The Context of Crash Occurrence: A Complexity-Infused Approach Integrating Semantic, Contextual, and Kinematic Features

Bruce Mehler; Bryan Reimer; Meng Wang; Pnina Gershon; Shannon C. Roberts; Zach Noonan

REVIEW 3 major objections 6 minor 1 cited by

The Context of Crash Occurrence: A Complexity-Infused Approach Integrating Semantic, Contextual, and Kinematic Features

T0 review · 3 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A two-stage framework that fuses semantic, kinematic, and contextual driving features with an encoder-derived 'complexity' representation predicts crash density with 90.15% accuracy, about 2.2 points above the same features without the…

desk verdict Useful LLM-vs-MTurk annotation data, but the headline crash-prediction accuracy is not interpretable until the authors disclose and fix the train/test split. read the letter →

arxiv 2411.17886 v2 pith:QYIWCC5X submitted 2024-11-26 cs.CV cs.LG

Meng Wang , Zach Noonan , Pnina Gershon , Bruce Mehler , Bryan Reimer , Shannon C. Roberts This is my paper · ORCID

classification cs.CVcs.LG

keywords crashpredictionroadwaycomplexitynaturalisticdrivinglargelanguagemodelsfeaturefusiondensitysemanticsegmentationkinematicfeatures

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to show that a driving scene's 'roadway complexity'—the combined effect of what is in the scene, the road context, and how the vehicle is moving—can be captured as a learnable representation and used to predict where crashes are more likely. It proposes a two-stage framework: first an encoder trained to reproduce a human- or LLM-assigned complexity score from semantic, kinematic, and contextual features, then a crash-density classifier that uses both the original features and the encoder's hidden features. On 10,407 frames sampled every 20 meters from 500 naturalistic driving clips, the full model reaches 90.15% accuracy versus 87.98% with only the original features. The paper also claims that complexity scores generated by a large language model support better crash prediction than crowdworker scores, making the annotation step scalable. If these results hold, the approach offers a practical route to real-time crash-risk monitoring from ordinary vehicle sensors and cameras.

What carries the argument

The load-bearing mechanism is the complexity-infused encoder: a fully connected neural network with 32 hidden neurons trained to regress a 0-10 'complexity index' from semantic, kinematic, and contextual features, with the ReLU-activated hidden layer extracted as a new feature vector. This representation is what injects the latent context of roadway complexity into the second-stage crash-density classifier; the paper's ablation shows the classifier benefits from these infused features but not from the plain complexity index itself.

What would settle it

Re-run the experiments with a clip-level train/test split—putting all frames from each video clip in only one side—and compare test accuracy; if the 90.15% drops substantially, the frame-level split inflated the result.

Watch

Extended reading notes

Core claim

The central claim is that hidden context matters: predicting crash density from raw scene and driving features is improved when a neural encoder first compresses those features into a 'complexity-infused' representation guided by an explicit complexity index. In the paper's design, 17 semantic segmentation features, 9 CAN-bus kinematic features, and 19 LLM-derived contextual features are normalized and fed into a fully connected encoder with 32 hidden neurons whose output is trained to match a 0-10 complexity rating; the post-ReLU hidden layer becomes the infused feature set. The downstream classifier is a Random Forest whose inputs are the original features plus these hidden features, and it yields the best reported test accuracy of 90.15%, compared with 87.98% for the same classifier on original features alone. Ablations show that all three feature families contribute, that LLM-generated complexity annotations outperform crowd annotations for this downstream task, and that the encoder's hidden representation—rather than the raw complexity index—is what drives the gain.

Load-bearing premise

The headline accuracy rests on treating every 20-meter frame as an independent sample, even though frames come from only 500 video clips and share road context and the same crash-density label; grouping frames by clip for the train/test split could change the number.

