pith. sign in

arxiv: 2508.13881 · v2 · submitted 2025-08-19 · 💻 cs.RO

Driving Style Recognition Like an Expert Using Semantic Privileged Information from Large Language Models

Zhaokun Chen , Chaopeng Zhang , Xiaohan Li , Wenshuo Wang , Gentiane Venture , Junqiang Xi This is my paper

Pith reviewed 2026-05-18 22:28 UTC · model grok-4.3

classification 💻 cs.RO
keywords driving style recognitionsemantic privileged informationlarge language modelsSVM+DriBehavGPTcar followinglane changing
0
0 comments X

The pith

Incorporating LLM-generated semantic descriptions as privileged information during training improves driving style recognition to better align with human expert judgments.

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

Existing driving style recognition systems depend on low-level sensor data and miss the rich semantic reasoning of human experts. This leads to misalignment with expert judgments. The paper introduces DriBehavGPT to generate natural language descriptions of behaviors using LLMs. These are encoded and used as privileged information in SVM+ training to help the model learn expert patterns. This results in F1 improvements of 7.6% for car-following and 7.9% for lane-changing, with efficient inference on sensor data only.

Core claim

By using semantic privileged information from large language models in the training phase of SVM+, the framework enables driving style classifiers to approximate the semantic reasoning of human experts, leading to higher accuracy on sensor-based inference without added runtime cost.

What carries the argument

Semantic Privileged Information (SPI) from LLM-generated natural language descriptions of driving behaviors, integrated into SVM+ training.

If this is right

  • Improved F1 scores of 7.6% in car-following and 7.9% in lane-changing scenarios compared to conventional methods.
  • The model maintains efficiency by using only sensor data during inference.
  • Enhances alignment between algorithmic classifications and expert judgments.
  • Supports more interpretable human-centric autonomous driving systems.

Where Pith is reading between the lines

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

  • This method could extend to other areas of autonomous systems where high-level semantic knowledge from experts is available only at training time.
  • Similar privileged information techniques might help bridge gaps in other machine learning applications involving human-like reasoning.
  • Testing with varied LLMs could reveal how description quality affects the final recognition performance.

Load-bearing premise

The LLM-generated descriptions faithfully represent the semantic reasoning patterns of human driving experts without hallucinations or biases.

What would settle it

Collecting ratings from human driving experts on classifications from both the SPI model and a baseline model across the same set of driving scenarios to check for closer alignment with the SPI version.

Figures

Figures reproduced from arXiv: 2508.13881 by Chaopeng Zhang, Gentiane Venture, Junqiang Xi, Wenshuo Wang, Xiaohan Li, Zhaokun Chen.

Figure 1
Figure 1. Figure 1: The difference between an expert and an algorithm for driving style [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The LUSPI framework for driving style recognition. The input consists of the [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: DriBehavGPT for semantically describing driving behaviors. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: The driving routes of experiments in Changchun, China. [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: Illustration of key time-point in lane-changing behavior definition. [PITH_FULL_IMAGE:figures/full_fig_p005_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Three-stage annotation workflow for driving style labeling. [PITH_FULL_IMAGE:figures/full_fig_p006_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Prompt design for DriBehavGPT in car-following ( [PITH_FULL_IMAGE:figures/full_fig_p007_8.png] view at source ↗
Figure 10
Figure 10. Figure 10: Illustration of chain-of-thought reasoning by DriBehavGPT in a lane [PITH_FULL_IMAGE:figures/full_fig_p009_10.png] view at source ↗
read the original abstract

