HMVLM: Multistage Reasoning-Enhanced Vision-Language Model for Long-Tailed Driving Scenarios

Daming Wang; Kangliang Chen; Lu Deng; Weihao Gu; Xing Pan; Yuhao Song; Zijian He

REVIEW 4 major objections 6 minor 1 cited by

HMVLM: Multistage Reasoning-Enhanced Vision-Language Model for Long-Tailed Driving Scenarios

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read A 3B vision-language model with staged reasoning and trajectory smoothing reaches a 7.7367 Rater Feedback Score, taking 2nd place in the 2025 Waymo E2E Driving Challenge.

desk verdict A top-2 Waymo leaderboard score that is unverifiable from the paper: no baseline, no ablations, and a causal claim that needs experimental support. read the letter →

arxiv 2506.05883 v1 pith:GMQUNRXF submitted 2025-06-06 cs.CV cs.AI

Daming Wang , Yuhao Song , Zijian He , Kangliang Chen , Xing Pan , Lu Deng , Weihao Gu This is my paper · ORCID

classification cs.CVcs.AI

keywords HMVLMvision-languagemodelend-to-endautonomousdrivingchain-of-thoughtpromptingWaymoOpenDatasettrajectorysmoothingfast-slowarchitecturelong-tailedscenarios

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

This paper argues that a relatively small vision-language model can act as the deliberative planner in an autonomous-driving stack and handle long-tail scenes if its output is organized as structured reasoning. HMVLM feeds five surround-view images and a four-second history of ego velocity and acceleration into Qwen2.5-VL-3B, then forces the model to answer in the fixed order: scene understanding, driving decision, trajectory inference. A trajectory refiner based on Savitzky-Golay filtering removes jitter and sharp turns while preserving key points and endpoints. On the Waymo Open Dataset, the method scores 7.7367 on the Rater Feedback Score, taking second place in the 2025 Vision-based End-to-End Driving Challenge and beating the public baseline by 2.77%. The paper's point is that prompt structure and output post-processing can improve a VLM planner's generalization to rare, safety-critical events without a larger model.

What carries the argument

The central mechanism is the multi-stage chain-of-thought format: a fixed, special-token-delimited answer structure, <DESC START> ... <DESC END> <DECI START> ... <DECI END> <TRAJ START> ... <TRAJ END>, which compels the VLM to reason in the order Scene Understanding → Driving Decision → Trajectory Inference. This turns token generation into an interpretable chain and aligns training labels with the same structure. It is supported by two secondary mechanisms: selective five-view prompting that injects four seconds of ego velocity and acceleration into the text prompt, and an adaptive Savitzky-Golay trajectory refiner that removes late-stage jitter and sharp turns while preserving endpoints and key geometric features.

What would settle it

Audit a random sample of the auto-generated VQA pairs from Section 2.1 by comparing each stated driving decision with the ground-truth trajectory and navigation input that prompted it, or re-run the trained model on the Waymo validation split with the three-stage CoT tokens replaced by a flat prompt; a substantial mismatch rate or a flat RFS would indicate the claimed reasoning-driven gain is not coming from the stated mechanism.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that a vision-language trajectory planner can be made competitive on long-tail driving by tightening the prompt-design and output-format loop rather than by scaling the model. The core is a three-stage chain-of-thought format with dedicated special tokens that forces the model to emit a scene description, then a natural-language driving decision, then a sequence of BEV waypoints for the trajectory. Around that core, the authors add selective five-view prompting with an embedded four-second ego-kinematics history, and a trajectory-refinement stage that applies adaptive Savitzky-Golay filtering with z-score outlier removal, key-point preservation, and strict endpoint maintenance. The measured outcome is an RFS of 7.7367 on the Waymo E2E benchmark, with the highest category scores in construction (8.6663) and single-lane maneuvers (8.3563) and lower scores in spotlight (6.7269) and cyclist (7.3925) scenes.

Load-bearing premise

The load-bearing premise is that the auto-labeled VQA training examples, generated by an unspecified pre-trained LLM on the Waymo data without human verification or reported quality statistics, are accurate enough to teach the model the mapping from visual scenes to human-like driving decisions.