Editorial extensions

If this is right

Adding complexity-infused features to all three original feature families raises Random Forest test accuracy from 87.98% to 90.15%, a gain the paper reports as statistically significant.
The best result requires all three feature families together, indicating that semantic scene content, vehicle kinematics, and road context carry complementary crash-risk information.
LLM-generated complexity labels outperform crowd-sourced labels when used to train the encoder, so an automated annotation pipeline can replace or augment human labeling without losing predictive accuracy.
The complexity-infused features only improve predictions when combined with original features; used alone they underperform, so the hidden representation is complementary rather than a substitute.

Reading between the lines

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

A grouped train/test split at the video-clip level would test whether the 90.15% accuracy survives the removal of spatial-temporal leakage; this is a direct extension the paper does not report.
The same encoder-plus-auxiliary-label design could inject other scarce judgments—perceived risk, driver workload, near-miss severity—into crash models.
The pattern linking lower speeds to high-density areas suggests the model may partly encode congestion; separating congestion from collision risk would sharpen the interpretation.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Two things to know. First, the paper has a genuinely useful new comparison: LLM-generated complexity ratings versus MTurk ratings for the same driving scenes, used as training targets for a small encoder. Second, the central accuracy claim — 90.15% with complexity-infused features versus 87.98% without — is not yet interpretable, because the paper never says whether the 70/30 split is at the clip level or the frame level. With 10,407 frames drawn from only 500 clips at one frame per 20 meters, a frame-level split guarantees that clips appear in both train and test. Adjacent frames share the same roadway, lighting, vehicle kinematics, and the same KDE-smoothed crash-density label over a 1000 m radius. The model can exploit clip-specific cues instead of learning generalizable features. The reader's stress-test calculation is sound: under a random frame split, the probability that a clip has no test frame is about 1%. The paper contains no clip-level evaluation, no code, and no statement about grouping. So the headline number, and the LLM/MTurk comparison in Table 4, are not reliable as reported. This is the load-bearing problem, not a quibble.

What is actually new and worth credit: the integration of semantic segmentation counts, CAN-bus kinematics, and LLM-derived contextual descriptions into one crash-density classifier, with the complexity index used as an auxiliary target for a hidden layer. The encoder-plus-hidden-features recipe is standard, but the specific combination and the annotation-source comparison are not present in the cited literature. The ablation structure is clear, and the SHAP analysis gives a readable picture of which features drive the classes. The authors also note the data are naturalistic and balanced enough across the three density classes.

Minor issues: no confidence intervals anywhere; McNemar's test is mentioned but not described in a way that accounts for correlated test frames; and the free parameters (KDE radius, density thresholds, hidden units) are reasonable but unexamined. On circularity, the complexity index is externally annotated, so the encoder is not directly predicting crash density; the main worry is that LLM ratings may correlate with visual features that also predict crashes, which is a confound, not a fatal one.

Who this is for: researchers working on crash-risk prediction from naturalistic driving data and on replacing crowdworkers with LLMs for scene annotation. It deserves a serious referee, but the referee should demand a clip-level split or grouped cross-validation, per-clip metrics, and confidence intervals before the empirical claims can be evaluated. If the gain survives clip-level evaluation, this is a solid incremental contribution.

Referee Report

3 major / 6 minor

Summary. This paper proposes a two-stage framework for predicting crash-density levels from fused semantic (OneFormer), kinematic (CAN bus), and contextual (GPT-generated) features. In the first stage, a fully connected encoder is trained to predict an externally annotated roadway-complexity index; its hidden-layer outputs are then used as additional 'complexity-infused' features for a downstream crash-density classifier. On a test split of 10,407 frames drawn from 500 MIT-AVT video clips, the authors report 87.98% accuracy with original features and 90.15% after adding complexity-infused features. They also report that LLM-generated complexity annotations lead to better downstream accuracy than MTurk annotations, and they provide ablations over feature sets and encoder variants.