Existing driving style recognition systems largely depend on low-level sensor-derived features for training, neglecting the rich semantic reasoning capability inherent to human experts. This discrepancy results in a fundamental misalignment between algorithmic classifications and expert judgments. To bridge this gap, we propose a novel framework that integrates Semantic Privileged Information (SPI) derived from large language models (LLMs) to align recognition outcomes with human-interpretable reasoning. First, we introduce DriBehavGPT, an interactive LLM-based module that generates natural-language descriptions of driving behaviors. These descriptions are then encoded into machine learning-compatible representations via text embedding and dimensionality reduction. Finally, we incorporate them as privileged information into Support Vector Machine Plus (SVM+) for training, enabling the model to approximate human-like interpretation patterns. Experiments across diverse real-world driving scenarios demonstrate that our SPI-enhanced framework outperforms conventional methods, achieving F1-score improvements of 7.6% (car-following) and 7.9% (lane-changing). Importantly, SPI is exclusively used during training, while inference relies solely on sensor data, ensuring computational efficiency without sacrificing performance. These results highlight the pivotal role of semantic behavioral representations in improving recognition accuracy while advancing interpretable, human-centric driving systems.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript proposes a framework for driving style recognition that uses an LLM-based module (DriBehavGPT) to generate natural-language descriptions of driving behaviors as Semantic Privileged Information (SPI). These descriptions are encoded via text embedding and dimensionality reduction, then incorporated into SVM+ training to align classifications with human expert semantic reasoning. SPI is used only during training; inference relies solely on sensor data. Experiments on real-world car-following and lane-changing scenarios report F1-score gains of 7.6% and 7.9% over conventional methods.

Significance. If the central assumption holds, the work could meaningfully advance human-centric autonomous driving by injecting high-level semantic knowledge into standard ML pipelines without runtime cost. The training-only use of SPI is a practical design choice that supports efficient deployment.

major comments (2)
  1. [Abstract and Method] Abstract and § on DriBehavGPT: The performance improvements are attributed to SPI from DriBehavGPT faithfully capturing expert semantic reasoning. However, the manuscript provides no human expert review, inter-rater agreement scores, or quantitative fidelity metrics on the generated descriptions, leaving open the possibility that gains arise from LLM noise or bias rather than true alignment (as highlighted by the weakest assumption in the reader's report).
  2. [Experiments] Experiments section: The abstract states F1 improvements of 7.6% (car-following) and 7.9% (lane-changing) but supplies no dataset size, baseline details, statistical tests, or confound controls. This absence prevents verification that the data support the central claim of outperforming conventional methods.
minor comments (1)
  1. Define all acronyms (e.g., SPI, SVM+, LLM) on first use and ensure consistent notation for the embedding and dimensionality-reduction steps.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address each major comment point by point below, with clear indications of planned revisions to strengthen the work.

read point-by-point responses

  1. Referee: [Abstract and Method] Abstract and § on DriBehavGPT: The performance improvements are attributed to SPI from DriBehavGPT faithfully capturing expert semantic reasoning. However, the manuscript provides no human expert review, inter-rater agreement scores, or quantitative fidelity metrics on the generated descriptions, leaving open the possibility that gains arise from LLM noise or bias rather than true alignment (as highlighted by the weakest assumption in the reader's report).

    Authors: We agree that the absence of direct human expert validation or quantitative fidelity metrics for the DriBehavGPT-generated descriptions represents a gap in substantiating the claim of alignment with expert semantic reasoning. The manuscript currently relies on downstream performance gains as indirect support for the utility of the SPI. To address this, we will revise the relevant sections to explicitly discuss the potential for LLM-generated bias or noise, include qualitative examples of the natural-language descriptions, and add a limitations paragraph acknowledging the lack of inter-rater agreement scores. While a full human evaluation study is beyond the scope of the current work, these changes will clarify the assumptions and provide a more balanced presentation. revision: partial

  2. Referee: [Experiments] Experiments section: The abstract states F1 improvements of 7.6% (car-following) and 7.9% (lane-changing) but supplies no dataset size, baseline details, statistical tests, or confound controls. This absence prevents verification that the data support the central claim of outperforming conventional methods.

