pith. sign in

arxiv: 2506.05442 · v2 · pith:LR5TWA5Znew · submitted 2025-06-05 · 💻 cs.CV · cs.AI

Structured Labeling Enables Faster Vision-Language Models for End-to-End Autonomous Driving

Hao Jiang , Chuan Hu , Yukang Shi , Yuan He , Ke Wang , Xi Zhang , Zhipeng Zhang This is my paper

Pith reviewed 2026-05-22 00:11 UTC · model grok-4.3

classification 💻 cs.CV cs.AI
keywords vision-language modelsautonomous drivingstructured labelsNuScenes-SFastDriveend-to-end drivinginference speedcompact models
0
0 comments X

The pith

Structured concise labels let a 0.9B VLM match larger models on driving decisions with over 10x speedup.

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

The paper argues that loose, redundant language descriptions in existing driving datasets hinder vision-language models and slow their inference. It creates NuScenes-S, a structured and concise reformatting of NuScenes data, then trains FastDrive, a compact 0.9 billion parameter model, to read these machine-friendly labels and output driving decisions. This yields roughly 20 percent higher accuracy on decision tasks while running more than ten times faster than baselines such as LLaVA-1.5 that use unstructured text. A reader would care because the approach suggests end-to-end autonomous driving could run in real time on modest hardware rather than requiring massive models and compute.

Core claim

FastDrive, a 0.9B-parameter vision-language model, processes structured concise descriptions from the NuScenes-S dataset to generate machine-friendly driving decisions. It delivers competitive performance with approximately 20% accuracy gains on decision-making tasks and over 10x inference speedup relative to larger VLMs exceeding 7B parameters that handle unstructured language.

What carries the argument

NuScenes-S structured dataset, which converts loose NuScenes descriptions into concise, machine-friendly representations that allow compact VLMs like FastDrive to understand scenes and produce decisions efficiently.

If this is right

  • Compact VLMs become practical for real-time end-to-end autonomous driving on edge hardware.
  • Including scene annotations such as weather and time of day measurably improves decision accuracy.
  • Reducing language redundancy through structured labeling can outweigh gains from simply scaling model size.
  • Inference speed improvements of 10x or more make deployment in production vehicles more feasible.

Where Pith is reading between the lines

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

  • The same structured-labeling tactic could accelerate VLMs in other real-time control domains such as robotics or drone navigation.
  • Lower parameter counts might reduce training energy and data-center costs for autonomous systems.
  • Machine-friendly outputs could simplify integration between perception modules and downstream planners.

Load-bearing premise

Converting original NuScenes language descriptions into structured concise labels preserves every detail required for safe and accurate driving decisions.

What would settle it

Real-vehicle tests in which FastDrive produces more unsafe maneuvers or collisions than a larger unstructured VLM baseline would disprove the performance and safety claims.

Figures

Figures reproduced from arXiv: 2506.05442 by Chuan Hu, Hao Jiang, Ke Wang, Xi Zhang, Yuan He, Yukang Shi, Zhipeng Zhang.

Figure 1
Figure 1. Figure 1: Existing VLMs heavily rely on massive-parameter [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The dataset construction process of the NuScenes-S dataset. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: An annotation example of the NuScenes-S dataset. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: The framework of the FastDrive model for end-to-end autonomous driving. [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Examples of ablation studies on the impact of scene annotations on driving decisions. The red decision represents a [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
read the original abstract