Significance. If the reported numbers are unbiased, the paper makes a useful empirical contribution by combining scene semantics, vehicle kinematics, and contextual descriptors for crash-density prediction, and by demonstrating that LLM-generated complexity annotations can substitute for crowdworkers. The two-stage feature-infusion idea is interesting, and the appendix provides unusually detailed prompts, model architectures, and training configurations. However, the central claims rest on a train/test split that is not shown to preserve clip identity, and the paper does not establish that the encoder's hidden features were generated without seeing the test frames. These issues directly affect the headline 90.15% versus 87.98% result and the LLM-versus-MTurk comparison, so the empirical contribution is not yet established.

major comments (3)

[Sections 3.1 and 3.4.2] The frame-level independence assumption is violated. Section 3.1 states that 10,407 frames were extracted from 500 clips at one frame per 20 meters, giving roughly 20.8 frames per clip. Section 3.4.2 says only that 'the dataset was split into 70% for training and 30% for testing' and does not state that clips were kept intact. Under a random frame split, a clip with 20 frames appears in both training and test with probability about 1 - 0.7^20 ≈ 0.999. Adjacent frames from the same clip share roadway, lighting, vehicle kinematics, and a crash-density label computed from a 1000 m KDE, so they are highly correlated. The model can exploit clip-level cues rather than learning generalizable complexity representations, inflating the reported accuracy and the differences in Tables 2 and 4. Please repeat the entire evaluation with a clip-stratified split or clip-level cross-validation and report the results, and explicitly state whether every frame of a clip was assigned to the same side of the split.
[Section 3.4.2] It is unclear whether the complexity-infused features for the test frames were generated by an encoder trained only on the training split. The text says 'After generating the complexity-infused features from the encoder, they were used to predict the level of crash density' and then describes splitting the dataset, which suggests the features may have been computed before the split. If the encoder is trained on all frames and then its hidden features are extracted for test frames, test information leaks into the downstream prediction model through the features. Please clarify the order of operations, and if the encoder was trained on the full data, retrain it within the training split only and regenerate all reported numbers.
[Section 4.2] The statistical significance claim is not supported by reported evidence. The paper states that McNemar's test was conducted and that improvements were statistically significant, but no test statistics, p-values, or details of the comparison procedure are provided. Moreover, standard McNemar assumes independent observations, which is not satisfied by the frame-level data. Please report the test details or use a cluster-robust permutation test at the clip level, and state which model pairs were compared.

minor comments (6)

[Section 3.1] The listed scenario counts do not sum to 500: 100 highway + 100 rural + 100 urban + 75 bridge + 75 overpass + 75 crash hotspot = 525. Please correct the totals or the category counts.
[Section 3.3.2] The KDE radius (1000 m) and the crash-density thresholds (0.5 and 2) appear to be chosen based on the observed distribution, but the paper does not state whether these were fixed a priori or selected from the data. Please clarify and, if possible, report sensitivity of the main results to these choices.
[Tables 1-5] All results come from a single 70/30 split, so no measure of variance is available. Reporting confidence intervals or repeated-split results would help the reader assess whether the differences between models are stable.
[Appendix 7] The GPT-based feature generation uses three runs per image with majority voting, but the paper does not report how often the three runs disagreed. This information would help assess the reliability of the contextual features and the complexity index.
[Equation (3)] The kernel density formula would be clearer if the bandwidth were defined explicitly; currently 'radius' serves as both the truncation distance and the bandwidth, and the notation '(radius)^2' is easy to misread.
[Throughout] There are several typographical issues, including '1 .10' and '0.84 0 .86' in Table 1 and the heading 'V ehicle Kinematic Features' in Section 3.2.2. A careful proofread is needed.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the complexity index is externally annotated and the crash-density labels are independently derived from historical crash KDE, so the encoder's hidden features are auxiliary-task representations rather than restatements of the target.

full rationale

The derivation chain is not circular. The encoder is trained to predict an externally collected complexity index (GPT-generated or MTurk-averaged ratings, Section 3.3.1); the crash-density ground truth is a KDE over historical crash GPS data with a 1000 m radius (Section 3.3.2). Neither target is defined in terms of the other, and the complexity-infused features are hidden-layer activations produced while regressing onto the complexity index (Section 3.4.1). Using those activations as additional predictors of crash density is an auxiliary-task/transfer representation, not a fitted-input-called-prediction scheme: no parameter of the crash model is fit to the crash labels and then evaluated on those same labels. The only self-citations are to the MIT-AVT/Ding et al. dataset [5] and the lead-car-region definition [27]; both are ordinary external inputs and do not carry the paper's predictive claim. The frame-level 70/30 split described in Section 3.4.2 is a potential evaluation-leakage threat because frames are sampled every 20 m from 500 clips, but this is a data-splitting validity issue, not a circularity: it does not make the crash prediction equivalent to the model's own inputs by construction. Therefore the circularity score is 0.

