Pith. sign in

REVIEW 4 major objections 5 minor 4 cited by

LD-Scene: LLM-Guided Diffusion for Controllable Generation of Adversarial Safety-Critical Driving Scenarios

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read By letting a large language model write the attack objective as code, LD-Scene guides a latent diffusion model to generate realistic, natural-language-controllable crash scenarios, reaching a 40.75% adversarial-ego collision rate in…

desk verdict LD-Scene is a useful integration of LLM-generated guidance with latent diffusion, but the headline SOTA claim rests on a best-of-10 selection protocol that baselines did not receive, so the margin is not yet trustworthy. read the letter →

arxiv 2505.11247 v2 pith:HER34L67 submitted 2025-05-16 cs.AI cs.LGcs.RO

Mingxing Peng , Yuting Xie , Xusen Guo , Ruoyu Yao , Hai Yang , Jun Ma This is my paper · ORCID
classification cs.AIcs.LGcs.RO
keywords largelanguagemodelsdiffusiontrafficsimulationsafety-criticalscenariogenerationautonomousdrivingadversarialchain-of-thoughtlatent
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

LD-Scene claims that a user who can describe a dangerous driving event in plain language should be able to obtain a realistic, closed-loop traffic simulation in which a chosen background vehicle attacks the ego vehicle in exactly that way. The paper's central result is that pairing a latent diffusion model, which has learned realistic driving behavior from the nuScenes dataset, with an LLM that writes the adversarial objective as Python code lifts the adversarial-ego collision rate to 40.75%, well above the 15-28% range achieved by prior generators. This matters because safety-critical scenarios are rare in real driving data, and current generators either require expert-designed objectives or retrained classifiers for each new attack type. The paper also shows that the user can tune the attack's intensity and specify the maneuver, such as a high-speed overtake or a sharp turn. If the claim holds, LD-Scene turns adversarial scenario generation into a natural-language prompting task accessible to non-experts.

What carries the argument

The load-bearing object is the LLM guidance module: a zero-shot Chain-of-Thought prompt that decomposes a user query into three steps — classify the adversarial level (weak, medium, or strong), assign loss weights from predefined ranges, and complete a code template that produces a guidance loss class — followed by an LLM-based debugger that runs unit tests on the generated code and iteratively repairs it using the traceback. This guidance loss is injected into a latent diffusion model at inference time by decoding the predicted clean latent into trajectories, computing the gradient of the objective with respect to the latent, and shifting the denoising update; the gradient is applied separately to the adversarial vehicle's latent and to the non-adversarial vehicles' latents, with separate realism losses keeping background agents plausible. A pretrained graph-based VAE provides the latent space and the autoregressive kinematic decoder, while DDIM accelerates sampling. The code debugger is what makes the pipeline reliable: it lifts the success rate of generated guidance functions from 69.4% to 95.0% for the model used in the main experiments.

What would settle it

Run LD-Scene with a single candidate trajectory per non-ego agent, removing the best-of-10 guidance-loss-plus-feasibility selection, and measure the Adv-Ego collision rate on the nuScenes validation split; if the rate falls toward the 15-28% band of the baselines, the reported advantage comes from the selection protocol rather than the learned guidance. Conversely, grant the same best-of-10 selection to AdvSim, Strive, DiffScene, and Safe-Sim and check whether the 40.75% gap persists.

Watch

Extended reading notes

Core claim

LD-Scene's central claim is that an LLM-based guidance generation module, consisting of a Chain-of-Thought code generator and an iterative code debugger, can translate a natural-language user query into a differentiable adversarial loss function, and that this loss can steer the denoising process of a latent diffusion model to produce adversarial safety-critical scenarios that are both realistic and precisely controllable. Evaluated in closed-loop simulation with a rule-based planner on the nuScenes validation split, the method achieves a 40.75% adversarial-ego collision rate, a 12.52% adversarial offroad rate, and a 229.40-second average simulation time, which the paper reports as state-of-the-art relative to AdvSim, Strive, DiffScene, and Safe-Sim. The paper further claims that generation is controllable along two axes: the adversarial level, inferred from intensity descriptors in the query and mapped to loss-weight ranges, and the adversarial behavior, determined by the structure of the generated loss function.

Load-bearing premise

The headline 40.75% collision rate rests on the inference protocol described in Section 4.3: ten candidate trajectories per agent are generated and the one minimizing guidance loss while passing a physical-feasibility filter is kept, and the baselines are not given the same selection advantage.