Vision-Language Models (VLMs) offer a promising approach to end-to-end autonomous driving due to their human-like reasoning capabilities. However, troublesome gaps remains between current VLMs and real-world autonomous driving applications. One major limitation is that existing datasets with loosely formatted language descriptions are not machine-friendly and may introduce redundancy. Additionally, high computational cost and massive scale of VLMs hinder the inference speed and real-world deployment. To bridge the gap, this paper introduces a structured and concise benchmark dataset, NuScenes-S, which is derived from the NuScenes dataset and contains machine-friendly structured representations. Moreover, we present FastDrive, a compact VLM baseline with 0.9B parameters. In contrast to existing VLMs with over 7B parameters and unstructured language processing(e.g., LLaVA-1.5), FastDrive understands structured and concise descriptions and generates machine-friendly driving decisions with high efficiency. Extensive experiments show that FastDrive achieves competitive performance on structured dataset, with approximately 20% accuracy improvement on decision-making tasks, while surpassing massive parameter baseline in inference speed with over 10x speedup. Additionally, ablation studies further focus on the impact of scene annotations (e.g., weather, time of day) on decision-making tasks, demonstrating their importance on decision-making tasks in autonomous driving.

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

3 major / 2 minor

Summary. The paper introduces NuScenes-S, a structured and concise benchmark dataset derived from NuScenes with machine-friendly representations (e.g., weather, time, object lists), and proposes FastDrive, a compact 0.9B-parameter VLM that processes these structured inputs to generate driving decisions. It claims competitive performance with approximately 20% accuracy improvement on decision-making tasks and over 10x inference speedup relative to larger unstructured VLMs such as LLaVA-1.5, supported by ablations on scene annotations.

Significance. If the empirical results hold with proper validation, the work could meaningfully advance practical VLM deployment in autonomous driving by showing that structured labeling enables much smaller and faster models. This addresses key barriers of computational cost and input redundancy, with potential for more efficient end-to-end systems.

major comments (3)
  1. [Abstract] Abstract: The central claims of ~20% accuracy improvement on decision-making tasks and >10x speedup are presented without any experimental protocol, baseline details, dataset statistics, error bars, or evaluation metrics, preventing verification of the results.
  2. [§3] Dataset construction: Converting NuScenes free-form descriptions to structured concise labels lacks any quantitative fidelity metric, human validation study, or analysis of retained safety-critical information (e.g., pedestrian intent or multi-agent interactions), so reported gains may partly reflect task simplification.
  3. [§5] §5 (Experiments): Ablation studies on scene annotations (weather, time of day) are described as demonstrating importance but provide no specific quantitative results, tables, or controls, undermining assessment of their contribution to the decision-making claims.
minor comments (2)
  1. [Abstract] Abstract: Grammatical issue - 'troublesome gaps remains' should read 'troublesome gaps remain'.
  2. [Abstract] Abstract: The phrase 'massive parameter baseline' is imprecise; explicitly name the compared models and their parameter counts.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback. We address each major comment point by point below, clarifying existing content where appropriate and outlining specific revisions to improve transparency and rigor.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central claims of ~20% accuracy improvement on decision-making tasks and >10x speedup are presented without any experimental protocol, baseline details, dataset statistics, error bars, or evaluation metrics, preventing verification of the results.

    Authors: We agree that the abstract should better contextualize the claims for readers. The full experimental protocol, baselines (including LLaVA-1.5), dataset statistics from NuScenes-S, decision accuracy as the primary metric, and results with standard deviations are detailed in Sections 4 and 5. In the revised version we will expand the abstract to briefly note the evaluation metric, key baselines, and that results are averaged over multiple runs, while retaining conciseness. revision: yes

  2. Referee: [§3] Dataset construction: Converting NuScenes free-form descriptions to structured concise labels lacks any quantitative fidelity metric, human validation study, or analysis of retained safety-critical information (e.g., pedestrian intent or multi-agent interactions), so reported gains may partly reflect task simplification.

    Authors: We acknowledge that the current Section 3 describes the conversion process and provides examples but does not include quantitative fidelity metrics or human validation. We will add an analysis comparing information retention (with emphasis on safety-critical elements such as pedestrian intent and multi-agent interactions) and report results from a small-scale human validation study in the revised manuscript to address potential concerns about task simplification. revision: yes

  3. Referee: [§5] §5 (Experiments): Ablation studies on scene annotations (weather, time of day) are described as demonstrating importance but provide no specific quantitative results, tables, or controls, undermining assessment of their contribution to the decision-making claims.