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

The central claim depends on several numbers chosen by the authors: the KDE radius, the crash-density class boundaries, the encoder size, and the feature-variability cutoff. These are fitted to the data rather than derived from theory. No new physical entities are introduced; 'complexity-infused features' are a learned representation, not a postulated object.

free parameters (5)

KDE radius = 1000 m
Chosen for the crash-density heatmap; no sensitivity analysis is provided.
Crash density thresholds = Low <0.5, Medium 0.5-2, High 2-10
Defined post hoc from the skewed distribution in Fig. 5; they directly define the classification task.
Encoder hidden units = 32 (16 also tested)
Selected by validation performance; the hidden features used downstream depend on this choice.
Semantic feature variability cutoff = 90% zeros
Features with more than 90% zero values were removed, reducing 50 to 17 features; this affects all downstream models.
Learning rates and epochs = 0.0003-0.005, 1000-2000
Hyperparameters chosen during training; no sensitivity analysis is reported.

assumptions (4)

domain assumption Historical crash GPS locations aggregated with a 1000 m KDE provide a valid per-location proxy for crash likelihood.
The ground truth is a spatial density of past crashes, not a causal or predictive risk measure for the specific driving moment; Section 3.3.2.
domain assumption Frames sampled every 20 meters from the same video clip can be treated as independent samples for training and evaluation.
The split in Section 3.4.2 is described at the frame level; if not grouped by clip, temporal and spatial autocorrelation invalidates the accuracy estimate.
domain assumption The LLM-generated complexity index is a meaningful label that can be predicted from the input features.
The encoder is trained against this annotation in Section 3.4.1; if the rating is arbitrary, the complexity-infused features are noise.
domain assumption OneFormer panoptic segmentation outputs and GPT-4o contextual answers are accurate enough for feature generation.
Semantic and contextual features are generated by third-party models without ground-truth verification in Section 3.2.

how reviews work

0 comments

Cite this review

Pith. "Pith review of The Context of Crash Occurrence: A Complexity-Infused Approach Integrating Semantic, Contextual, and Kinematic Features." pith.science (2026). https://pith.science/paper/QYIWCC5X

@misc{pith2026241117886,
  author       = {Pith},
  title        = {Pith review of: The Context of Crash Occurrence: A Complexity-Infused Approach Integrating Semantic, Contextual, and Kinematic Features},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QYIWCC5X}},
  note         = {Machine review of arXiv:2411.17886}
}

read the original abstract

Understanding the context of crash occurrence in complex driving environments is essential for improving traffic safety and advancing automated driving. Previous studies have used statistical models and deep learning to predict crashes based on semantic, contextual, or vehicle kinematic features, but none have examined the combined influence of these factors. In this study, we term the integration of these features ``roadway complexity''. This paper introduces a two-stage framework that integrates roadway complexity features for crash prediction. In the first stage, an encoder extracts hidden contextual information from these features, generating complexity-infused features. The second stage uses both original and complexity-infused features to predict crash likelihood, achieving an accuracy of 87.98\% with original features alone and 90.15\% with the added complexity-infused features. Ablation studies confirm that a combination of semantic, kinematic, and contextual features yields the best results, which emphasize their role in capturing roadway complexity. Additionally, complexity index annotations generated by the Large Language Model outperform those by Amazon Mechanical Turk, highlighting the potential of AI-based tools for accurate, scalable crash prediction systems.

Figures

Figures reproduced from arXiv: 2411.17886 by the authors.