Editorial extensions

If this is right

HMVLM's ranking shows that a 3B-parameter vision-language planner can reach second place on long-tail driving through imitation learning alone, without reinforcement learning or a modular perception stack.
The three-stage reasoning output makes the planner's decisions inspectable, letting an operator read the scene summary and driving decision before accepting the trajectory.
The trajectory refiner is reported to remove late-stage oscillations and reduce collision events, which is the safety-relevant payoff of the smoothing stage.
Because HMVLM is designed as the slow branch of a fast-slow architecture, its natural deployment path is to pair it with a millisecond-latency fast controller, which the paper names as future work.
The category breakdown indicates that structured reasoning helps most in clearly defined contexts such as construction and single-lane maneuvers, while dynamic interactions like spotlight and cyclist scenes remain the harder tail.

Reading between the lines

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

The paper does not ablate the three upgrades separately, so an untested possibility is that the trajectory refiner, which directly modifies the scored waypoints, accounts for a large share of the 2.77% gain rather than the chain-of-thought format.
Because the auto-labeling pipeline gives the labeler the ground-truth trajectory and navigation input before asking for a driving decision, the VLM may learn to rationalize given trajectories instead of generating new plans; a random human audit of the VQA labels would test this.
A cheap external test of the recipe is to swap Qwen2.5-VL-3B for another open vision-language model and rerun the same five-view, three-stage, smoothing pipeline on the Waymo validation split.
The gap between the RFS ranking and the reported ADE values suggests that human raters may reward interpretable, human-aligned behavior more than raw trajectory accuracy, a trade-off the paper leaves unquantified.

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

Colleague – here's my read on 2506.05883.

The honest summary: this is a competition report, not a complete scientific paper. The headline claim — HMVLM gets RFS 7.74, 2nd place in the Waymo E2E challenge, 2.77% above the public baseline — is the one thing people will quote, and it's unverifiable from the text. The baseline score is never given, no competitor results appear anywhere, and there are no ablations isolating any of the three 'upgrades.' So the causal story ('these upgrades enable...') is not supported by evidence in the paper.

What's genuine: the writing is clear, the fast–slow framing is honest (this is the slow branch only), and the three components — five-view prompting with 4s kinematics, multi-stage CoT with special tokens, and Savitzky-Golay-based trajectory smoothing — are all sensible engineering choices. The paper also reports per-category RFS numbers, which at least let you see where the model is strong (construction, single-lane) and weak (spotlight, cyclists). If the leaderboard result is real, it's a useful data point for the VLM-driving community. The model itself may be fine; the paper just doesn't give us the means to check.

The soft spots are serious. Missing baseline and competitor numbers make the 2.77% arithmetic impossible to verify. No ablation means we can't know whether the improvement comes from the three proposed techniques, from fine-tuning Qwen2.5-VL on Waymo data at all, or from the auto-labeled VQA corpus described in Sec 2.1. That auto-labeling pipeline is itself a concern: an unnamed pre-trained LLM generates the training annotations, with no human verification, no statistics, no quality control. If those labels leak trajectory or navigation information into the captions, the 'scene understanding' stage could be trivial. And the constant-velocity completion for short trajectories (Sec 3.1) is the kind of post-processing that can materially change ADE/RFS; its impact is unreported. No code, no data, no error bars. Single run on a leaderboard.

Who is this for? Someone tracking VLM-based driving systems who wants a quick snapshot of a strong competition result. It doesn't deserve acceptance as a self-contained contribution, but it does deserve a serious referee: a top-2 leaderboard result in this area should get review, with the clear expectation that the authors add the baseline number, ablations, and annotation quality analysis. I'd send it out with those demands.

Referee Report

4 major / 6 minor

Summary. The manuscript describes HMVLM, a vision-language-model-based slow planner for end-to-end driving, built by fine-tuning Qwen2.5-VL-3B on part of the Waymo Open Dataset. Three contributions are claimed: selective five-view prompting with embedded ego-kinematics history, multi-stage chain-of-thought prompting using special tokens, and spline-based trajectory post-processing. The paper reports a Rater Feedback Score (RFS) of 7.7367, second place in the 2025 Waymo Vision-based End-to-End Driving Challenge, and an unspecified 2.77% improvement over the public baseline. The manuscript is a short competition-style report; it provides no ablations, no baseline scores, no error bars, and no annotation-quality statistics.

