
  • noisy gating noise scale σ(r) and Gaussian injection
    Training-time noise for exploration; ablation shows collision drop vs plain gating (Table V).
  • learning rate 1e-4, weight decay 1e-3, batch size 24, image size 704×256
    Standard training hyperparameters that affect final metrics; not predicted by theory.
assumptions (4)
  • domain assumption Ego planning is adequately explained by three semantic factors—agent interaction, road geometry, and navigation intent—that can be tokenized separately from perception outputs.
    Stated throughout Introduction and Sec. II-A as the motivation for partition T={T_inter,T_geo,T_intent}.
  • ad hoc to paper Independent experts with no cross-expert attention during representation learning preserve useful specialization without critical loss of cross-factor information.
    Sec. II-B explicitly forbids cross-expert attention to prevent ‘semantic leakage’; this is a design premise, not proven.
  • ad hoc to paper The interaction expert must remain always active; geometry and intent experts may be sparsely selected.
    Sec. II-C states interaction is always activated as the basic branch for temporal agent information.
  • domain assumption nuScenes open-loop L2/collision and NeuroNCAP closed-loop scores are valid proxies for planning quality under the paper’s camera-only, no-ego-status protocol.
    Evaluation protocol Sec. III-A/B following prior E2E planning papers; known open-loop limitations are accepted.
invented entities (3)
  • Semantic Mixture-of-Planners (three full motion-planning expert branches)
    purpose: Specialize complete planning heads on interaction, geometry, and intent token groups rather than MoE layers inside FFNs.
    Core architectural proposal of the letter; evidence is internal ablations and baseline deltas only.
  • Semantics-aware router with separate motion and planning weight heads
    purpose: Produce distinct expert mixture weights for agent motion prediction vs ego planning from ego-centric cross-attention.
    Eq. 5–10; dual routing is paper-specific and only validated inside their experiments.
  • Semantic token partition operator into interaction/geometry/intent groups
    purpose: Convert coupled scene tokens into factor-aligned expert inputs (Eq. 1–2).
    Defined by the authors from detection/map/command tokens; t-SNE (Fig. 2) is qualitative support only.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PrismAD: Decoupled Planning via Semantic Mixture-of-Planners for End-to-End Autonomous Driving." pith.science (2026). https://pith.science/paper/ZSBVNI62

@misc{pith2026260710336,
  author       = {Pith},
  title        = {Pith review of: PrismAD: Decoupled Planning via Semantic Mixture-of-Planners for End-to-End Autonomous Driving},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZSBVNI62}},
  note         = {Machine review of arXiv:2607.10336}
}
abstract

This letter presents PrismAD, a decoupled end-to-end autonomous driving framework based on a Semantic Mixture-of-Planners. Existing planners usually aggregate heterogeneous scene tokens into a coupled representation space, forcing a single planning branch to jointly model agent interaction, road geometry, and driving intention. Such coupling may weaken factor-specific reasoning and obscure the contribution of different planning cues. To address this limitation, PrismAD partitions scene tokens into interaction, geometry, and intent groups, and assigns them to independent planning experts with the same architecture but separate parameters. Each expert learns a specialized motion-planning representation, while a semantics-aware router adaptively aggregates expert predictions with separate routing weights for motion prediction and ego planning. Sparse top-$K$ activation with noisy gating is further introduced to improve routing robustness and reduce unnecessary expert computation. Extensive experiments on the nuScenes open-loop dataset and NeuroNCAP closed-loop benchmark demonstrate that PrismAD exhibits competitive performance. Our code will be released soon.

Figures

Figures reproduced from arXiv: 2607.10336 by the authors.

Figure 1
Figure 1. Comparison of our approach with existing methods. (a) Traditional end-to-end Paradigm. (b) Classical MoE-based Paradigm. (c) Our proposed Paradigm. with surrounding agents, local road geometry, and high-level navigation intention. Most existing end-to-end planners[4], [5], [6], [7] aggregate these heterogeneous scene tokens into a single coupled representation space. Although this unified design simplifies the model… view at source ↗
Figure 2
Figure 2. t-SNE visualization of interaction, geometry, and intent ego tokens [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Architecture of PrismAD. The perception backbone extracts sparse scene tokens from multi-view images, which are partitioned into interaction, [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Qualitative analysis results in the closed-loop evaluation. [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Qualitative analysis results of PrismAD on the nuScenes. [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

29 extracted references · 3 linked inside Pith

  1. [1]

    Planning-oriented autonomous driving,

    Y . Hu, J. Yang, L. Chen, K. Li, C. Sima, X. Zhu, S. Chai, S. Du, T. Lin, W. Wang, L. Lu, X. Jia, Q. Liu, J. Dai, Y . Qiao, and H. Li, “Planning-oriented autonomous driving,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023

    2023

  2. [2]