    Authors: We acknowledge that additional experimental details are necessary for full verification and reproducibility. The manuscript describes real-world car-following and lane-changing datasets but does not explicitly report sample sizes, list all baselines with implementation details, or include statistical significance testing in the provided sections. We will revise the experiments section to report dataset sizes (number of trajectories and scenarios), specify the conventional baselines (including standard SVM and other classifiers), add statistical tests (e.g., paired significance tests on F1 scores), and discuss potential confounds such as scenario variability or sensor noise. These updates will be incorporated to directly support the reported improvements. revision: yes

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper's core derivation introduces DriBehavGPT to generate natural-language driving behavior descriptions from an external LLM, encodes them via text embedding and dimensionality reduction, and feeds the result as privileged information exclusively into SVM+ training (with inference using only sensor data). The reported F1 improvements (7.6% and 7.9%) are presented as outcomes of experiments on real-world data rather than any equation or step that reduces by construction to a fitted parameter, self-definition, or self-citation chain. No load-bearing self-citations, uniqueness theorems, or ansatzes are invoked that collapse the claimed alignment with expert reasoning back to the inputs. The framework remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claim rests on the premise that LLM outputs can serve as faithful proxies for expert semantic reasoning and that these proxies transfer usefully through the SVM+ training stage.

axioms (1)
  • domain assumption LLM-generated natural-language descriptions of driving behaviors accurately reflect the semantic reasoning patterns of human experts
    This premise is required for the privileged information to align algorithmic outputs with expert judgments as stated in the abstract.
invented entities (1)
  • DriBehavGPT no independent evidence
    purpose: Interactive LLM-based module that generates natural-language descriptions of driving behaviors from sensor data
    The module is introduced by the authors to produce the semantic privileged information.

pith-pipeline@v0.9.0 · 5762 in / 1368 out tokens · 58109 ms · 2026-05-18T22:28:28.089284+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

65 extracted references · 65 canonical work pages · 3 internal anchors

  1. [1]

    Intelligent in-vehicle inter- action technologies,

    P. K. Murali, M. Kaboli, and R. Dahiya, “Intelligent in-vehicle inter- action technologies,” Advanced Intelligent Systems , vol. 4, no. 2, p. 2100122, 2022

    work page 2022

  2. [2]

    Satisfactory driving mode clas- sification based on pedal operation characteristics,

    Y . Kim, S. B. Choi, J. J. Oh, and J. Eo, “Satisfactory driving mode clas- sification based on pedal operation characteristics,” IEEE Transactions on Intelligent Vehicles, vol. 9, no. 1, pp. 2988–2998, 2023

    work page 2023

  3. [3]

    Drive mode optimization and path planning for plug-in hybrid electric vehicles,

    C.-K. Chau, K. Elbassioni, and C.-M. Tseng, “Drive mode optimization and path planning for plug-in hybrid electric vehicles,” IEEE Transac- tions on Intelligent Transportation Systems , vol. 18, no. 12, pp. 3421– 3432, 2017

    work page 2017

  4. [4]

    Personalized adaptive cruise control based on online driving style recognition technology and model predictive control,

    B. Gao, K. Cai, T. Qu, Y . Hu, and H. Chen, “Personalized adaptive cruise control based on online driving style recognition technology and model predictive control,” IEEE transactions on vehicular technology , vol. 69, no. 11, pp. 12 482–12 496, 2020

    work page 2020

  5. [5]

    Typical- driving-style-oriented personalized adaptive cruise control design based on human driving data,

    B. Zhu, Y . Jiang, J. Zhao, R. He, N. Bian, and W. Deng, “Typical- driving-style-oriented personalized adaptive cruise control design based on human driving data,” Transportation research part C: emerging technologies, vol. 100, pp. 274–288, 2019

    work page 2019

  6. [6]

    Develop- ment and evaluation of two learning-based personalized driver models for car-following behaviors,

    W. Wang, D. Zhao, J. Xi, D. J. LeBlanc, and J. K. Hedrick, “Develop- ment and evaluation of two learning-based personalized driver models for car-following behaviors,” in 2017 American Control Conference (ACC). IEEE, 2017, pp. 1133–1138

    work page 2017

  7. [7]

    Study of semi-active suspension control strategy based on driving behaviour characteristics,

    W. Wang and J. Xi, “Study of semi-active suspension control strategy based on driving behaviour characteristics,” International Journal of Vehicle Design, vol. 68, no. 1-3, pp. 141–161, 2015

    work page 2015

  8. [8]