Editorial extensions

If this is right

  • A user can specify an adversarial event in plain language and receive a closed-loop simulation in which the specified behavior occurs, without writing any objective function or retraining a classifier.
  • The adversarial intensity of the generated scenarios scales with the query wording: weak queries yield longer time-to-collision and lower accelerations, while strong queries produce shorter time-to-collision and higher accelerations.
  • Because the adversarial vehicle remains fixed during the scenario and the ego planner reacts through a closed loop, generated scenarios can expose planner vulnerabilities such as yield-versus-pass decisions at junctions and ramps.
  • The debugger reduces dependence on the specific LLM: all evaluated models improve their guidance-code success rates, making the framework's reliability less sensitive to LLM choice.

Reading between the lines

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

  • The best-of-10 candidate selection described in the inference protocol means the reported 40.75% collision rate measures the best of ten guided samples, not the average sample; a single-sample evaluation would likely report a lower rate, and giving the same selection protocol to the baselines is the cleanest way to test whether the advantage is intrinsic to the guidance.
  • The guidance template could be extended beyond collision events to other safety-critical conditions such as near-misses with specified margins, traffic-rule violations that do not involve contact, or sensor-level stressors, since the LLM only needs to generate loss code within the provided template.
  • The framework's cost scales with LLM token usage for code generation and debugging; as cheaper code-capable models improve, the per-scenario cost should fall without changing the diffusion backbone.
  • A stricter test of controllability would measure whether the requested collision type or maneuver is actually achieved in the closed-loop rollout, not just whether the generated loss function contains the corresponding code structure.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes LD-Scene, a framework that combines a latent diffusion model for multi-agent trajectory generation with an LLM-based guidance module. The LLM translates a natural-language user query into an adversarial guidance loss function through chain-of-thought code generation and an automated code debugger. The guidance loss is used to perturb the denoising process so that a designated adversarial vehicle collides with an ego vehicle controlled by a rule-based planner. The authors evaluate on the nuScenes dataset in closed-loop simulation, reporting that LD-Scene achieves a 40.75% Adv-Ego collision rate, outperforming AdvSim, Strive, DiffScene, and Safe-Sim (Table 1). They also report ablation studies on the guidance components and the debugger, a controllability study of adversarial level and behavior, and an analysis of key parameters. The central claim is that LD-Scene achieves state-of-the-art controllable generation of realistic, diverse, and effective adversarial safety-critical driving scenarios.

Significance. If the claims are substantiated, the paper makes a useful contribution: it offers a user-friendly, natural-language interface for adversarial scenario generation, which is currently a gap in the diffusion-based trajectory generation literature. The combination of LLM-generated guidance code with a latent diffusion model is a sensible and timely integration, and the closed-loop evaluation against several representatives (optimization-based and diffusion-based) is relevant. The debugger module with quantitative success-rate improvements is a practical strength. However, the headline performance claim rests on an evaluation protocol that is not yet convincingly fair, so the significance of the empirical advance is currently uncertain. The paper's underlying idea is sound and the empirical claims are in principle testable, which makes the weaknesses fixable rather than fatal.

