Pith. sign in

REVIEW 3 major objections 6 minor 28 references

Chat2Scenic turns regulatory text into executable driving scenarios at a 76.42% compilation success rate, roughly 2.5x the best prior method, by generating scripts component by component with retrieval-augmented prompting.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-02 02:12 UTC pith:QIF6GVWF

load-bearing objection Useful framework with a genuinely novel iterative component-wise RAG design, but the headline CSR/FA are best-on-benchmark numbers and a data-leakage check is needed before trusting them. the 3 major comments →

arxiv 2607.14387 v1 pith:QIF6GVWF submitted 2026-07-15 cs.AI cs.RO

Chat2Scenic: An Iterative RAG-Based Framework for Scenario Generation in Autonomous Driving

Yuan Gao , Wenting Miao , Mattia Piccinini , Haoyu Wang , Qunying Song , Johannes Betz This is my paper
classification cs.AI cs.RO
keywords autonomous drivingscenario generationretrieval-augmented generationlarge language modelsSceniccomponent-wise generationcompilation success rateregulatory test scenarios
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper proposes an iterative, retrieval-augmented framework, Chat2Scenic, to generate executable driving-scenario scripts from natural-language regulatory descriptions. The central claim is that breaking each scenario into logical components (global settings, spatial relations, ego behavior, objects, restrictions) and generating them one at a time, with retrieval of code snippets and documentation at each step, makes the generation reliable enough for simulation-based testing. On a new benchmark of 123 scenarios drawn from U.S. and international vehicle-safety regulations and public leaderboard tasks, the best configuration achieves a 76.42% compilation success rate and 58.17% framework accuracy, compared with 30.08% and 16.26% CSR for prior retrieval-assemble and full-script generation methods. If this result is robust, it would let engineers turn regulatory text into simulation-ready tests with far less manual authoring.

Core claim

The paper's central discovery, on its own terms, is that component-wise iterative generation, combined with retrieval-augmented in-context learning, resolves the trade-off between compilation reliability and generalizability in DSL-based scenario generation. Rather than assembling pre-existing snippets (which compiles but does not generalize) or generating a full script at once (which generalizes but rarely compiles), the framework parses a query into a logical structure, generates each component in dependency order, and feeds the accumulated code back as context for later components. A dual retriever supplies semantically similar code snippets and documentation chunks to ground each generat

What carries the argument

The load-bearing mechanism is a logical structure schema S = {G, R, E, O, T} that decomposes any scenario description into global configuration (map, weather, vehicles) and scenario components (spatial relations, ego behavior, objects, restrictions). The generation module produces code component by component in a fixed dependency order, passing each generated component's code into the prompt for the next one, so compatibility is maintained incrementally. A dual-retriever RAG module fetches the most relevant code snippets (by semantic similarity) and documentation chunks (by hybrid BM25 plus embedding search fused with Reciprocal Rank Fusion) for each component description. These retrieved co

Load-bearing premise

The reported 76.42% compilation success and 58.17% framework accuracy come from the single best configuration (C11) chosen among 12 ablations on the same 123-scenario benchmark it is then evaluated against; the load-bearing premise is that this selection does not materially overfit the benchmark and that the human layer-scoring is a stable measure of quality.