Significance. If the reported leaderboard result is taken at face value, the paper demonstrates that a 3B-parameter VLM fine-tuned solely on Waymo data can rank competitively on an external autonomous-driving benchmark. The use of an external benchmark is a genuine strength and avoids any circularity in evaluation. The structured-reasoning prompting scheme and the trajectory-refinement idea are plausible and potentially useful to practitioners. However, the scientific contribution is substantially weakened by the absence of controlled experiments: the paper's central claim that the three named upgrades 'enable' the reported performance is not testable from the material presented. No code, model checkpoints, or data are released, so the empirical result itself is not independently reproducible within the manuscript. The paper is best read as a challenge report rather than a mechanistic study.

major comments (4)

[Section 3.2, Table 1] The public baseline score is never stated, and no competitor scores are listed. Consequently, the claimed 2.77% improvement over the baseline and the second-place claim cannot be checked even arithmetically from the manuscript. The authors should report the baseline RFS (overall and per category) alongside the HMVLM values, together with the scores of the other ranked systems or at least the first- and third-place entries.
[Abstract and Sections 2.2-2.3] The central causal claim that selective five-view prompting, multi-stage CoT, and spline smoothing 'enable' the RFS result is unsupported by any ablation. No experiment removes the selective view selection, no variant strips the <DESC>/<DECI>/<TRAJ> structure to a single-stage prompt, and no post-processing run disables the smoothing. Without these variants, the reported RFS could be driven by fine-tuning the base VLM on Waymo data, by the auto-labeled VQA corpus, or by the constant-velocity completion step described in Section 3.1. The authors should add ablations, including a base Qwen2.5-VL-3B model fine-tuned without the proposed upgrades, and report RFS for each variant.
[Section 2.1] The auto-labeled VQA training data are generated by an unspecified pre-trained LLM, and no human verification, annotation statistics, or quality-control measures are reported. Because these annotations are the training signal for the scene-understanding and driving-decision stages, their quality is load-bearing for the claimed reasoning behavior. The authors should identify the labeling model, report the number of samples, and provide a verification protocol or a sample of human-checked annotations.
[Section 3.1 and Section 3.2] The inference procedure is stochastic (temperature 0.01, top-p 0.7, top-k 50), and the trajectories are post-processed with a constant-velocity completion step. Only a single leaderboard submission is reported, so the statement that HMVLM 'consistently maintains a leading position' is unsupported. The authors should report repeated inference runs with variance statistics, or at minimum state explicitly that the leaderboard score is a single run and avoid the word 'consistently'.

minor comments (6)

[Abstract] The abstract contains typographical spacing errors: '4 shistory' should be '4s history', and 'aRater Feedback Score' should be 'a Rater Feedback Score'.
[Section 1] There are formatting and wording issues: 'V AD[4]' should be 'VAD[4]', and the sentence 'while Senna pairs a Senna[5] planner' repeats 'Senna'.
[Section 1] The phrase 'fewer than < 0.003%' is redundant; 'occur in fewer than 0.003% of daily driving' would be sufficient.
[Section 2.3] The title and abstract call the post-processing 'spline-based', but the method description in Section 2.3 says 'adaptive Savitzky-Golay filtering', which is polynomial smoothing rather than spline fitting. The terminology should be aligned.
[Section 2.1] The 'systematic comparison' of open-source VLMs that is said to justify the choice of Qwen2.5-VL-3B is mentioned but no results are reported. A small table or a citation with the comparison results would make the selection verifiable.
[Section 3.2] The phrase 'Although its ADE scores are slightly higher compared to some competitors' refers to competitor ADE values that are not reported in the table; either add those values or remove the comparison.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the central RFS claim is measured by the external Waymo challenge benchmark and is not derived from the paper's own fitted inputs.

full rationale

The paper's central claim, a Rater Feedback Score of 7.7367 with second place in the 2025 Waymo Vision-based End-to-End Driving Challenge, is an externally evaluated leaderboard result, not a quantity derived from the paper's assumptions or fitted parameters. The described upgrades (selective five-view prompting, multi-stage CoT, and spline-based trajectory smoothing) are architectural and prompting components of a trained system; the paper never reduces any metric to these components by construction, and no equation in the paper defines RFS in terms of the model's own outputs. The auto-labeled VQA training data described in Section 2.1 are training-time supervision, and the challenge evaluation is held out and independent of that labeling process. Citations to prior work, including CoT prompting and the Qwen2.5-VL backbone, support standard techniques rather than carrying the paper's conclusion. The absence of ablations isolating individual upgrades is a scientific evidence limitation, not circularity, because it does not show that the claimed result is equivalent to its inputs by definition. Under the hard rules requiring a quotable reduction, no circular step is present.