major comments (4)
  1. [§4.3, Table 1] The headline Adv-Ego Collision Rate of 40.75% is obtained after a best-of-10 selection step: the paper states that the final trajectory is selected as the one that minimizes the guidance loss while satisfying physical feasibility constraints. Since the guidance loss is explicitly designed to encourage collision between the adversarial and ego vehicle, this selection protocol mechanically raises the measured collision rate even if the underlying conditional distribution is no more adversarial than that of the baselines. The three diffusion baselines (DiffScene, Safe-Sim) are not reported to receive the same selection protocol, and the optimization-based baselines (AdvSim, Strive) follow a different search procedure. The reported margin over Safe-Sim (40.75% vs 27.81%) may therefore reflect protocol asymmetry rather than model quality. To substantiate the SOTA claim, the authors should either apply the same best-of-10 selection to the diffusion baselines, or report the Adv-Ego collision rate without selection, or both; they should also report the mean (not best) collision rate over the 10 candidates.
  2. [§4.3, Eq. (6)-(7)] The 'physical feasibility constraints' from Peng et al. (2025) are load-bearing for the selection procedure, but they are not defined anywhere in this manuscript. The reader cannot tell how many of the 10 candidate trajectories are rejected, how the filter interacts with the guidance-loss selection, or whether the same filter is applied to the baselines. Given that the feasibility filter is self-cited to the authors' own prior work, this is a reproducibility gap that directly affects the validity of Table 1. The constraints should be stated explicitly in the paper (or in an appendix), and the authors should report the rejection rate of the candidate trajectories.
  3. [§4.4, Table 1] All metrics in Table 1 are reported as single point estimates without error bars, confidence intervals, or significance tests, and the number of evaluated scenarios is not stated. Some gaps between methods are small (e.g., Other Offroad is 17.95% for LD-Scene vs 16.64% for Strive), so the claimed 'significant advantages' are not supported by the reported data. The authors should report the number of scenes, the standard deviation or confidence intervals across scenes or random seeds, and perform pairwise significance tests (e.g., bootstrap or paired tests) for at least the headline Adv-Ego Collision Rate.
  4. [§4.5.2, Fig. 5(a), §4.3] Fig. 5(a) reports that with the debugger, GPT-4o achieves a 95.0% success rate in generating executable guidance functions, meaning 5% of user queries fail to produce valid code. The paper does not state whether scenarios generated from failed code are included in the evaluation of Table 1, or whether the 40.75% collision rate is conditioned on successful code generation. If failures are excluded, the evaluation is only over successful queries, which biases the collision rate upward and differs from the baselines, which use predefined guidance functions with no generation failure. The authors should either include the failures in the denominator or clearly condition the reported metrics on successful code generation and discuss the resulting bias.
minor comments (5)
  1. [§3.4, Eq. (5)] In Eq. (5), the notation Σg is used without defining the covariance matrix or how the gradient g is scaled; the relation between Σ and the diffusions steps is unclear. Please define all symbols.
  2. [§4.7, Fig. 8 caption] The text under Fig. 8 refers to 'Adv Offroad collision rate', which conflates an offroad metric with a collision rate; the y-axis label should be simply 'Adv Offroad Rate'.
  3. [Fig. 2(c)] There is a typo in the Reasoning Prompt: 'guidance loss funtion class' should be 'guidance loss function class'.
  4. [§4.2] The AdvAcceleration metric is defined only as 'acceleration magnitude' without specifying whether it is peak, mean, or per-scenario averaged; please clarify the exact computation.
  5. [§4.3] The adversarial vehicle selection strategy is described as 'closest to the ego vehicle in the initial state' and is said to follow Strive; however, Strive changes the adversarial vehicle dynamically, while the paper says it is fixed. Please clarify the divergence from Strive and its potential effect on the comparison.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: LD-Scene's adversarial guidance loss directly optimizes collision, which is the intended task, and the reported Adv-Ego collision rate is an independent closed-loop outcome, not a restatement of the training objective or a fitted parameter.

full rationale

The paper's central derivation—training an LDM on nuScenes trajectory distributions and using an LLM-generated collision-inducing guidance loss to steer denoising—is self-contained and externally grounded. The guidance loss is explicitly defined from vehicle collision penalties (Eq. 6) and the evaluation metric Adv-Ego Collision Rate is measured in closed-loop simulation with a rule-based planner; it is not equal by construction to the guidance loss, since collision is a downstream outcome rather than the loss value itself. The inference-time selection of the candidate trajectory that minimizes the guidance loss is an optimization/search protocol, not a hidden definition or a fitted parameter renamed as a prediction; it may affect the magnitude of the reported rate and fairness relative to baselines, but that is a protocol concern, not circularity. The citations to the authors' prior work—Peng et al. (2025) for physical feasibility constraints and Strive for the pretrained VAE architecture—are implementation references; they are not load-bearing uniqueness theorems, and the central claim does not reduce to a self-citation chain. No equation in the paper reduces to its own input, no fitted value is renamed as a prediction, and no ansatz is smuggled in solely through citation. Therefore, no circular step is found.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central claim rests primarily on the frozen Strive VAE latent space, the rule-based planner, the unstated feasibility constraints, and hand-set weight ranges for adversarial levels. These are the main imported assumptions; no new physical or conceptual entities are posited.