What would settle it

Run the final C11 configuration and the same human scoring protocol on a fresh set of regulatory descriptions that were not part of the 123-scenario benchmark. If the compilation success rate falls substantially below 76.4% (for example, below 50%), the headline numbers are at least partly artifacts of benchmark-driven configuration selection.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Regulatory descriptions that previously required manual translation into simulation code can be converted to executable scenarios with a majority success rate, reducing the cost of building diverse test suites.
  • The component-wise structure means a single failed component can be regenerated or corrected without discarding the whole scenario, easing debugging and human-in-the-loop refinement.
  • The proposed benchmark (123 scenarios) and the layer-based human scoring protocol give the research community a shared, repeatable way to compare scenario-generation methods.
  • Because the framework relies on prompting rather than fine-tuning, it can be applied to different LLMs, but the large gap between proprietary and open models suggests that model capability is currently the binding constraint.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same component-wise generation idea likely transfers to other scenario DSLs (e.g., OpenScenario) and other simulators, so the contribution may generalize beyond the specific language evaluated here.
  • The paper selects the best of 12 ablation configurations on the same benchmark it uses for the final comparison; an independent holdout set would clarify whether the 76.42% CSR reflects true generalizability or benchmark-specific tuning.
  • The human layer-scoring protocol is expensive and subjective; a future automated metric (e.g., a learned verifier that checks alignment against the description) could make the benchmark cheaper and more reproducible.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper proposes Chat2Scenic, an iterative retrieval-augmented generation framework that converts natural-language regulation-style scenario descriptions into executable Scenic programs. The framework parses a description into logical components (global configuration, spatial relations, ego, objects, restrictions), retrieves code snippets and documentation through a dual RAG architecture, and generates components iteratively using contextual prompting, chain-of-thought, few-shot in-context learning, and retrieval-augmented in-context learning. The authors introduce a 123-scenario benchmark drawn from CARLA Leaderboard, NHTSA, and UN regulations, with metrics comprising CSR, response time, token usage, and human-scored layer alignment, plus FA = CSR × SQ. Using Gemini-3-Flash with configuration C11, they report CSR 76.42%, SQ 76.12%, FA 58.17%, outperforming ChatScene (30.08% CSR) and NL2Scenic (16.26% CSR) on the same benchmark.

Significance. If the reported results are unbiased, the contribution is significant: it addresses a real bottleneck in DSL-based scenario generation, introduces a reusable regulation-grounded benchmark, and provides an open-source implementation. Strengths include the public code release, the use of simulator compilation as an objective filter for CSR, and the component-wise iterative design, which is well motivated by the failure modes of full-script generation. The ablation over prompting techniques and the multi-backbone evaluation are also useful. However, the empirical claim currently depends on unaddressed evaluation-hygiene issues: possible overlap between the retrieval database and benchmark scenarios, selection of the best configuration on the same benchmark, and unmeasured human-evaluator variance. These issues are fixable, but the headline numbers cannot be taken at face value until they are resolved.

major comments (3)
  1. [§II-B.1 and §III-A.1] The RAG code database is built from "official Scenic sources" (§II-B.1), while the benchmark deliberately includes 24 CARLA Leaderboard scenarios (§III-A.1). The paper never states that the source files or scenario descriptions corresponding to the 123 benchmark items were excluded from the database. Since CARLA Leaderboard scenarios are official Scenic examples and §III-B.2 notes that ChatScene's database is built from CARLA Leaderboard scenarios, the retriever could return code that is effectively the ground-truth solution for a subset of test items, inflating both CSR and SQ. Please report a decontamination analysis, e.g., remove any snippet whose source scenario appears in the benchmark and re-run, or show retrieval outputs for benchmark queries and verify that no ground-truth snippet is returned.
  2. [§III-B.1, Table III and Table IV] Configuration C11 was selected as the best of 12 ablation configurations evaluated on the complete 123-scenario benchmark (Table III) and then used as the final result in Table IV. The headline 76.42% CSR and 58.17% FA are therefore selected-on-test estimates, not unbiased predictive estimates for new regulation descriptions. This also makes the comparison to ChatScene and NL2Scenic unfair to the extent that those baselines were not configuration-selected on this benchmark. Please either split the benchmark into development and held-out test sets, use nested cross-validation, or at minimum report the distribution over configurations and the selection rule, together with confidence intervals.
  3. [§III-A.3] FA is defined as CSR × SQ, where SQ is the average of human layer-alignment scores. The manuscript reports no detail on the human evaluation: number of evaluators, whether they were blind to the framework or method, whether the same evaluator scored all systems, or inter-rater reliability. Since the headline FA depends directly on these subjective scores, unmeasured evaluator variance could change the comparison. Please provide the evaluation protocol and agreement statistics (e.g., Cohen's kappa or per-layer variance), and ideally report SQ with confidence intervals.
minor comments (6)
  1. [Table I] The table header for RAG-ICL and the distinction between the 'Assemble' and 'Generation' rows are difficult to follow; consider simplifying the column layout or adding a footnote.
  2. [§I-C] The contribution text says '120+ scenario descriptions' while the benchmark has 123; please be consistent.
  3. [§II-C.1] SettingsDetector is described as using 'rule-based keyword matching patterns and example-based reasoning,' but the rules and examples are not provided. This limits reproducibility; please include them in an appendix or supplementary material.
  4. [Table III] For configurations with very low CSR (e.g., C2, C3), the layer-alignment averages are based on a handful of compiled scenarios and contain many dashes. Please clarify how SQ is computed in these cases and consider reporting the number of compiled scenarios.
  5. [§III-A.3] Response time and token usage are reported as averages without variance. Adding standard deviations or confidence intervals would help assess stability across scenarios.
  6. [Abstract and §I-A] The claim 'first iterative retrieval-augmented framework' is strong given that prior work such as Talk2Traffic and Rubavicius already combine interaction and retrieval. Please soften the claim or clarify the specific novelty.