Assumptions & free parameters 7 free parameters · 5 assumptions · 0 invented entities

The central claim depends on a set of undocumented design choices and domain assumptions. The most consequential are the reliance on auto-annotated supervision without quality control, the post-hoc trajectory completion and smoothing whose parameters are undisclosed, and the absence of a baseline in the comparison. No new physical or conceptual entities are introduced.

free parameters (7)

Learning rate = 2e-5
Chosen for fine-tuning Qwen2.5-VL-3B; not derived from theory, and no sensitivity analysis is provided (Section 3.1).
Inference sampling parameters (top-p, top-k, temperature) = 0.7, 50, 0.01
Hand-picked settings for VLLM inference; no ablation on their effect on trajectory quality (Section 3.1).
Savitzky-Golay window size = adaptive, unspecified
The paper states the window size is adaptively selected but never specifies the range or criterion, directly affecting the smoothed output (Section 2.3).
Directional-change threshold for keypoint preservation = 25 degrees
Used to decide which geometric features to preserve during smoothing; no justification or alternative thresholds are tested (Section 2.3).
Z-score threshold for outlier removal = unspecified
Outlier removal uses z-score thresholding but the threshold value is not reported (Section 2.3).
Ego kinematics history length = 4 seconds
The 4 s history of velocity and acceleration is embedded in the prompt; no comparison to other history lengths is provided (Section 2.1).
Number of camera views = 5 views (3 front, 2 side)
The choice of five specified views is presented as a contribution, but no comparison to using all views or fewer views is shown (Section 2.1).

assumptions (5)

standard math Savitzky-Golay filtering assumes the underlying trajectory is locally smooth and polynomial in segments; z-score outlier removal assumes approximate normality of errors.
Invoked in Section 2.3 without stating the conditions under which these assumptions hold for driving trajectories.
domain assumption The fast-slow architecture and the use of a VLM as a slow planner improves driving in long-tail scenarios.
Motivated from prior work (DriveVLM-Dual, Senna) but not tested by a control condition in this paper.
domain assumption The auto-annotations produced by a pre-trained LLM on the Waymo dataset are accurate enough to serve as supervision for driving decisions.
Section 2.1 states a pre-trained LLM constructs VQA tasks, but no human verification, annotation accuracy statistics, or quality control is reported.
domain assumption The Rater Feedback Score (RFS) is a valid proxy for real-world driving safety and is more meaningful than ADE.
Section 3.2 argues RFS captures long-tail generalization, but the paper provides no evidence linking RFS to actual safety outcomes.
ad hoc to paper The three-stage CoT format with special tokens <DESC>, <DECI>, <TRAJ> is an effective decomposition for driving.
No comparison to alternative prompt structures, stage orderings, or output formats is provided.

how reviews work

0 comments

Cite this review

Pith. "Pith review of HMVLM: Multistage Reasoning-Enhanced Vision-Language Model for Long-Tailed Driving Scenarios." pith.science (2026). https://pith.science/paper/GMQUNRXF

@misc{pith2026250605883,
  author       = {Pith},
  title        = {Pith review of: HMVLM: Multistage Reasoning-Enhanced Vision-Language Model for Long-Tailed Driving Scenarios},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GMQUNRXF}},
  note         = {Machine review of arXiv:2506.05883}
}

read the original abstract

We present HaoMo Vision-Language Model (HMVLM), an end-to-end driving framework that implements the slow branch of a cognitively inspired fast-slow architecture. A fast controller outputs low-level steering, throttle, and brake commands, while a slow planner-a large vision-language model-generates high-level intents such as "yield to pedestrian" or "merge after the truck" without compromising latency. HMVLM introduces three upgrades: (1) selective five-view prompting with an embedded 4s history of ego kinematics, (2) multi-stage chain-of-thought (CoT) prompting that enforces a Scene Understanding -> Driving Decision -> Trajectory Inference reasoning flow, and (3) spline-based trajectory post-processing that removes late-stage jitter and sharp turns. Trained on the Waymo Open Dataset, these upgrades enable HMVLM to achieve a Rater Feedback Score (RFS) of 7.7367, securing 2nd place in the 2025 Waymo Vision-based End-to-End (E2E) Driving Challenge and surpassing the public baseline by 2.77%.

Figures

Figures reproduced from arXiv: 2506.05883 by the authors.

**Figure 1.** Overall diagram of HMVLM. 2. Methodology HMVLM is an end-to-end autonomous driving framework built upon a general-purpose vision-language model. The proposed method progressively reasons from scene understanding to driving decisions and trajectory inference by leveraging multi-stage Chain-of-Thought (CoT) prompting[13] and trajectory refinement [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗

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

NoRD: A Data-Efficient Vision-Language-Action Model that Drives without Reasoning
cs.AI 2026-02 conditional novelty 5.0 of 10

A reasoning-free VLA trained with <60% of typical driving data reaches near-state-of-the-art trajectory scores when fine-tuned with Dr. GRPO instead of GRPO.