    Authors: The ablation results are presented in Section 5 with tables showing accuracy changes when removing individual annotations (weather, time of day) relative to the full structured input. We will revise the text to explicitly reference these quantitative values, clarify the control conditions, and discuss the magnitude of each annotation's contribution to decision accuracy. revision: yes

Circularity Check

0 steps flagged

No significant circularity; empirical claims rest on new dataset and model evaluations

full rationale

The paper introduces NuScenes-S as a structured reformatting of NuScenes descriptions and presents FastDrive as a compact VLM trained and evaluated on it. No equations, derivations, or first-principles predictions appear in the provided text. Performance claims (accuracy gains, speedups) are presented as direct experimental outcomes against baselines rather than quantities forced by fitting or self-definition. No load-bearing self-citations, uniqueness theorems, or ansatzes are invoked to justify core results. The work is self-contained as standard empirical ML research: new data representation plus model training, with results measured on held-out tasks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Central claims rest on the untested premise that structured labels retain full decision-relevant information; no free parameters or new entities are mentioned.

axioms (1)
  • domain assumption Structured and concise scene annotations preserve all information needed for accurate driving decisions
    The paper's performance claims depend on this premise being true; it is invoked when the authors state that NuScenes-S is machine-friendly without loss of utility.

pith-pipeline@v0.9.0 · 5774 in / 1169 out tokens · 61022 ms · 2026-05-22T00:11:36.907359+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

40 extracted references · 40 canonical work pages · 14 internal anchors

  1. [1]

    Scenario understanding of traffic scenes through large visual language models,

    R. Esteban, L. Jannik, N. Uhlemann, and M. Lienkamp, “Scenario understanding of traffic scenes through large visual language models,”

    work page

  2. [2]

    Available: https://arxiv.org/abs/2501.17131

    [Online]. Available: https://arxiv.org/abs/2501.17131

    work page arXiv

  3. [3]

    Vision language models in autonomous driving: A survey and outlook.arXiv preprint arXiv:2310.14414, 2023

    X. Zhou, M. Liu, E. Yurtsever, B. L. Zagar, W. Zimmer, H. Cao, and A. C. Knoll, “Vision language models in autonomous driving: A survey and outlook,” 2024. [Online]. Available: https://arxiv.org/abs/2310.14414

    work page arXiv 2024

  4. [5]

    Drivelm: Driving with graph visual question answering.arXiv preprint arXiv:2312.14150,

    C. Sima, K. Renz, K. Chitta, L. Chen, H. Zhang, C. Xie, J. Beißwenger, P. Luo, A. Geiger, and H. Li, “Drivelm: Driving with graph visual question answering,” 2025. [Online]. Available: https://arxiv.org/abs/2312.14150

    work page arXiv 2025

  5. [6]

    DriveVLM: The Convergence of Autonomous Driving and Large Vision-Language Models

    X. Tian, J. Gu, B. Li, Y . Liu, Y . Wang, Z. Zhao, K. Zhan, P. Jia, X. Lang, and H. Zhao, “Drivevlm: The convergence of autonomous driving and large vision-language models,” 2024. [Online]. Available: https://arxiv.org/abs/2402.12289

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  6. [12]

    Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone

    M. Abdin, J. Aneja, H. Awadalla, and A. Awadallah, “Phi-3 technical report: A highly capable language model locally on your phone,” 2024. [Online]. Available: https://arxiv.org/abs/2404.14219

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  7. [13]

    Qwen Technical Report

    J. Bai, S. Bai, Y . Chu, Z. Cui, K. Dang, and X. Deng, “Qwen technical report,” 2023. [Online]. Available: https://arxiv.org/abs/2309.16609

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  8. [14]