    Human-centered feed-forward control of a vehicle steering system based on a driver’s path-following characteristics,

    W. Wang, J. Xi, C. Liu, and X. Li, “Human-centered feed-forward control of a vehicle steering system based on a driver’s path-following characteristics,” IEEE transactions on intelligent transportation systems, vol. 18, no. 6, pp. 1440–1453, 2016

    work page 2016

  9. [9]

    A learning-based approach for lane departure warning systems with a personalized driver model,

    W. Wang, D. Zhao, W. Han, and J. Xi, “A learning-based approach for lane departure warning systems with a personalized driver model,” IEEE Transactions on Vehicular Technology, vol. 67, no. 10, pp. 9145–9157, 2018

    work page 2018

  10. [10]

    Online driving style recognition using fuzzy logic,

    D. D ¨orr, D. Grabengiesser, and F. Gauterin, “Online driving style recognition using fuzzy logic,” in 17th international IEEE conference on intelligent transportation systems (ITSC) . IEEE, 2014, pp. 1021–1026

    work page 2014

  11. [11]

    Study on the driving style adaptive vehicle longitudinal control strategy,

    J. Huang, Y . Chen, X. Peng, L. Hu, and D. Cao, “Study on the driving style adaptive vehicle longitudinal control strategy,” IEEE/CAA Journal of Automatica Sinica , vol. 7, no. 4, pp. 1107–1115, 2020

    work page 2020

  12. [12]

    A driving-style-oriented adaptive control strategy based pso-fuzzy expert algorithm for a plug-in hybrid electric vehicle,

    X. Lin, K. Li, and L. Wang, “A driving-style-oriented adaptive control strategy based pso-fuzzy expert algorithm for a plug-in hybrid electric vehicle,” Expert Systems with Applications , vol. 201, p. 117236, 2022

    work page 2022

  13. [13]

    Comprehensive iden- tification of driving style based on vehicle’s driving cycle recognition,

    K. Liang, Z. Zhao, W. Li, J. Zhou, and D. Yan, “Comprehensive iden- tification of driving style based on vehicle’s driving cycle recognition,” IEEE Transactions on Vehicular Technology, vol. 72, no. 1, pp. 312–326, 2022

    work page 2022

  14. [14]

    Classifying travel- ers’ driving style using basic safety messages generated by connected vehicles: Application of unsupervised machine learning,

    A. Mohammadnazar, R. Arvin, and A. J. Khattak, “Classifying travel- ers’ driving style using basic safety messages generated by connected vehicles: Application of unsupervised machine learning,” Transportation research part C: emerging technologies , vol. 122, p. 102917, 2021

    work page 2021

  15. [15]

    Incorporating driving style recognition into mpc for energy management of plug-in hybrid electric buses,

    X. Tian, Y . Cai, X. Sun, Z. Zhu, Y . Wang, and Y . Xu, “Incorporating driving style recognition into mpc for energy management of plug-in hybrid electric buses,” IEEE Transactions on Transportation Electrifica- tion, vol. 9, no. 1, pp. 169–181, 2022. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 11

    work page 2022

  16. [16]

    Recognition of trip-based aggres- sive driving: A system integrated with gaussian mixture model structured of factor-analysis, and hierarchical clustering,

    J. Wang, W. Xu, T. Fu, and R. Jiang, “Recognition of trip-based aggres- sive driving: A system integrated with gaussian mixture model structured of factor-analysis, and hierarchical clustering,” IEEE Transactions on Intelligent Transportation Systems , vol. 23, no. 11, pp. 20 442–20 451, 2022

    work page 2022

  17. [17]

    An embedded driving style recognition approach: Leveraging knowledge in learning,

    C. Zhang, W. Wang, Z. Ju, Z. Chen, G. Venture, and J. Xi, “An embedded driving style recognition approach: Leveraging knowledge in learning,” IEEE Transactions on Intelligent Vehicles , 2024

    work page 2024

  18. [18]

    A theory of semantic information,

    Y . Zhong, “A theory of semantic information,” China communications, vol. 14, no. 1, pp. 1–17, 2017

    work page 2017

  19. [19]