**Figure 1.** The model structure. The model takes raw images and CAN-Bus signals as inputs to generate semantic, contextual, and kinematic [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗

**Figure 2.** The raw roadway scene image and OneFormer algorithm [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** The prompt used in collecting contextual features with [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Figures from the paper (4 more)

**Figure 4.** Figure 4: Crash density heatmap (2018-2022) in Massachusetts, displayed in red, where darker colors indicate a higher crash density. Five [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

**Figure 5.** Figure 5: The distribution of crash density value. [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]

**Figure 6.** Figure 6: The SHAP values of the 10 most influential features for each class of the best-performing model. [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]

**Figure 7.** Figure 7: Visualization of the crash density prediction neural net [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

Simulating the Unseen: Crash Prediction Must Learn from What Did Not Happen
cs.LG 2025-05 conditional novelty 4.0 of 10

Crash prediction should learn from near-miss events and synthetic counterfactual scenarios, not just recorded crashes.

Reference graph

Works this paper leans on

33 extracted references · 28 canonical work pages · cited by 1 Pith paper

[1]

Gpt-4 technical report

Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ah- mad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, et al. Gpt-4 technical report. arXiv preprint arXiv:2303.08774 ,

arXiv
[2]

What characterizes” situations” in situation awareness? findings from a human- centered investigation

Philipp Michael Markus Peter Asteriou, Heike Christiane Kotsios, and Philipp Wintersberger. What characterizes” situations” in situation awareness? findings from a human- centered investigation. In Proceedings of the 16th Interna- tional Conference on Automotive User Interfaces and Inter- active Vehicular Applications, pages 216–226, 2024. 1

work page 2024
[3]

The cityscapes dataset for semantic urban scene understanding

Marius Cordts, Mohamed Omran, Sebastian Ramos, Timo Rehfeld, Markus Enzweiler, Rodrigo Benenson, Uwe Franke, Stefan Roth, and Bernt Schiele. The cityscapes dataset for semantic urban scene understanding. In Proc. of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016. 3

work page 2016
[4]

Drivellm: Charting the path toward full au- tonomous driving with large language models

Yaodong Cui, Shucheng Huang, Jiaming Zhong, Zhenan Liu, Yutong Wang, Chen Sun, Bai Li, Xiao Wang, and Amir Khajepour. Drivellm: Charting the path toward full au- tonomous driving with large language models. IEEE Trans- actions on Intelligent Vehicles, 2023. 3

work page 2023
[5]

Mit-avt clustered driving scene dataset: Evaluating perception systems in real-world natu- ralistic driving scenarios

Li Ding, Michael Glazer, Meng Wang, Bruce Mehler, Bryan Reimer, and Lex Fridman. Mit-avt clustered driving scene dataset: Evaluating perception systems in real-world natu- ralistic driving scenarios. In 2020 IEEE Intelligent Vehicles Symposium (IV), pages 232–237. IEEE, 2020. 3 9

work page 2020
[6]

Mit ad- vanced vehicle technology study: Large-scale naturalistic driving study of driver behavior and interaction with automa- tion

Lex Fridman, Daniel E Brown, Michael Glazer, William An- gell, Spencer Dodd, Benedikt Jenik, Jack Terwilliger, Alek- sandr Patsekin, Julia Kindelsberger, Li Ding, et al. Mit ad- vanced vehicle technology study: Large-scale naturalistic driving study of driver behavior and interaction with automa- tion. IEEE Access, 7:102021–102038, 2019. 3

work page 2019
[7]

Environmental factors affecting the frequency of road traffic accidents: a case study of sub-urban area of pakistan

Hafiz Mohkum Hammad, Muhammad Ashraf, Farhat Abbas, Hafiz Faiq Bakhat, Saeed A Qaisrani, Muhammad Mubeen, Shah Fahad, and Muhammad Awais. Environmental factors affecting the frequency of road traffic accidents: a case study of sub-urban area of pakistan. Environmental Science and Pollution Research, 26:11674–11685, 2019. 1

work page 2019
[8]