    Visual Instruction Tuning

    H. Liu, C. Li, Q. Wu, and Y . J. Lee, “Visual instruction tuning,” 2023. [Online]. Available: https://arxiv.org/abs/2304.08485

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  9. [15]

    Improved Baselines with Visual Instruction Tuning

    H. Liu, C. Li, Y . Li, and Y . J. Lee, “Improved baselines with visual instruction tuning,” 2024. [Online]. Available: https: //arxiv.org/abs/2310.03744

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  10. [16]

    InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Tasks

    Z. Chen, J. Wu, W. Wang, W. Su, G. Chen, S. Xing, M. Zhong, Q. Zhang, X. Zhu, L. Lu, B. Li, P. Luo, T. Lu, Y . Qiao, and J. Dai, “Internvl: Scaling up vision foundation models and aligning for generic visual-linguistic tasks,” 2024. [Online]. Available: https://arxiv.org/abs/2312.14238

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  11. [17]

    MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Models

    D. Zhu, J. Chen, X. Shen, X. Li, and M. Elhoseiny, “Minigpt-4: Enhancing vision-language understanding with advanced large language models,” 2023. [Online]. Available: https://arxiv.org/abs/2304.10592

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  12. [18]

    Learning Transferable Visual Models From Natural Language Supervision

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark, G. Krueger, and I. Sutskever, “Learning transferable visual models from natural language supervision,” 2021. [Online]. Available: https://arxiv.org/abs/2103.00020

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  13. [19]

    Holistic autonomous driving understanding by bird’s-eye-view injected multi- modal large models,

    X. Ding, J. Han, H. Xu, X. Liang, W. Zhang, and X. Li, “Holistic autonomous driving understanding by bird’s-eye-view injected multi- modal large models,” 2024. [Online]. Available: https://arxiv.org/abs/ 2401.00988

    work page arXiv 2024

  14. [20]

    Visual chatgpt: Talking, drawing and editing with visual foundation models,

    C. Wu, S. Yin, W. Qi, X. Wang, Z. Tang, and N. Duan, “Visual chatgpt: Talking, drawing and editing with visual foundation models,”

    work page

  15. [21]

    Visual ChatGPT: Talking, Drawing and Editing with Visual Foundation Models

    [Online]. Available: https://arxiv.org/abs/2303.04671

    work page internal anchor Pith review Pith/arXiv arXiv

  16. [22]

    MoE-LLaVA: Mixture of Experts for Large Vision-Language Models

    B. Lin, Z. Tang, Y . Ye, J. Huang, J. Zhang, Y . Pang, P. Jin, M. Ning, J. Luo, and L. Yuan, “Moe-llava: Mixture of experts for large vision-language models,” 2024. [Online]. Available: https: //arxiv.org/abs/2401.15947

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  17. [23]

    The entity-deduction arena: A playground for probing the conversational reasoning and planning capabilities of LLMs,

    Y . Zhang, J. Lu, and N. Jaitly, “The entity-deduction arena: A playground for probing the conversational reasoning and planning capabilities of LLMs,” 2024. [Online]. Available: https://openreview. net/forum?id=PfrpYGKGPL

    work page 2024

  18. [24]

    An in-depth look at gemini’s language abilities,

    S. N. Akter, Z. Yu, A. Muhamed, T. Ou, A. B ¨auerle, ´Angel Alexander Cabrera, K. Dholakia, C. Xiong, and G. Neubig, “An in-depth look at gemini’s language abilities,” 2023. [Online]. Available: https://arxiv.org/abs/2312.11444

    work page arXiv 2023

  19. [25]

    Llm4drive: A survey of large language models for autonomous driving.ArXiv, abs/2311.01043, 2023

    Z. Yang, X. Jia, H. Li, and J. Yan, “Llm4drive: A survey of large language models for autonomous driving,” 2024. [Online]. Available: https://arxiv.org/abs/2311.01043

    work page arXiv 2024

  20. [26]

    Dilu: A knowledge-driven approach to au- tonomous driving with large language models