    Au- tonomous driving: cognitive construction and situation understanding,

    S. Chen, Z. Jian, Y . Huang, Y . Chen, Z. Zhou, and N. Zheng, “Au- tonomous driving: cognitive construction and situation understanding,” Science China Information Sciences , vol. 62, pp. 1–27, 2019

    work page 2019

  20. [20]

    Driving style alignment for llm-powered driver agent,

    R. Yang, X. Zhang, A. Fernandez-Laaksonen, X. Ding, and J. Gong, “Driving style alignment for llm-powered driver agent,” in 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2024, pp. 11 318–11 324

    work page 2024

  21. [21]

    Recent advances in natural language processing via large pre-trained language models: A survey,

    B. Min, H. Ross, E. Sulem, A. P. B. Veyseh, T. H. Nguyen, O. Sainz, E. Agirre, I. Heintz, and D. Roth, “Recent advances in natural language processing via large pre-trained language models: A survey,” ACM Computing Surveys, vol. 56, no. 2, pp. 1–40, 2023

    work page 2023

  22. [22]

    Chain-of-thought prompting elicits reasoning in large language models,

    J. Wei, X. Wang, D. Schuurmans, M. Bosma, F. Xia, E. Chi, Q. V . Le, D. Zhou et al. , “Chain-of-thought prompting elicits reasoning in large language models,” Advances in neural information processing systems , vol. 35, pp. 24 824–24 837, 2022

    work page 2022

  23. [23]

    On the theory of learnining with privi- leged information,

    D. Pechyony and V . Vapnik, “On the theory of learnining with privi- leged information,” Advances in neural information processing systems , vol. 23, 2010

    work page 2010

  24. [24]

    The neural and computational bases of semantic cognition,

    M. A. L. Ralph, E. Jefferies, K. Patterson, and T. T. Rogers, “The neural and computational bases of semantic cognition,” Nature reviews neuroscience, vol. 18, no. 1, pp. 42–55, 2017

    work page 2017

  25. [25]

    Understanding common human driving semantics for autonomous vehicles,

    Y . Xia, M. Geng, Y . Chen, S. Sun, C. Liao, Z. Zhu, Z. Li, W. Y . Ochieng, P. Angeloudis, M. Elhajj et al., “Understanding common human driving semantics for autonomous vehicles,” Patterns, 2023

    work page 2023

  26. [26]

    Dimensions of driver stress,

    E. Gulian, G. Matthews, A. I. Glendon, D. Davies, and L. Debney, “Dimensions of driver stress,” Ergonomics, vol. 32, no. 6, pp. 585–602, 1989

    work page 1989

  27. [27]

    Decision- making style, driving style, and self-reported involvement in road traffic accidents,

    D. J. French, R. J. West, J. Elander, and J. M. Wilding, “Decision- making style, driving style, and self-reported involvement in road traffic accidents,” Ergonomics, vol. 36, no. 6, pp. 627–644, 1993

    work page 1993

  28. [28]

    Driving style analysis using primitive driving patterns with bayesian nonparametric approaches,

    W. Wang, J. Xi, and D. Zhao, “Driving style analysis using primitive driving patterns with bayesian nonparametric approaches,” IEEE Trans- actions on Intelligent Transportation Systems , vol. 20, no. 8, pp. 2986– 2998, 2018

    work page 2018

  29. [29]

    Ex- ploring behavioral patterns of lane change maneuvers for human-like autonomous driving,

    Y . Chen, G. Li, S. Li, W. Wang, S. E. Li, and B. Cheng, “Ex- ploring behavioral patterns of lane change maneuvers for human-like autonomous driving,” IEEE Transactions on Intelligent Transportation Systems, vol. 23, no. 9, pp. 14 322–14 335, 2021

    work page 2021

  30. [30]

    Shareable driving style learning and analysis with a hierarchical latent model,

    C. Zhang, W. Wang, Z. Chen, J. Zhang, L. Sun, and J. Xi, “Shareable driving style learning and analysis with a hierarchical latent model,” IEEE Transactions on Intelligent Transportation Systems, vol. 25, no. 9, pp. 11 471–11 484, 2024

    work page 2024

  31. [31]