Reference graph

Works this paper leans on

14 extracted references · 5 canonical work pages · cited by 1 Pith paper

[1]

Shuai Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wenbin Ge, Sibo Song, Kai Dang, Peng Wang, Shijie Wang, Jun Tang, et al. Qwen2. 5-vl technical report. arXiv preprint arXiv:2502.13923, 2025. 2

arXiv 2025
[2]

Vlmevalkit: An open-source toolkit for evaluating large multi-modality models

Haodong Duan, Junming Yang, Yuxuan Qiao, Xinyu Fang, Lin Chen, Yuan Liu, Xiaoyi Dong, Yuhang Zang, Pan Zhang, Jiaqi Wang, et al. Vlmevalkit: An open-source toolkit for evaluating large multi-modality models. In Proceedings of the 32nd ACM international conference on multimedia , pages 11198–11201, 2024. 2

work page 2024
[3]

Emma: End-to-end multimodal model for autonomous driving

Jyh-Jing Hwang, Runsheng Xu, Hubert Lin, Wei-Chih Hung, Jingwei Ji, Kristy Choi, Di Huang, Tong He, Paul Covington, Benjamin Sapp, et al. Emma: End-to-end multimodal model for autonomous driving. arXiv preprint arXiv:2410.23262,

arXiv
[4]

Vad: Vectorized scene representa- tion for efficient autonomous driving

Bo Jiang, Shaoyu Chen, Qing Xu, Bencheng Liao, Jiajie Chen, Helong Zhou, Qian Zhang, Wenyu Liu, Chang Huang, and Xinggang Wang. Vad: Vectorized scene representa- tion for efficient autonomous driving. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 8340–8350, 2023. 1

2023
[5]

Senna: Bridging large vision-language mod- els and end-to-end autonomous driving

Bo Jiang, Shaoyu Chen, Bencheng Liao, Xingyu Zhang, Wei Yin, Qian Zhang, Chang Huang, Wenyu Liu, and Xing- gang Wang. Senna: Bridging large vision-language mod- els and end-to-end autonomous driving. arXiv preprint arXiv:2410.22313, 2024. 1

arXiv
[6]

Alphadrive: Unleashing the power of vlms in autonomous driving via reinforcement learning and reason- ing

Bo Jiang, Shaoyu Chen, Qian Zhang, Wenyu Liu, and Xing- gang Wang. Alphadrive: Unleashing the power of vlms in autonomous driving via reinforcement learning and reason- ing. arXiv preprint arXiv:2503.07608, 2025. 3

arXiv 2025
[7]

Efficient memory management for large language model serving with pagedattention

Woosuk Kwon, Zhuohan Li, Siyuan Zhuang, Ying Sheng, Lianmin Zheng, Cody Hao Yu, Joseph Gonzalez, Hao Zhang, and Ion Stoica. Efficient memory management for large language model serving with pagedattention. In Pro- ceedings of the 29th Symposium on Operating Systems Prin- ciples, pages 611–626, 2023. 3

work page 2023
[8]

Devel- oping and validating an adaptive multi-layer vehicle trajec- tory reconstruction method for outlier removal

Ruijie Li, Zuduo Zheng, Dong Ngoduy, and Linbo Li. Devel- oping and validating an adaptive multi-layer vehicle trajec- tory reconstruction method for outlier removal. Transporta- tion Research Part C: Emerging Technologies, 171:104946,

work page

Show all 14 references

[9]

A survey on hallucination in large vision-language models

Hanchao Liu, Wenyuan Xue, Yifei Chen, Dapeng Chen, Xiu- tian Zhao, Ke Wang, Liping Hou, Rongjun Li, and Wei Peng. A survey on hallucination in large vision-language models. arXiv preprint arXiv:2402.00253, 2024. 2

2024 arXiv
[10]

Deepspeed: System optimizations enable train- ing deep learning models with over 100 billion parameters

Jeff Rasley, Samyam Rajbhandari, Olatunji Ruwase, and Yuxiong He. Deepspeed: System optimizations enable train- ing deep learning models with over 100 billion parameters. In Proceedings of the 26th ACM SIGKDD international con- ference on knowledge discovery & data mining, pag...