free parameters (4)
  • Number of candidate trajectories per scene = 10
    Chosen as a trade-off between performance and runtime (Fig. 8a); collision rate improves slightly with more samples.
  • Number of diffusion steps = 20
    Chosen because more denoising steps degrade adversariality and realism and increase time (Fig. 8b).
  • Adversarial level loss-weight ranges = Weak [0,1], Medium [1,10], Strong [10,100]
    Hand-defined in the Reasoning Prompt (Fig. 2c); the controllability result (Table 3) directly depends on these ranges, which are not derived from data.
  • Safety buffer d_buffer and map threshold p_i = Not reported
    The collision penalty (Eq. 6) and map penalty (Eq. 7) depend on these constants, but their values are not given; they affect the realism losses.
assumptions (5)
  • domain assumption The pretrained Strive VAE (frozen encoders and decoder) provides a latent space in which gradient guidance on the denoised latent produces physically plausible trajectories.
    Section 3.2 states the encoders and decoder are directly adopted from Strive (Rempe et al., 2022); the entire guided generation operates in this latent space and no fine-tuning or validation of the latent space is reported.
  • domain assumption The rule-based lane-graph planner (Montemerlo et al., 2008) is a representative ego planner for evaluating adversarial scenarios.
    Section 4.3 describes the closed-loop setup; results may not transfer to learned planners.
  • ad hoc to paper The 'physical feasibility constraints' from the authors' prior work (Peng et al., 2025) are well-defined and filter unrealistic trajectories without biasing the comparison.
    Section 4.3 references these constraints without defining them; the final trajectory selection depends on them.
  • domain assumption GPT-4o's generated code, after the debugger loop, implements the user's intent with sufficient reliability.
    Section 3.3 relies on the LLM for both loss function generation and debugging; the success rate is 95% after debugging, leaving a residual failure rate.
  • standard math Standard diffusion model and DDIM sampling equations (Ho et al., 2020; Song et al., 2020) are correct.
    Eqs. (1)-(4) use standard DDPM/DDIM formulations without modification.

how reviews work

0 comments
Cite this review

Pith. "Pith review of LD-Scene: LLM-Guided Diffusion for Controllable Generation of Adversarial Safety-Critical Driving Scenarios." pith.science (2026). https://pith.science/paper/HER34L67

@misc{pith2026250511247,
  author       = {Pith},
  title        = {Pith review of: LD-Scene: LLM-Guided Diffusion for Controllable Generation of Adversarial Safety-Critical Driving Scenarios},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HER34L67}},
  note         = {Machine review of arXiv:2505.11247}
}
read the original abstract

Ensuring the safety and robustness of autonomous driving systems necessitates a comprehensive evaluation in safety-critical scenarios. However, these safety-critical scenarios are rare and difficult to collect from real-world driving data, posing significant challenges to effectively assessing the performance of autonomous vehicles. Typical existing methods often suffer from limited controllability and lack user-friendliness, as extensive expert knowledge is essentially required. To address these challenges, we propose LD-Scene, a novel framework that integrates Large Language Models (LLMs) with Latent Diffusion Models (LDMs) for user-controllable adversarial scenario generation through natural language. Our approach comprises an LDM that captures realistic driving trajectory distributions and an LLM-based guidance module that translates user queries into adversarial loss functions, facilitating the generation of scenarios aligned with user queries. The guidance module integrates an LLM-based Chain-of-Thought (CoT) code generator and an LLM-based code debugger, enhancing the controllability and robustness in generating guidance functions. Extensive experiments conducted on the nuScenes dataset demonstrate that LD-Scene achieves state-of-the-art performance in generating realistic, diverse, and effective adversarial scenarios. Furthermore, our framework provides fine-grained control over adversarial behaviors, thereby facilitating more effective testing tailored to specific driving scenarios.

Figures

Figures reproduced from arXiv: 2505.11247 by the authors.