    A new learning paradigm: Learning using privileged information,

    V . Vapnik and A. Vashist, “A new learning paradigm: Learning using privileged information,” Neural networks, vol. 22, no. 5-6, pp. 544–557, 2009

    work page 2009

  32. [32]

    Multi-frequency representation enhancement with privilege information for video super-resolution,

    F. Li, L. Zhang, Z. Liu, J. Lei, and Z. Li, “Multi-frequency representation enhancement with privilege information for video super-resolution,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 12 814–12 825

    work page 2023

  33. [33]

    Person re-identification with metric learning using privileged information,

    X. Yang, M. Wang, and D. Tao, “Person re-identification with metric learning using privileged information,” IEEE Transactions on Image Processing, vol. 27, no. 2, pp. 791–805, 2017

    work page 2017

  34. [34]

    When does privileged information explain away label noise?

    G. Ortiz-Jimenez, M. Collier, A. Nawalgaria, A. N. D’Amour, J. Berent, R. Jenatton, and E. Kokiopoulou, “When does privileged information explain away label noise?” in International Conference on Machine Learning. PMLR, 2023, pp. 26 646–26 669

    work page 2023

  35. [35]

    Ensemble deep random vector func- tional link network using privileged information for alzheimer’s disease diagnosis,

    M. Ganaie and M. Tanveer, “Ensemble deep random vector func- tional link network using privileged information for alzheimer’s disease diagnosis,” IEEE/ACM Transactions on Computational Biology and Bioinformatics, 2022

    work page 2022

  36. [36]

    Privileged modality distillation for vessel border detection in intracoronary imaging,

    Z. Gao, J. Chung, M. Abdelrazek, S. Leung, W. K. Hau, Z. Xian, H. Zhang, and S. Li, “Privileged modality distillation for vessel border detection in intracoronary imaging,” IEEE transactions on medical imaging, vol. 39, no. 5, pp. 1524–1534, 2019

    work page 2019

  37. [37]

    Learning and refining of privileged information- based rnns for action recognition from depth sequences,

    Z. Shi and T.-K. Kim, “Learning and refining of privileged information- based rnns for action recognition from depth sequences,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2017, pp. 3461–3470

    work page 2017

  38. [38]

    Exploring privileged information from simple actions for complex action recognition,

    F. Liu, X. Xu, T. Zhang, K. Guo, and L. Wang, “Exploring privileged information from simple actions for complex action recognition,” Neu- rocomputing, vol. 380, pp. 236–245, 2020

    work page 2020

  39. [39]

    Learning using partially available privileged informa- tion and label uncertainty: Application in detection of acute respiratory distress syndrome,

    E. Sabeti, J. Drews, N. Reamaroon, E. Warner, M. W. Sjoding, J. Gryak, and K. Najarian, “Learning using partially available privileged informa- tion and label uncertainty: Application in detection of acute respiratory distress syndrome,” IEEE journal of biomedical and health informatics , vol. 25, no. 3, pp. 784–796, 2020

    work page 2020

  40. [40]

    Drive like a human: Rethinking autonomous driving with large language models,

    D. Fu, X. Li, L. Wen, M. Dou, P. Cai, B. Shi, and Y . Qiao, “Drive like a human: Rethinking autonomous driving with large language models,” in Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2024, pp. 910–919

    work page 2024

  41. [41]

    Drive as you speak: Enabling human-like interaction with large language models in au- tonomous vehicles,

    C. Cui, Y . Ma, X. Cao, W. Ye, and Z. Wang, “Drive as you speak: Enabling human-like interaction with large language models in au- tonomous vehicles,” in Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision , 2024, pp. 902–909

    work page 2024

  42. [42]

    Stock intelligent investment strategy based on support vector machine parameter optimization algorithm,

    X. Li and Y . Sun, “Stock intelligent investment strategy based on support vector machine parameter optimization algorithm,” Neural Computing and Applications, vol. 32, no. 6, pp. 1765–1775, 2020

    work page 2020

  43. [43]