The effect of traffic on situation aware- ness and mental workload: Simulator-based study

Xueqin Hao, Zhiguo Wang, Fan Yang, Ying Wang, Yanru Guo, and Kan Zhang. The effect of traffic on situation aware- ness and mental workload: Simulator-based study. In Engi- neering Psychology and Cognitive Ergonomics: 7th Interna- tional Conference, EPCE 2007, Held as Part of HCI Interna- tional 2007, Beijing, China, July 22-27, 2007. Proceedings 7, pages ...

work page 2007

Show all 33 references

[9]

An image-based crash risk prediction model using visual attention mapping and a deep convolutional neural network

Chengyu Hu, Wenchen Yang, Chenglong Liu, Rui Fang, Zhongyin Guo, and Bijiang Tian. An image-based crash risk prediction model using visual attention mapping and a deep convolutional neural network. Journal of Transporta- tion Safety & Security, 15(1):1–23, 2023. 1

2023
[10]

Oneformer: One transformer to rule universal image segmentation

Jitesh Jain, Jiachen Li, Mang Tik Chiu, Ali Hassani, Nikita Orlov, and Humphrey Shi. Oneformer: One transformer to rule universal image segmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2989–2998, 2023. 3

2023
[11]

Effects of hazard exposure and roadway com- plexity on young and older driver situation awareness and performance

David Kaber, Yu Zhang, Sangeun Jin, Prithima Mosaly, and Megan Garner. Effects of hazard exposure and roadway com- plexity on young and older driver situation awareness and performance. Transportation research part F: traffic psy- chology and behaviour, 15(5):600–611, 2012. 1

2012
[12]

A system of vision sensor based deep neu- ral networks for complex driving scene analysis in support of crash risk assessment and prevention

Muhammad Monjurul Karim, Yu Li, Ruwen Qin, and Zhaozheng Yin. A system of vision sensor based deep neu- ral networks for complex driving scene analysis in support of crash risk assessment and prevention. arXiv preprint arXiv:2106.10319, 2021. 1

2021 arXiv
[13]

Evaluating the nature of distractive driving factors towards road traffic accident

Khawar Khan, Syed Bilal Zaidi, and Asad Ali. Evaluating the nature of distractive driving factors towards road traffic accident. Civil Engineering Journal, 6(8):1555–1580, 2020. 1

2020
[14]

Panoptic segmentation

Alexander Kirillov, Kaiming He, Ross Girshick, Carsten Rother, and Piotr Dollar. Panoptic segmentation. In Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019. 3

2019
[15]

Lundberg, Gabriel Erion, Hugh Chen, Alex De- Grave, Jordan M

Scott M. Lundberg, Gabriel Erion, Hugh Chen, Alex De- Grave, Jordan M. Prutkin, Bala Nair, Ronit Katz, Jonathan Himmelfarb, Nisha Bansal, and Su-In Lee. From local expla- nations to global understanding with explainable ai for trees. Nature Machine Intelligence, 2(1):2522–5839...

2020
[16]

Note on the sampling error of the differ- ence between correlated proportions or percentages

Quinn McNemar. Note on the sampling error of the differ- ence between correlated proportions or percentages. Psy- chometrika, 12(2):153–157, 1947. 7

1947
[17]

Urban street structure and traffic safety.Journal of safety research, 62:63–71, 2017

Dinesh Mohan, Shrikant I Bangdiwala, and Andres Villave- ces. Urban street structure and traffic safety.Journal of safety research, 62:63–71, 2017. 1

2017
[18]

Most: Multi-modality scene to- kenization for motion prediction

Norman Mu, Jingwei Ji, Zhenpei Yang, Nate Harada, Hao- tian Tang, Kan Chen, Charles R Qi, Runzhou Ge, Kratarth Goel, Zoey Yang, et al. Most: Multi-modality scene to- kenization for motion prediction. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Reco...