Figure 1
Figure 1. Overall framework of LD-Scene. During the training stage, an LDM learns the distribution of realistic driving trajectories conditioned on the latent representation of historical scene input. During the inference stage, given a user query, an LLM-based code generator produces an adversarial loss function. This loss function is then validated by an LLM-based debugger through a closed-loop unit testing process and subs… view at source ↗
Figure 2
Figure 2. Prompts in the guidance generation module. (a) System Prompt: it specifies the task, inputs, objectives, and important instructions for generating the guidance loss function. (b) Code Generation Prompt: it provides a structured template for generating a guidance loss, incorporating predefined loss functions and some code generation tips. (c) Reasoning Prompt: it outlines a step-by-step reasoning process to determine… view at source ↗
Figure 3
Figure 3. Visualizations of overall performance among scenario generation models. (a)–(d): Distributions of longitudi￾nal/lateral speed and acceleration. (e): Radar chart comparing adversariality, behavior plausibility, and generation efficiency. LD-Scene achieves the best overall balance. LD-Scene achieves the strongest adversarial effectiveness while maintaining high levels of realism and generation efficiency, thus demonst… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Adversarial safety-critical scenarios generated by LD-Scene. Examples include cut-in maneuvers, junction interactions, U-turns, counterflow driving, ramp merges, and roundabout conflicts, demonstrating our model’s capability to generate diverse and realistic safety-cri…
Figure 5
Figure 5. Figure 5: Ablation study on the effectiveness of the debugger module. (a): Success rates of guidance code execution with and without the debugger. The debugger significantly improves success rates across all LLMs. (b): Average token input and output statistics per model. (c): To…
Figure 6
Figure 6. Figure 6: Visualization of results under different adversarial levels. and lower medians from weak to strong levels. Meanwhile, [PITH_FULL_IMAGE:figures/full_fig_p014_6.png]
Figure 7
Figure 7. Figure 7: Case studies on adversarial safety-critical scenario generation based on different user queries, including normal collisions, high-speed overtaking collisions, and sharp-turn collisions. The results demonstrate the capability of our LD-Scene for controllable adversaria…
Figure 8
Figure 8. Figure 8: Quantitative results of key parameters on adversarial scenario generation, evaluated in terms of adversariality (Adv-Ego collision rate), realism (Adv Offroad collision rate), and efficiency (time consumption). 4.7. Analysis of the Impact of Key Parameters We analyze t…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 4 Pith papers

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

  1. CCFM: Collision-Constrained Flow Matching for Safety-Critical Scenario Generation

    cs.CV 2026-07 accept novelty 6.0 of 10

    Hard-constrained flow matching with a heuristic selector and Gauss-Newton projection produces type-controllable collisions at 46–83 % rates for AV safety evaluation.

  2. SPHINX: First Explain, Then Explore

    cs.CV 2026-06 unverdicted novelty 6.0 of 10

    SPHINX generates adversarial driving scenarios by first diagnosing policy failures with XAI and then using VLMs to create targeted critics and scenarios, yielding robustness gains over prior methods on AV benchmarks.

  3. Generative AI for Testing of Autonomous Driving Systems: A Survey

    cs.SE 2025-08 conditional novelty 5.0 of 10

    A systematic survey that organizes 91 studies of generative AI for autonomous driving testing into six scenario-based tasks and catalogs 27 limitations.

  4. A Survey on the Applications of Generative Artificial Intelligence in Automated Driving Systems Test Scenario Generation Methods

    cs.SE 2025-12 reject novelty 4.0 of 10

    A literature survey of scenario-generation methods for ADS testing that adds an unvalidated AII/RAS/OCS metric suite and ODD-difficulty schema, undermined by inconsistent calculations in the worked examples.

Reference graph

Works this paper leans on

15 extracted references · 4 canonical work pages · cited by 4 Pith papers

  1. [4]

    IEEE Robotics and Automation Letters 6, 1551–1558

    Multimodal safety-critical scenarios generation for decision-making algorithms evaluation. IEEE Robotics and Automation Letters 6, 1551–1558. Ding, W., Chen, B., Xu, M., Zhao, D., a. Learning to Collide: An adaptive safety-critical scenarios generating method, in: the Proceedings of IEEE/RSJ International Conference on Intelligent Robots and Systems, pp. ...

    work page 2023

  2. [9]

    10684–10695

    High-resolution image synthesis with latent diffusion models, in: the Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 10684–10695. Scanlon,J.M.,Kusano,K.D.,Daniel,T.,Alderson,C.,Ogle,A.,Victor,T.,2021. Waymosimulateddrivingbehaviorinreconstructedfatalcrashes within an autonomous vehicle operating domain. Accident Ana...

    work page 2021

  3. [10]

    IEEE Transactions on Neural Networks 20, 61–80

    The graph neural network model. IEEE Transactions on Neural Networks 20, 61–80. Schulman,J.,Wolski,F.,Dhariwal,P.,Radford,A.,Klimov,O.,2017. Proximalpolicyoptimizationalgorithms. arXivpreprintarXiv:1707.06347. Song, J., Meng, C., Ermon, S.,

    arXiv 2017

  4. [11]

    arXiv preprint arXiv:2010.02502

    Denoising diffusion implicit models. arXiv preprint arXiv:2010.02502 . Suo, S., Wong, K., Xu, J., Tu, J., Cui, A., Casas, S., Urtasun, R.,

    arXiv 2010

  5. [13]

    arXiv preprint arXiv:2309.16292

    DILU: A knowledge-driven approach to autonomous driving with large language models. arXiv preprint arXiv:2309.16292 . Xia, J., Xu, C., Xu, Q., Wang, Y., Chen, S.,

    arXiv

  6. [14]