Circularity Check

0 steps flagged

No significant circularity: the reported results are empirical benchmark measurements with independent SOTA baselines, and no derivation step reduces to its own inputs.

full rationale

Chat2Scenic is an empirical systems paper; its central claims are benchmark measurements, not a mathematical derivation. The only composite metric, FA = CSR × SQ, is an explicit definition rather than a quantity that is derived from or equivalent to the inputs, and its components are measured independently: CSR by executing Scenic programs in CARLA, and SQ by human layer-wise evaluation. The RAG code database is stated to come from official Scenic sources, while the benchmark is drawn from CARLA Leaderboard, NHTSA, and UN regulations; no quoted passage establishes that benchmark scenarios were already present in the retrieval database. The comparison against ChatScene and NL2Scenic on the same benchmark provides external grounding for the framework's relative improvement. The self-citation [28] is used only as an explanatory aside for why Gemini-3-Flash outperforms Gemini-3-Pro and is not load-bearing for the headline CSR/FA numbers. Selecting configuration C11 on the same benchmark used for final reporting is a legitimate model-selection/generalization-risk concern, but it is not a circularity in the derivation-chain sense: no fitted parameter is renamed as a prediction. Similarly, the absence of an explicit de-duplication statement is a potential data-leakage risk, not a demonstrated circular step. Under the rule that circularity requires a quotable reduction (Eq. X = Eq. Y by construction, or a fitted input presented as a prediction), no such reduction is present.

Axiom & Free-Parameter Ledger

5 free parameters · 6 axioms · 0 invented entities

No new physical entities are introduced; the contribution is an engineering framework and benchmark. The central claim rests on several hand-set parameters (retrieval K, confidence threshold, defaults, selected prompting configuration) and on domain assumptions about LLM capability, database coverage, and the validity of the human-evaluation metric.

free parameters (5)
  • Code retriever top-K = 3
    Section II-B.3: code retriever returns top-3 snippets per component; hand-chosen and affects how much context the generator receives.
  • SettingsDetector confidence threshold = 0.6
    Algorithm 1 line 8: only detected settings with confidence >= 0.6 are used; hand-set threshold determines when defaults replace detected values.
  • Default global settings = Town05, ClearNoon, Lincoln mkz
    Algorithm 1 line 11 and Section II-C.1: low-confidence or null settings fall back to these defaults; performance depends on how often these defaults are correct for the benchmark scenarios.
  • Prompting configuration C11 = CP+CoT+ICL+CodeICL (DocICL off)
    Table III: selected as best on the same 123-scenario benchmark, then used for headline results in Table IV; selection on the evaluation set can inflate reported CSR/FA.
  • Human Scenario Quality scoring = Average over six layers
    Section III-A.3: SQ is derived from human layer scores; no inter-rater reliability or scoring rubric details are provided, so the FA = CSR x SQ headline depends on this subjective measurement.
axioms (6)
  • domain assumption Pre-trained LLMs, prompted with CP/CoT/ICL/RAG-ICL, can translate component descriptions into compilable Scenic code.
    The entire generation module assumes no fine-tuning is needed and that prompting is sufficient. The strong dependence on Gemini-3-Flash and near-zero results for open-source models make this a load-bearing assumption.
  • ad hoc to paper Concatenation of individually generated Scenic components yields a valid, executable Scenic program.
    Algorithm 1 line 22 concatenates G, R, E, O, T; the paper's CSR depends on cross-component compatibility, but no formal or empirical argument is given beyond context accumulation.
  • domain assumption The official Scenic code snippet and documentation database covers the component types needed for all 123 benchmark scenarios.
    Section II-B: RAG databases built from official Scenic sources; if a scenario component has no near code or doc, generation cannot succeed, bounding CSR.
  • domain assumption Human layer-based scoring is a valid measure of semantic fidelity.
    Section III-A.3: SQ uses the 6-layer model but no inter-annotator agreement, no evaluator count, and no validation of this metric as a proxy for scenario correctness.
  • domain assumption Benchmark scenario texts from CARLA, NHTSA, and UN regulations are complete and unambiguous enough to serve as ground truth.
    Table II and Section III-A.1: if source descriptions omit details or contain ambiguities, human SQ scores reflect guesswork as much as framework accuracy.
  • domain assumption The same prompting strategy is appropriate for all LLMs without per-model tuning.
    Table IV evaluates all models under the C11 configuration; poor open-source performance may reflect prompt-template mismatch rather than intrinsic model capability.