    L. Wen, D. Fu, X. Li, X. Cai, T. Ma, P. Cai, M. Dou, B. Shi, L. He, and Y . Qiao, “Dilu: A knowledge-driven approach to autonomous driving with large language models,” 2024. [Online]. Available: https://arxiv.org/abs/2309.16292

    work page arXiv 2024

  21. [27]

    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,”

    work page

  22. [28]

    Drive like a human: Rethink- ing autonomous driving with large language models

    [Online]. Available: https://arxiv.org/abs/2307.07162

    work page arXiv

  23. [30]

    Lmdrive: Closed-loop end-to-end driving with large language models,

    H. Shao, Y . Hu, L. Wang, S. L. Waslander, Y . Liu, and H. Li, “Lmdrive: Closed-loop end-to-end driving with large language models,”

    work page

  24. [31]

    Available: https://arxiv.org/abs/2312.07488

    [Online]. Available: https://arxiv.org/abs/2312.07488

    work page arXiv

  25. [32]

    Drivemlm: Aligning multi-modal large language models with behavioral planning states for au- tonomous driving

    W. Wang, J. Xie, C. Hu, H. Zou, J. Fan, W. Tong, Y . Wen, S. Wu, H. Deng, Z. Li, H. Tian, L. Lu, X. Zhu, X. Wang, Y . Qiao, and J. Dai, “Drivemlm: Aligning multi-modal large language models with behavioral planning states for autonomous driving,” 2023. [Online]. Available: https://arxiv.org/abs/2312.09245

    work page arXiv 2023

  26. [33]

    Senna: Bridging Large Vision-Language Models and End-to-End Autonomous Driving

    B. Jiang, S. Chen, B. Liao, X. Zhang, W. Yin, Q. Zhang, C. Huang, W. Liu, and X. Wang, “Senna: Bridging large vision-language models and end-to-end autonomous driving,” 2024. [Online]. Available: https://arxiv.org/abs/2410.22313

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  27. [34]

    Continuously learning, adapting, and improving: A dual-process approach to autonomous driving,

    J. Mei, Y . Ma, X. Yang, L. Wen, X. Cai, X. Li, D. Fu, B. Zhang, P. Cai, M. Dou, B. Shi, L. He, Y . Liu, and Y . Qiao, “Continuously learning, adapting, and improving: A dual-process approach to autonomous driving,” 2024. [Online]. Available: https://arxiv.org/abs/2405.15324

    work page arXiv 2024

  28. [35]

    Vision meets robotics: The KITTI dataset,

    A. Geiger, P. Lenz, C. Stiller, and R. Urtasun, “Vision meets robotics: The kitti dataset,” The International Journal of Robotics Research, vol. 32, no. 11, pp. 1231–1237, 2013. [Online]. Available: https://doi.org/10.1177/0278364913491297

    work page doi:10.1177/0278364913491297 2013

  29. [36]

    Scalability in perception for autonomous driving: Waymo open dataset,

    P. Sun, H. Kretzschmar, X. Dotiwalla, A. Chouard, and V . Patnaik, “Scalability in perception for autonomous driving: Waymo open dataset,” 2020. [Online]. Available: https://arxiv.org/abs/1912.04838

    work page arXiv 2020

  30. [37]

    nuScenes: A multimodal dataset for autonomous driving

    H. Caesar, V . Bankiti, A. H. Lang, S. V ora, V . E. Liong, Q. Xu, A. Krishnan, Y . Pan, G. Baldan, and O. Beijbom, “nuscenes: A multimodal dataset for autonomous driving,” 2020. [Online]. Available: https://arxiv.org/abs/1903.11027

    work page internal anchor Pith review Pith/arXiv arXiv 2020

  31. [38]

    Talk2car: Taking control of your self-driving car,

    T. Deruyttere, S. Vandenhende, D. Grujicic, L. Van Gool, and M.-F. Moens, “Talk2car: Taking control of your self-driving car,” in Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP) . 9 Association for Computational Linguistics, 2019...