    Vad: Vectorized scene representa- tion 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 representa- tion for efficient autonomous driving,”2023 IEEE/CVF International Conference on Computer Vision (ICCV), pp. 8306–8316, 2023

    2023

  3. [3]

    Sparsedrive: End-to-end autonomous driving via sparse scene representation,

    W. Sun, X. Lin, Y . Shi, C. Zhang, H. Wu, and S. Zheng, “Sparsedrive: End-to-end autonomous driving via sparse scene representation,”2025 IEEE International Conference on Robotics and Automation (ICRA), pp. 8795–8801, 2024

    2025

  4. [4]

    Genad: Generative end- to-end autonomous driving,

    W. Zheng, R. Song, X. Guo, and L. Chen, “Genad: Generative end- to-end autonomous driving,” inEuropean Conference on Computer Vision, 2024

    2024

  5. [5]

    Impromptu vla: Open weights and open data for driving vision- language-action models,

    H. Chi, H. ang Gao, Z. Liu, J. Liu, C. Liu, J. Li, K. Yang, Y . Yu, Z. Wang, W. Li, L. Wang, X. Hu, H. Sun, H. Zhao, and H. Zhao, “Impromptu vla: Open weights and open data for driving vision- language-action models,”ArXiv, vol. abs/2505.23757, 2025

    arXiv 2025

  6. [6]

    Don’t shake the wheel: Momentum- aware planning in end-to-end autonomous driving,

    Z. Song, C. Jia, L. Liu, H. Pan, Y . Zhang, J. Wang, X. Zhang, S. Xu, L. Yang, and Y . Luo, “Don’t shake the wheel: Momentum- aware planning in end-to-end autonomous driving,”2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 22 432–22 441, 2025

    2025

  7. [7]

    Bridging past and future: End-to-end autonomous driving with historical prediction and plan- ning,

    B. Zhang, N. Song, X. Jin, and L. Zhang, “Bridging past and future: End-to-end autonomous driving with historical prediction and plan- ning,”2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 6854–6863, 2025

    2025

  8. [8]

    Artemis: Autoregressive end-to-end trajectory planning with mixture of experts for autonomous driving,

    R. Feng, N. Xi, D. Chu, R. Wang, Z. Deng, A. Wang, L. Lu, J. Wang, and Y . Huang, “Artemis: Autoregressive end-to-end trajectory planning with mixture of experts for autonomous driving,”IEEE Robotics and Automation Letters, vol. 11, no. 1, pp. 226–233, 2025

    2025

Show all 29 references
  1. [9]

    Expertad: Enhancing autonomous driving systems with mixture of experts,

    H. Jiang, X. Huang, Y . Lu, D. Wang, Y . Cao, C. Sha, B. Chen, K. Chen, and X. Peng, “Expertad: Enhancing autonomous driving systems with mixture of experts,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 40, no. 7, 2026, pp. 5378–5387

    2026

  2. [10]

    Drivemoe: Mixture-of-experts for vision-language-action model in end-to-end autonomous driving,

    Z. Yang, Y . Chai, X. Jia, Q. Li, Y . Shao, X. Zhu, H. Su, and J. Yan, “Drivemoe: Mixture-of-experts for vision-language-action model in end-to-end autonomous driving,”arXiv preprint arXiv:2505.16278, 2025

    2025 arXiv

  3. [11]

    Unimm- v2x: Moe-enhanced multi-level fusion for end-to-end cooperative autonomous driving,

    Z. Song, C. Xia, C. Wang, H. Yu, S. Zhou, and Z. Niu, “Unimm- v2x: Moe-enhanced multi-level fusion for end-to-end cooperative autonomous driving,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 40, no. 11, 2026, pp. 9135–9143

    2026

  4. [12]

    Samoe-vla: A scene adaptive mixture-of-experts vision- language-action model for autonomous driving,

    Z. You, H. Liu, C. Dang, Z. Wang, S. Ang, A. Wang, and Y . Wang, “Samoe-vla: A scene adaptive mixture-of-experts vision- language-action model for autonomous driving,”arXiv preprint arXiv:2603.08113, 2026

    2026

  5. [13]

    Generalizing motion planners with mixture of experts for autonomous driving,

    Q. Sun, H. Wang, J. Zhan, F. Nie, X. Wen, L. Xu, K. Zhan, P. Jia, X. Lang, and H. Zhao, “Generalizing motion planners with mixture of experts for autonomous driving,” in2025 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2025, pp. 6033–6039

    2025

  6. [14]

    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. Zhang, and X. Wang, “Diffusiondrive: Truncated diffusion model for end-to-end autonomous driving,” pp. 12 037– 12 047, 2025