pith-pipeline@v1.3.0-alltime-deepseek · 10941 in / 11302 out tokens · 117965 ms · 2026-08-02T02:12:52.747530+00:00 · methodology

0 comments
read the original abstract

Validating autonomous driving systems requires diverse, regulation-compliant test scenarios. In simulation-based testing, scenarios are defined as executable scripts. Yet automatically generating such scripts from regulatory descriptions remains an open challenge, and existing approaches face fundamental trade-offs. Retrieval-assemble methods achieve reasonable compilation rates but lack scalability, whereas retrieval-based full-script generation suffers from low compilation success rates. We present Chat2Scenic, the first iterative retrieval-augmented framework to generate scenario scripts in Domain Specific Language (DSL). Specifically, Chat2Scenic provides a chatbot interface that supports interactive scenario refinement and integrates Retrieval-augmented Generation (RAG) to ground scenario generation in regulatory knowledge and DSL syntax. Furthermore, we propose an open benchmark for scenario generation comprising 123 scenarios from various regulations, including NHTSA and United Nations Vehicle Regulations, as well as other sources. Extensive evaluation with State-of-the-Art (SOTA) Large Language Models (LLMs) demonstrates that Chat2Scenic achieves 76.42% Compilation Success Rate (CSR) and 58.17% Framework Accuracy (FA), outperforming existing methods (Retrieval Assemble with 30.08% CSR, 11.03% FA and Retrieval full script generation with 16.26% CSR, 10.86% FA). To facilitate future research, we release our code as open source at https://github.com/TUM-AVS/chat2scenic.

Figures

Figures reproduced from arXiv: 2607.14387 by Haoyu Wang, Johannes Betz, Mattia Piccinini, Qunying Song, Wenting Miao, Yuan Gao.

Figure 1
Figure 1. Figure 1: Comparison of DSL-based scenario generation ap [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the Chat2Scenic framework, consisting of three core modules: the [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Users can describe scenarios in natural language, [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 3
Figure 3. Figure 3: Chatbot UI for interactive refinement. B. RAG Module To enhance the LLM performance for domain-specific tasks, we integrate RAG supported by the LangChain3 framework. The RAG technique retrieves relevant exter￾nal knowledge and returns it to the LLM during scenario generation, improving accuracy and reducing hallucinations. Specifically, we construct two complementary databases: a code snippet database and… view at source ↗
Figure 4
Figure 4. Figure 4: Prompting strategies for component generation. [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Example scenarios from three sources: a) CARLA leaderboard, b) UN R171, c) NHTSA Crash. [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

28 extracted references · 6 linked inside Pith

  1. [1]

    A new taxonomy for automated driving: Structuring applications based on their operational design domain, level of automation and automation readiness,

    J. Betz, M. Lutwitzi, and S. Peters, “A new taxonomy for automated driving: Structuring applications based on their operational design domain, level of automation and automation readiness,” in2024 IEEE Intelligent Vehicles Symposium (IV), 2024

    2024

  2. [2]

    B ¨aumler and E

    M. B ¨aumler and E. Cuvillier,Generating Test Scenarios for Assessing Automated Driving Systems Using Scenario Fusion, ser. Schriftenreihe des Lehrstuhls Kraftfahrzeugtechnik. Cuvillier Verlag, 2024

    2024

  3. [3]

    Critical scenario identification for realistic testing of autonomous driving systems,

    Q. Song, K. Tan, P. Runeson, and S. Persson, “Critical scenario identification for realistic testing of autonomous driving systems,” Software Quality Journal, 2023