2024
[19]

Impact of road infrastructure land use and traffic operational characteristics on pedestrian fatality risk: A case study of kolkata, india

Dipanjan Mukherjee and Sudeshna Mitra. Impact of road infrastructure land use and traffic operational characteristics on pedestrian fatality risk: A case study of kolkata, india. Transportation in Developing Economies, 5(2):6, 2019. 1

2019
[20]

The impact of environmental complexity on drivers’ situation awareness

Sami Park, Yilun Xing, Kumar Akash, Teruhisa Misu, and Linda Ng Boyle. The impact of environmental complexity on drivers’ situation awareness. In Proceedings of the 14th International Conference on Automotive User Interfaces and Interactive Vehicular Applications, pages 131–13...

2022
[21]

Modeling the accuracy of traffic crash prediction models

Mohammad Hesam Rashidi, Soheil Keshavarz, Parham Pazari, Navid Safahieh, and Amir Samimi. Modeling the accuracy of traffic crash prediction models. IATSS research, 46(3):345–352, 2022. 1

2022
[22]

Critical reasons for crashes investigated in the national motor vehicle crash causation survey

Santokh Singh. Critical reasons for crashes investigated in the national motor vehicle crash causation survey. Technical report, U.S. Department of Transportation Technical Report (DOT HS 812 115), 2015. 1

2015
[23]

Con- textvlm: Zero-shot and few-shot context understanding for autonomous driving using vision language models

Shounak Sural, Ragunathan Rajkumar, et al. Con- textvlm: Zero-shot and few-shot context understanding for autonomous driving using vision language models. arXiv preprint arXiv:2409.00301, 2024. 3

2024 arXiv
[24]

The effect of traffic and road characteristics on road safety: A review and future research direction

Chao Wang, Mohammed A Quddus, and Stephen G Ison. The effect of traffic and road characteristics on road safety: A review and future research direction. Safety science, 57: 264–275, 2013. 1

2013
[25]

Identifying crash- prone traffic conditions under different weather on freeways

Chengcheng Xu, Wei Wang, and Pan Liu. Identifying crash- prone traffic conditions under different weather on freeways. Journal of safety research, 46:135–144, 2013

2013
[26]

Associ- ation rule analysis of factors contributing to extraordinarily severe traffic crashes in china.Journal of safety research, 67: 65–75, 2018

Chengcheng Xu, Jie Bao, Chen Wang, and Pan Liu. Associ- ation rule analysis of factors contributing to extraordinarily severe traffic crashes in china.Journal of safety research, 67: 65–75, 2018. 1

2018
[27]

Takeover context matters: Characterising context of takeovers in naturalistic driving using super cruise and au- topilot

Shiyan Yang, Angus McKerral, Megan Dawn Mulhall, Michael Graeme Lenn ´e, Bryan Reimer, and Pnina Ger- shon. Takeover context matters: Characterising context of takeovers in naturalistic driving using super cruise and au- topilot. In Proceedings of the 15th International Confer...

2023
[28]

High- wayllm: Decision-making and navigation in highway driv- ing with rl-informed language model

Mustafa Yildirim, Barkin Dagda, and Saber Fallah. High- wayllm: Decision-making and navigation in highway driv- ing with rl-informed language model. arXiv preprint arXiv:2405.13547, 2024. 3

2024 arXiv
[29]

Improving driver situation awareness pre- diction using human visual sensory and memory mechanism

Haibei Zhu, Teruhisa Misu, Sujitha Martin, Xingwei Wu, and Kumar Akash. Improving driver situation awareness pre- diction using human visual sensory and memory mechanism. In 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 6210–6216. IEEE,...

2021
[30]

3.2.3 and Sec

LLM Implementation Details The contextual features and the complexity index feature mentioned in Sec. 3.2.3 and Sec. 3.3.1 were generated us- ing the GPT-4o-2024-08-06 model 4. Batch API 5 was used for faster processing

2024
[31]

Annotation Comparisons Table 6 presents the descriptive statistics for the compari- son between AI-generated and human-generated annotated complexity indices, as discussed in Sec. 3.3.1. This compar- ison is based on the entire dataset without matching. It can be observed that...
[32]

3.4.2 is shown in Fig

Model Structure The neural network model structure mentioned in Sec. 3.4.2 is shown in Fig. 7
[33]

Encoder Training Details Regarding the training configurations of the encoders men- tioned in Sec

Model Training Details 10.1. Encoder Training Details Regarding the training configurations of the encoders men- tioned in Sec. 4.1, different setups were applied based on the input features and the output data types. The model was trained for 1000 epochs using Stochastic Grad...