    2025

  7. [15]

    nuscenes: A multimodal dataset for autonomous driving,

    H. Caesar, V . Bankiti, A. H. Lang, S. V ora, V . E. Liong, Q. Xu, A. Krishnan, Y . Pan, G. Baldan, and O. Beijbom, “nuscenes: A multimodal dataset for autonomous driving,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 11 621–11 631

    2020

  8. [16]

    Neuroncap: Photorealistic closed-loop safety testing for autonomous driving,

    W. Ljungbergh, A. Tonderski, J. Johnander, H. Caesar, K. ˚Astr¨om, M. Felsberg, and C. Petersson, “Neuroncap: Photorealistic closed-loop safety testing for autonomous driving,” inEuropean Conference on Computer Vision. Springer, 2024, pp. 161–177

    2024

  9. [17]

    Is ego status all you need for open-loop end-to-end autonomous driving?

    Z. Li, Z. Yu, S. Lan, J. Li, J. Kautz, T. Lu, and J. M. ´Alvarez, “Is ego status all you need for open-loop end-to-end autonomous driving?”2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 14 864–14 873, 2023

    2024

  10. [18]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,”2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 770–778, 2015

    2016

  11. [19]

    Decoupled weight decay regularization,

    I. Loshchilov and F. Hutter, “Decoupled weight decay regularization,” inInternational Conference on Learning Representations, 2017

    2017

  12. [20]

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

    Y .-Q. 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,”2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 14 749–14 759, 2023

    2024

  13. [21]

    Enhancing end-to-end autonomous driving with latent world model,

    Y . Li, L. Fan, J. He, Y . Wang, Y . Chen, Z. Zhang, and T. Tan, “Enhancing end-to-end autonomous driving with latent world model,” arXiv preprint arXiv:2406.08481, 2024

    2024 arXiv

  14. [22]

    Occworld: Learning a 3d occupancy world model for autonomous driving,

    W. Zheng, W. Chen, Y . Huang, B. Zhang, Y . Duan, and J. Lu, “Occworld: Learning a 3d occupancy world model for autonomous driving,” inEuropean conference on computer vision. Springer, 2024, pp. 55–72

    2024

  15. [23]

    Orion: A holistic end-to-end au- tonomous driving framework by vision-language instructed action generation,

    H. Fu, D. Zhang, Z. Zhao, J. Cui, D. Liang, C. Zhang, D. Zhang, H. Xie, B. Wang, and X. Bai, “Orion: A holistic end-to-end au- tonomous driving framework by vision-language instructed action generation,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2025

    2025

  16. [24]

    Omnidrive: A holistic vision-language dataset for autonomous driving with counterfactual reasoning,

    S. Wang, Z. Yu, X. Jiang, S. Lan, M. Shi, N. Chang, J. Kautz, Y . Li, and J. M. Alvarez, “Omnidrive: A holistic vision-language dataset for autonomous driving with counterfactual reasoning,” inProceedings of the computer vision and pattern recognition conference, 2025, pp. 22 ...

    2025

  17. [25]

    Unidrivevla: Unifying understanding, perception, and action planning for autonomous driving,

    Y . Li, L. Zhou, S. Yan, B. Liao, T. Yan, K. Xiong, L. Chen, H.- B. Xie, B. Wang, G. Chen, H. Ye, W. Liu, H. Sun, and X. Wang, “Unidrivevla: Unifying understanding, perception, and action planning for autonomous driving,” 2026

    2026

  18. [26]

    Uncad: Towards safe end-to-end autonomous driving via online map uncertainty,

    P. Yang, Y . Zheng, Q. Zhang, K. Zhu, Z. Xing, Q. Lin, Y .-F. Liu, Z. Su, and D. Zhao, “Uncad: Towards safe end-to-end autonomous driving via online map uncertainty,”2025 IEEE International Conference on Robotics and Automation (ICRA), pp. 6408–6415, 2025

    2025

  19. [27]

    Causalvad: De-confounding end-to-end autonomous driving via causal interven- tion,

    J. Tang, Z. Zhou, Z. He, J. Zhang, K. Zhang, and J. Pu, “Causalvad: De-confounding end-to-end autonomous driving via causal interven- tion,” 2026

    2026

  20. [28]

    Flowad: Ego-scene interactive modeling for autonomous driving,

    Anonymous, “Flowad: Ego-scene interactive modeling for autonomous driving,” inUnder review as a conference paper at ICLR 2026, 2026

    2026

  21. [29]

    Reasoning-vla: A fast and general vision- language-action reasoning model for autonomous driving,

    D. Zhang, Z. Yuan, Z. Chen, C.-T. Liao, Y . Chen, F. Shen, Q. Zhou, and T.-S. Chua, “Reasoning-vla: A fast and general vision- language-action reasoning model for autonomous driving,”ArXiv, vol. abs/2511.19912, 2025

    2025

Pith tools

Reviewed July 14, 2026 · model on record in the stance chip above.