    2023

  4. [4]

    Target: Automated scenario generation from traffic rules for testing autonomous vehicles,

    Y . Deng, J. Yao, Z. Tu, X. Zheng, M. Zhanget al., “Target: Automated scenario generation from traffic rules for testing autonomous vehicles,” arXiv preprint arXiv:2305.06018, 2023

    Pith/arXiv arXiv 2023

  5. [5]

    Chatscene: Knowledge-enabled safety- critical scenario generation for autonomous vehicles,

    J. Zhang, C. Xu, and B. Li, “Chatscene: Knowledge-enabled safety- critical scenario generation for autonomous vehicles,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024

    2024

  6. [6]

    Text2scenario: Text- driven scenario generation for autonomous driving test,

    X. Cai, X. Bai, Z. Cui, D. Xie, D. Fuet al., “Text2scenario: Text- driven scenario generation for autonomous driving test,”Automotive Innovation, 2026

    2026

  7. [7]

    Lmm- enhanced safety-critical scenario generation for autonomous driv- ing system testing from non-accident traffic videos,

    H. Tian, X. Han, Y . Zhou, G. Wu, A. Guoet al., “Lmm- enhanced safety-critical scenario generation for autonomous driv- ing system testing from non-accident traffic videos,”arXiv preprint arXiv:2406.10857, 2024

    Pith/arXiv arXiv 2024

  8. [8]

    Legend: A top- down approach to scenario generation of autonomous driving systems assisted by large language models,

    S. Tang, Z. Zhang, J. Zhou, L. Lei, Y . Zhouet al., “Legend: A top- down approach to scenario generation of autonomous driving systems assisted by large language models,” in39th IEEE/ACM International Conference on Automated Software Engineering, 2024

    2024

  9. [9]

    Scenicnl: generating proba- bilistic scenario programs from natural language,

    K. Elmaaroufi, D. Shanker, A. Cismaru, M. Vazquez-Chanlatte, A. Sangiovanni-Vincentelliet al., “Scenicnl: generating proba- bilistic scenario programs from natural language,”arXiv preprint arXiv:2405.03709, 2024

    Pith/arXiv arXiv 2024

  10. [10]

    Traffic scene generation from natural language description for autonomous vehicles with large language model,

    B.-K. Ruan, H.-T. Tsui, Y .-H. Li, and H.-H. Shuai, “Traffic scene generation from natural language description for autonomous vehicles with large language model,”arXiv preprint arXiv:2409.09575, 2024

    arXiv 2024

  11. [11]

    Conversational code generation: a case study of designing a dia- logue system for generating driving scenarios for testing autonomous vehicles,

    R. Rubavicius, A. V . Miceli-Barone, A. Lascarides, and S. Ramamoor- thy, “Conversational code generation: a case study of designing a dia- logue system for generating driving scenarios for testing autonomous vehicles,”arXiv preprint arXiv:2410.09829, 2024

    Pith/arXiv arXiv 2024

  12. [12]

    David vs. goliath: A comparative study of different- sized llms for code generation in the domain of automotive scenario generation,

    P. Bauerfeind, A. Salarpour, D. Fernandez, P. MohajerAnsari, J. Reschkeet al., “David vs. goliath: A comparative study of different- sized llms for code generation in the domain of automotive scenario generation,”arXiv preprint arXiv:2510.14115, 2025

    arXiv 2025

  13. [13]

    A survey on safety- critical driving scenario generation—a methodological perspective,

    W. Ding, C. Xu, M. Arief, H. Lin, B. Liet al., “A survey on safety- critical driving scenario generation—a methodological perspective,” IEEE Transactions on Intelligent Transportation Systems, 2023

    2023

  14. [14]

    1001 ways of scenario generation for testing of self-driving cars: A survey,

    B. Sch ¨utt, J. Ransiek, T. Braun, and E. Sax, “1001 ways of scenario generation for testing of self-driving cars: A survey,” in2023 IEEE Intelligent Vehicles Symposium (IV). IEEE, 2023

    2023

  15. [15]

    Foundation models in autonomous driving: A survey on scenario generation and scenario analysis,

    Y . Gao, M. Piccinini, Y . Zhang, D. Wang, K. Molleret al., “Foundation models in autonomous driving: A survey on scenario generation and scenario analysis,”IEEE Open Journal of Intelligent Transportation Systems, 2026