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [1]

Gpt-4 technical report

Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ah- mad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, et al. Gpt-4 technical report. arXiv preprint arXiv:2303.08774 ,

arXiv

[2] [2]

What characterizes” situations” in situation awareness? findings from a human- centered investigation

Philipp Michael Markus Peter Asteriou, Heike Christiane Kotsios, and Philipp Wintersberger. What characterizes” situations” in situation awareness? findings from a human- centered investigation. In Proceedings of the 16th Interna- tional Conference on Automotive User Interfaces and Inter- active Vehicular Applications, pages 216–226, 2024. 1

work page 2024

[3] [3]

The cityscapes dataset for semantic urban scene understanding

Marius Cordts, Mohamed Omran, Sebastian Ramos, Timo Rehfeld, Markus Enzweiler, Rodrigo Benenson, Uwe Franke, Stefan Roth, and Bernt Schiele. The cityscapes dataset for semantic urban scene understanding. In Proc. of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016. 3

work page 2016

[4] [4]

Drivellm: Charting the path toward full au- tonomous driving with large language models

Yaodong Cui, Shucheng Huang, Jiaming Zhong, Zhenan Liu, Yutong Wang, Chen Sun, Bai Li, Xiao Wang, and Amir Khajepour. Drivellm: Charting the path toward full au- tonomous driving with large language models. IEEE Trans- actions on Intelligent Vehicles, 2023. 3

work page 2023

[5] [5]

Mit-avt clustered driving scene dataset: Evaluating perception systems in real-world natu- ralistic driving scenarios

Li Ding, Michael Glazer, Meng Wang, Bruce Mehler, Bryan Reimer, and Lex Fridman. Mit-avt clustered driving scene dataset: Evaluating perception systems in real-world natu- ralistic driving scenarios. In 2020 IEEE Intelligent Vehicles Symposium (IV), pages 232–237. IEEE, 2020. 3 9

work page 2020

[6] [6]

Mit ad- vanced vehicle technology study: Large-scale naturalistic driving study of driver behavior and interaction with automa- tion

Lex Fridman, Daniel E Brown, Michael Glazer, William An- gell, Spencer Dodd, Benedikt Jenik, Jack Terwilliger, Alek- sandr Patsekin, Julia Kindelsberger, Li Ding, et al. Mit ad- vanced vehicle technology study: Large-scale naturalistic driving study of driver behavior and interaction with automa- tion. IEEE Access, 7:102021–102038, 2019. 3

work page 2019

[7] [7]

Environmental factors affecting the frequency of road traffic accidents: a case study of sub-urban area of pakistan

Hafiz Mohkum Hammad, Muhammad Ashraf, Farhat Abbas, Hafiz Faiq Bakhat, Saeed A Qaisrani, Muhammad Mubeen, Shah Fahad, and Muhammad Awais. Environmental factors affecting the frequency of road traffic accidents: a case study of sub-urban area of pakistan. Environmental Science and Pollution Research, 26:11674–11685, 2019. 1

work page 2019

[8] [8]

The effect of traffic on situation aware- ness and mental workload: Simulator-based study

Xueqin Hao, Zhiguo Wang, Fan Yang, Ying Wang, Yanru Guo, and Kan Zhang. The effect of traffic on situation aware- ness and mental workload: Simulator-based study. In Engi- neering Psychology and Cognitive Ergonomics: 7th Interna- tional Conference, EPCE 2007, Held as Part of HCI Interna- tional 2007, Beijing, China, July 22-27, 2007. Proceedings 7, pages ...

work page 2007