    9983–9989

    AdvDiffuser: Generating adversarial safety-critical driving scenarios via guided diffusion, in: the Proceedings of IEEE/RSJ International Conference on Intelligent Robots and Systems, pp. 9983–9989. Xu,C.,Zhao,D.,Sangiovanni-Vincentelli,A.,Li,B.,2023. DiffScene:Diffusion-basedsafety-criticalscenariogenerationforautonomousvehicles, in: The Second Workshop ...

    arXiv 2023

  7. [15]

    22490–22499

    LayoutDiffusion: Controllable diffusion model for layout-to-image generation, in: the Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 22490–22499. Zheng,L.,Yang,R.,YuWang,M.,Ma,J.,2025. Barrier-enhancedparallelhomotopictrajectoryoptimizationforsafety-criticalautonomousdriving. IEEE Transactions on Intelligent Transpo...

    work page 2025

  8. [2008]

    Journal of Field Robotics 25, 569–597

    Junior: The stanford entry in the urban challenge. Journal of Field Robotics 25, 569–597. Peng, M., Chen, K., Guo, X., Zhang, Q., Lu, H., Zhong, H., Chen, D., Zhu, M., Yang, H., 2024a. Diffusion models for intelligent transportation systems: A survey. arXiv preprint arXiv:2409.15816 . Peng,M.,Guo,X.,Chen,X.,Zhu,M.,Chen,K.,2024b. LC-LLM:Explainablelane-cha...

    arXiv

Show all 15 references
  1. [2019]

    8271–8277

    Generating adversarial driving scenarios in high-fidelity simulators, in: the Proceedings of IEEE International Conference on Robotics and Automation, pp. 8271–8277. Achiam,J.,Adler,S.,Agarwal,S.,Ahmad,L.,Akkaya,I.,Aleman,F.L.,Almeida,D.,Altenschmidt,J.,Altman,S.,Anadkat,S.,et...

    2023 arXiv

  2. [2020]

    11621–11631

    nuScenes: A multimodal dataset for autonomous driving, in: the Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 11621–11631. Chang,W.J.,Pittaluga,F.,Tomizuka,M.,Zhan,W.,Chandraker,M.,2024. SAFE-SIM:Safety-criticalclosed-looptrafficsimulati...

    2024

  3. [2021]

    5119–5125

    SimNet: Learning reactive self-driving simulations from real-world observations, in: the Proceedings of IEEE International Conference on Robotics and Automation, pp. 5119–5125. Bubeck,S.,Chadrasekaran,V.,Eldan,R.,Gehrke,J.,Horvitz,E.,Kamar,E.,Lee,P.,Lee,Y.T.,Li,Y.,Lundberg,S.,...

    2023 arXiv

  4. [2022]

    arXiv preprint arXiv:2205.09991

    Planning with diffusion for flexible behavior synthesis. arXiv preprint arXiv:2205.09991 . Jiang, C., Cornman, A., Park, C., Sapp, B., Zhou, Y., Anguelov, D., et al.,

    arXiv

  5. [2023]

    arXiv preprint arXiv:2307.07947

    Language conditioned traffic generation. arXiv preprint arXiv:2307.07947 . Wang, J., Pun, A., Tu, J., Manivasagam, S., Sadat, A., Casas, S., Ren, M., Urtasun, R.,

    arXiv

  6. [2024]

    arXiv preprint arXiv:2401.14196

    DeepSeek-Coder: When the large language model meets programming–the rise of code intelligence. arXiv preprint arXiv:2401.14196 . Guo, X., Yang, X., Peng, M., Lu, H., Zhu, M., Yang, H.,

    arXiv

  7. [2025]

    arXiv preprint arXiv:2505.00515

    Safety-critical traffic simulation with guided latent diffusion model. arXiv preprint arXiv:2505.00515 . Rempe,D.,Luo,Z.,BinPeng,X.,Yuan,Y.,Kitani,K.,Kreis,K.,Fidler,S.,Litany,O.,2023. TraceandPace:Controllablepedestriananimationvia guided trajectory diffusion, in: the Proceed...

    2023 arXiv

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.