    2026

  16. [16]

    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

    2017

  17. [17]

    Microscopic traffic simulation using sumo,

    P. A. Lopez, M. Behrisch, L. Bieker-Walz, J. Erdmann, Y .-P. Fl¨otter¨od et al., “Microscopic traffic simulation using sumo,” inThe 21st IEEE International Conference on Intelligent Transportation Systems. IEEE, 2018

    2018

  18. [18]

    Chatsumo: Large language model for automating traffic scenario generation in simulation of urban mobility,

    S. Li, T. Azfar, and R. Ke, “Chatsumo: Large language model for automating traffic scenario generation in simulation of urban mobility,” IEEE Transactions on Intelligent Vehicles, 2024

    2024

  19. [19]

    Multimodal large language model driven scenario testing for autonomous vehicles,

    Q. Lu, X. Wang, Y . Jiang, G. Zhao, M. Maet al., “Multimodal large language model driven scenario testing for autonomous vehicles,” arXiv preprint arXiv:2409.06450, 2024

    Pith/arXiv arXiv 2024

  20. [20]

    Generating traffic scenarios via in-context learning to learn better motion planner,

    A. Aiersilan, “Generating traffic scenarios via in-context learning to learn better motion planner,” inAAAI Conference on Artificial Intelligence, 2025

    2025

  21. [21]

    Sovar: Build generalizable scenarios from accident reports for autonomous driving testing,

    A. Guo, Y . Zhou, H. Tian, C. Fang, Y . Sunet al., “Sovar: Build generalizable scenarios from accident reports for autonomous driving testing,” inProceedings of the 39th IEEE/ACM International Confer- ence on Automated Software Engineering, 2024

    2024

  22. [22]

    Linguasim: Interactive multi-vehicle testing scenario generation via natural lan- guage instruction based on large language models,

    Q. Shi, Q. Meng, H. Cheng, Q. Xu, and J. Wang, “Linguasim: Interactive multi-vehicle testing scenario generation via natural lan- guage instruction based on large language models,”arXiv preprint arXiv:2510.08046, 2025

    arXiv 2025

  23. [23]

    Realistic corner case generation for autonomous vehicles with multimodal large language model,

    Q. Lu, M. Ma, X. Dai, X. Wang, and S. Feng, “Realistic corner case generation for autonomous vehicles with multimodal large language model,”arXiv preprint arXiv:2412.00243, 2024

    Pith/arXiv arXiv 2024

  24. [24]

    Automatic generation method for autonomous driving simulation scenarios based on large language model,

    X. Zhou, Y . Huang, J. Zhang, J. Shao, D. Panet al., “Automatic generation method for autonomous driving simulation scenarios based on large language model,” inInternational Conference on Artificial Intelligence and Autonomous Transportation. Springer, 2024

    2024

  25. [25]

    Scenic: a language for scenario specification and scene generation,

    D. J. Fremont, T. Dreossi, S. Ghosh, X. Yue, A. L. Sangiovanni- Vincentelliet al., “Scenic: a language for scenario specification and scene generation,” in40th ACM SIGPLAN conference on programming language design and implementation, 2019

    2019

  26. [26]

    From words to collisions: Llm-guided evaluation and adversarial generation of safety-critical driving scenarios,

    Y . Gao, M. Piccinini, K. Moller, A. Alanwar, and J. Betz, “From words to collisions: Llm-guided evaluation and adversarial generation of safety-critical driving scenarios,” in2025 IEEE 28th International Conference on Intelligent Transportation Systems (ITSC). IEEE, 2025

    2025

  27. [27]

    6-layer model for a structured description and categorization of urban traffic and environment,

    M. Scholtes, L. Westhofen, L. R. Turner, K. Lotto, M. Schuldeset al., “6-layer model for a structured description and categorization of urban traffic and environment,”IEEE Access, 2021

    2021

  28. [28]

    Nurisk: A visual question answering dataset for agent-level risk assessment in autonomous driving,

    Y . Gao, M. Piccinini, R. Brusnicki, Y . Zhang, and J. Betz, “Nurisk: A visual question answering dataset for agent-level risk assessment in autonomous driving,” in2026 IEEE International Conference on Robotics and Automation (ICRA), 2026

    2026