    Parameter selection of gaussian kernel for one-class svm,

    Y . Xiao, H. Wang, and W. Xu, “Parameter selection of gaussian kernel for one-class svm,” IEEE transactions on cybernetics, vol. 45, no. 5, pp. 941–953, 2014

    work page 2014

  44. [44]

    Smo-style algorithms for learning using privileged information

    D. Pechyony, R. Izmailov, A. Vashist, and V . Vapnik, “Smo-style algorithms for learning using privileged information.” Dmin, vol. 10, pp. 235–241, 2010

    work page 2010

  45. [45]

    GPT-4 Technical Report

    J. Achiam, S. Adler, S. Agarwal, L. Ahmad, I. Akkaya, F. L. Aleman, D. Almeida, J. Altenschmidt, S. Altman, S. Anadkat et al. , “Gpt-4 technical report,” arXiv preprint arXiv:2303.08774 , 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  46. [46]

    Behavioral correlates of individual differences in road-traffic crash risk: An examination of methods and findings

    J. Elander, R. West, and D. French, “Behavioral correlates of individual differences in road-traffic crash risk: An examination of methods and findings.” Psychological bulletin, vol. 113, no. 2, p. 279, 1993

    work page 1993

  47. [47]

    Comfort in automated driving: An analysis of preferences for different automated driving styles and their dependence on personality traits,

    H. Bellem, B. Thiel, M. Schrauf, and J. F. Krems, “Comfort in automated driving: An analysis of preferences for different automated driving styles and their dependence on personality traits,” Transportation research part F: traffic psychology and behaviour , vol. 55, pp. 90–100, 2018

    work page 2018

  48. [48]

    Effects of multisensory distractor interference on attentional driving,

    C.-T. Lin, Y . Tian, Y .-K. Wang, T.-T. N. Do, Y .-L. Chang, J.-T. King, K.-C. Huang, and L.-D. Liao, “Effects of multisensory distractor interference on attentional driving,” IEEE transactions on intelligent transportation systems, vol. 23, no. 8, pp. 10 395–10 403, 2022

    work page 2022

  49. [49]

    Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks

    N. Reimers, “Sentence-bert: Sentence embeddings using siamese bert- networks,” arXiv preprint arXiv:1908.10084 , 2019

    work page internal anchor Pith review Pith/arXiv arXiv 1908

  50. [50]

    Efficient Estimation of Word Representations in Vector Space

    T. Mikolov, “Efficient estimation of word representations in vector space,” arXiv preprint arXiv:1301.3781 , vol. 3781, 2013

    work page internal anchor Pith review Pith/arXiv arXiv 2013

  51. [51]

    Glove: Global vectors for word representation,

    J. Pennington, R. Socher, and C. D. Manning, “Glove: Global vectors for word representation,” in Proceedings of the 2014 conference on empirical methods in natural language processing (EMNLP) , 2014, pp. 1532–1543

    work page 2014

  52. [52]

    Uniform manifold approximation and pro- jection,

    J. Healy and L. McInnes, “Uniform manifold approximation and pro- jection,” Nature Reviews Methods Primers , vol. 4, no. 1, p. 82, 2024

    work page 2024

  53. [53]

    The highd dataset: A drone dataset of naturalistic vehicle trajectories on german highways for validation of highly automated driving systems,

    R. Krajewski, J. Bock, L. Kloeker, and L. Eckstein, “The highd dataset: A drone dataset of naturalistic vehicle trajectories on german highways for validation of highly automated driving systems,” in 2018 21st International Conference on Intelligent Transportation Systems (ITSC) . IEEE, 2018, pp. 2118–2125

    work page 2018

  54. [54]

    Follownet: a comprehensive benchmark for car-following behavior modeling,

    X. Chen, M. Zhu, K. Chen, P. Wang, H. Lu, H. Zhong, X. Han, X. Wang, and Y . Wang, “Follownet: a comprehensive benchmark for car-following behavior modeling,” Scientific data, vol. 10, no. 1, p. 828, 2023

    work page 2023

  55. [55]