    work page doi:10.18653/v1/d19-1215 2019

  32. [39]

    Language prompt for autonomous driving,

    D. Wu, W. Han, T. Wang, Y . Liu, X. Zhang, and J. Shen, “Language prompt for autonomous driving,” 2023. [Online]. Available: https://arxiv.org/abs/2309.04379

    work page arXiv 2023

  33. [40]

    Nuscenes-qa: A multi-modal visual question answering benchmark for autonomous driving scenario,

    T. Qian, J. Chen, L. Zhuo, Y . Jiao, and Y .-G. Jiang, “Nuscenes-qa: A multi-modal visual question answering benchmark for autonomous driving scenario,” 2024. [Online]. Available: https://arxiv.org/abs/2305. 14836

    work page 2024

  34. [41]

    Textual Explanations for Self-Driving Vehicles

    J. Kim, A. Rohrbach, T. Darrell, J. Canny, and Z. Akata, “Textual explanations for self-driving vehicles,” 2018. [Online]. Available: https://arxiv.org/abs/1807.11546

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  35. [42]

    Explainable object-induced action decision for autonomous vehicles,

    Y . Xu, X. Yang, L. Gong, H.-C. Lin, T.-Y . Wu, Y . Li, and N. Vasconcelos, “Explainable object-induced action decision for autonomous vehicles,” 2020. [Online]. Available: https://arxiv.org/abs/ 2003.09405

    work page arXiv 2020

  36. [43]

    Drama: Joint risk localization and captioning in driving,

    S. Malla, C. Choi, I. Dwivedi, J. H. Choi, and J. Li, “Drama: Joint risk localization and captioning in driving,” 2022. [Online]. Available: https://arxiv.org/abs/2209.10767

    work page arXiv 2022

  37. [44]

    Rank2tell: A multimodal driving dataset for joint importance ranking and reasoning,

    E. Sachdeva, N. Agarwal, S. Chundi, S. Roelofs, J. Li, M. Kochenderfer, C. Choi, and B. Dariush, “Rank2tell: A multimodal driving dataset for joint importance ranking and reasoning,” 2023. [Online]. Available: https://arxiv.org/abs/2309.06597

    work page arXiv 2023

  38. [45]

    Are vlms ready for autonomous driving? an empirical study from the reliability, data, and metric perspectives.arXiv preprint arXiv:2501.04003, 2025

    S. Xie, L. Kong, Y . Dong, C. Sima, W. Zhang, Q. A. Chen, Z. Liu, and L. Pan, “Are vlms ready for autonomous driving? an empirical study from the reliability, data, and metric perspectives,” 2025. [Online]. Available: https://arxiv.org/abs/2501.04003

    work page arXiv 2025

  39. [46]

    Expanding Performance Boundaries of Open-Source Multimodal Models with Model, Data, and Test-Time Scaling

    Z. Chen, W. Wang, Y . Cao, Y . Liu, Z. Gao, E. Cui, J. Zhu, S. Ye, H. Tian, Z. Liu, L. Gu, X. Wang, Q. Li, Y . Ren, Z. Chen, J. Luo, J. Wang, T. Jiang, B. Wang, C. He, B. Shi, X. Zhang, H. Lv, Y . Wang, W. Shao, P. Chu, Z. Tu, T. He, Z. Wu, H. Deng, J. Ge, K. Chen, K. Zhang, L. Wang, M. Dou, L. Lu, X. Zhu, T. Lu, D. Lin, Y . Qiao, J. Dai, and W. Wang, “Ex...

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  40. [47]

    Qwen2.5 technical report,

    Qwen, :, A. Yang, B. Yang, B. Zhang, B. Hui, and B. Zheng, “Qwen2.5 technical report,” 2025. [Online]. Available: https://arxiv.org/abs/2412. 15115

    work page 2025