2020
[11]

What is a savitzky-golay filter?[lecture notes]

Ronald W Schafer. What is a savitzky-golay filter?[lecture notes]. IEEE Signal processing magazine , 28(4):111–117,
[12]

Drivevlm: The convergence of autonomous driving and large vision-language models

Xiaoyu Tian, Junru Gu, Bailin Li, Yicheng Liu, Yang Wang, Zhiyong Zhao, Kun Zhan, Peng Jia, Xianpeng Lang, and Hang Zhao. Drivevlm: The convergence of autonomous driving and large vision-language models. arXiv preprint arXiv:2402.12289, 2024. 1

2024 arXiv
[13]

Chain-of-thought prompting elicits reasoning in large lan- guage models

Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. Chain-of-thought prompting elicits reasoning in large lan- guage models. Advances in neural information processing systems, 35:24824–24837, 2022. 2

2022
[14]

Llava-o1: Let vision language models reason step- by-step

Guowei Xu, Peng Jin, Li Hao, Yibing Song, Lichao Sun, and Li Yuan. Llava-o1: Let vision language models reason step- by-step. arXiv preprint arXiv:2411.10440, 2024. 2

2024 arXiv

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1]

Shuai Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wenbin Ge, Sibo Song, Kai Dang, Peng Wang, Shijie Wang, Jun Tang, et al. Qwen2. 5-vl technical report. arXiv preprint arXiv:2502.13923, 2025. 2

arXiv 2025

[2] [2]

Vlmevalkit: An open-source toolkit for evaluating large multi-modality models

Haodong Duan, Junming Yang, Yuxuan Qiao, Xinyu Fang, Lin Chen, Yuan Liu, Xiaoyi Dong, Yuhang Zang, Pan Zhang, Jiaqi Wang, et al. Vlmevalkit: An open-source toolkit for evaluating large multi-modality models. In Proceedings of the 32nd ACM international conference on multimedia , pages 11198–11201, 2024. 2

work page 2024

[3] [3]

Emma: End-to-end multimodal model for autonomous driving

Jyh-Jing Hwang, Runsheng Xu, Hubert Lin, Wei-Chih Hung, Jingwei Ji, Kristy Choi, Di Huang, Tong He, Paul Covington, Benjamin Sapp, et al. Emma: End-to-end multimodal model for autonomous driving. arXiv preprint arXiv:2410.23262,

arXiv

[4] [4]

Vad: Vectorized scene representa- tion for efficient autonomous driving

Bo Jiang, Shaoyu Chen, Qing Xu, Bencheng Liao, Jiajie Chen, Helong Zhou, Qian Zhang, Wenyu Liu, Chang Huang, and Xinggang Wang. Vad: Vectorized scene representa- tion for efficient autonomous driving. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 8340–8350, 2023. 1

2023

[5] [5]

Senna: Bridging large vision-language mod- els and end-to-end autonomous driving

Bo Jiang, Shaoyu Chen, Bencheng Liao, Xingyu Zhang, Wei Yin, Qian Zhang, Chang Huang, Wenyu Liu, and Xing- gang Wang. Senna: Bridging large vision-language mod- els and end-to-end autonomous driving. arXiv preprint arXiv:2410.22313, 2024. 1

arXiv

[6] [6]

Alphadrive: Unleashing the power of vlms in autonomous driving via reinforcement learning and reason- ing

Bo Jiang, Shaoyu Chen, Qian Zhang, Wenyu Liu, and Xing- gang Wang. Alphadrive: Unleashing the power of vlms in autonomous driving via reinforcement learning and reason- ing. arXiv preprint arXiv:2503.07608, 2025. 3

arXiv 2025

[7] [7]

Efficient memory management for large language model serving with pagedattention

Woosuk Kwon, Zhuohan Li, Siyuan Zhuang, Ying Sheng, Lianmin Zheng, Cody Hao Yu, Joseph Gonzalez, Hao Zhang, and Ion Stoica. Efficient memory management for large language model serving with pagedattention. In Pro- ceedings of the 29th Symposium on Operating Systems Prin- ciples, pages 611–626, 2023. 3

work page 2023

[8] [8]

Devel- oping and validating an adaptive multi-layer vehicle trajec- tory reconstruction method for outlier removal

Ruijie Li, Zuduo Zheng, Dong Ngoduy, and Linbo Li. Devel- oping and validating an adaptive multi-layer vehicle trajec- tory reconstruction method for outlier removal. Transporta- tion Research Part C: Emerging Technologies, 171:104946,

work page