    Human-like autonomous car-following model with deep reinforcement learning,

    M. Zhu, X. Wang, and Y . Wang, “Human-like autonomous car-following model with deep reinforcement learning,” Transportation research part C: emerging technologies , vol. 97, pp. 348–368, 2018

    work page 2018

  56. [56]

    Captur- ing car-following behaviors by deep learning,

    X. Wang, R. Jiang, L. Li, Y . Lin, X. Zheng, and F.-Y . Wang, “Captur- ing car-following behaviors by deep learning,” IEEE Transactions on Intelligent Transportation Systems, vol. 19, no. 3, pp. 910–920, 2017

    work page 2017

  57. [57]

    Segmentation and clustering of car-following behavior: Recognition of driving patterns,

    B. Higgs and M. Abbas, “Segmentation and clustering of car-following behavior: Recognition of driving patterns,” IEEE Transactions on Intel- ligent Transportation Systems, vol. 16, no. 1, pp. 81–90, 2014

    work page 2014

  58. [58]

    An ensemble deep learning approach for driver lane change intention inference,

    Y . Xing, C. Lv, H. Wang, D. Cao, and E. Velenis, “An ensemble deep learning approach for driver lane change intention inference,” Transportation Research Part C: Emerging Technologies , vol. 115, p. 102615, 2020

    work page 2020

  59. [59]

    Prediction of lane- changing maneuvers with automatic labeling and deep learning,

    V . Mahajan, C. Katrakazas, and C. Antoniou, “Prediction of lane- changing maneuvers with automatic labeling and deep learning,” Trans- portation research record, vol. 2674, no. 7, pp. 336–347, 2020. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 12

    work page 2020

  60. [60]

    Interaction- aware cut-in behavior prediction and risk assessment for autonomous driving,

    J. Zhang, G. Li, Z. Deng, H. Yu, J. P. Huissoon, and D. Cao, “Interaction- aware cut-in behavior prediction and risk assessment for autonomous driving,” IFAC-PapersOnLine, vol. 53, no. 5, pp. 656–663, 2020

    work page 2020

  61. [61]

    Interaction-awareness based inten- tion inference of lag vehicle in lane changing decision-making process for autonomous driving,

    G. Yan, H. Yu, C. Zhang, and J. Xi, “Interaction-awareness based inten- tion inference of lag vehicle in lane changing decision-making process for autonomous driving,” in 2023 IEEE 6th International Conference on Industrial Cyber-Physical Systems (ICPS) . IEEE, 2023, pp. 1–8

    work page 2023

  62. [62]

    A time-efficient approach for decision-making style recognition in lane-changing behavior,

    S. Yang, W. Wang, C. Lu, J. Gong, and J. Xi, “A time-efficient approach for decision-making style recognition in lane-changing behavior,” IEEE Transactions on Human-Machine Systems , vol. 49, no. 6, pp. 579–588, 2019

    work page 2019

  63. [63]

    Overtaking behaviour of vehicles on undivided roads in non-lane based mixed traffic conditions,

    G. Asaithambi and G. Shravani, “Overtaking behaviour of vehicles on undivided roads in non-lane based mixed traffic conditions,” Journal of traffic and transportation engineering (English edition) , vol. 4, no. 3, pp. 252–261, 2017

    work page 2017

  64. [64]

    A comparison of headway and time to collision as safety indicators,

    K. V ogel, “A comparison of headway and time to collision as safety indicators,” Accident analysis & prevention , vol. 35, no. 3, pp. 427– 433, 2003

    work page 2003

  65. [65]

    Dbscan revisited, revisited: why and how you should (still) use dbscan,

    E. Schubert, J. Sander, M. Ester, H. P. Kriegel, and X. Xu, “Dbscan revisited, revisited: why and how you should (still) use dbscan,” ACM Transactions on Database Systems (TODS) , vol. 42, no. 3, pp. 1–21, 2017. Zhaokun Chen received the B.S. degree in In- telligent Automotive Engineering from the Harbin Institute of Technology, Weihai, China, in 2022. He...

    work page 2017