Pith. sign in

REVIEW 4 major objections 5 minor 52 references

One is Plenty: A Polymorphic Feature Interpreter for Immutable Heterogeneous Collaborative Perception

T0 review · 4 major / 5 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 single interpreter can map any neighbor's features into the ego vehicle's semantic space.

desk verdict A solid prompt-tuning architecture for heterogeneous collaborative perception, but the abstract's numbers don't match the tables and the generalization claim is only tested with resolution variants of the same backbone families. read the letter →

arxiv 2411.16799 v2 pith:SV2G4ZWE submitted 2024-11-25 cs.CV

Yuchen Xia , Quan Yuan , Guiyang Luo , Xiaoyuan Fu , Yang Li , Xuanhan Zhu , Tianyou Luo , Siheng Chen
show 1 more author
Jinglin Li
This is my paper · ORCID
classification cs.CV
keywords collaborativeperceptionimmutableheterogeneityfeatureinterpreterprompttuningpolymorphismautonomousdriving3DobjectdetectionOPV2V
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 addresses an openness problem in collaborative perception: when vehicles with different, fixed perception networks share intermediate features, those features are not mutually understandable, and prior solutions either retrain a full interpreter for every new agent type or pay a two-stage semantic-loss penalty through a standard semantic space. PolyInter is proposed as a single polymorphic feature interpreter that maps any neighbor's features into the ego vehicle's semantic space in one stage, using one shared interpreter network plus a general prompt and per-agent specific prompts. New agent types are integrated by fine-tuning only their specific prompt and a small resizer, with the interpreter network and general prompt frozen. On the OPV2V dataset the paper reports precision improvements of up to 11.1% AP@0.7 over existing interpreters while training as little as 1.4% of PolyInter's parameters when adapting to a new agent.

What carries the argument

PolyInter's load-bearing object is the polymorphic interpreter: one interpreter network plus a shared general prompt and a stack of agent-specific prompts, all hosted on the ego agent. The interpreter network is built from a Channel Selection Module, which computes a cosine-similarity matrix between ego and neighbor BEV feature channels and reweights the neighbor feature accordingly, and a Spatial Attention Module, which applies fused axial attention to align positions into the ego semantic space. The prompts are what make the design polymorphic: the general prompt G is shared by all agents and is trained to encode cross-agent commonalities through an adversarial loss and a style loss, while each specific prompt S_i is trained with a single-agent detection loss to retain encoder-specific semantics. Phase I trains the interpreter and all prompts jointly with a collaborative detection loss; Phase II freezes everything except the new agent's specific prompt and a resizer, and the specific prompt can be low-rank decomposed to reduce trainable parameters to about 1.4% of the interpreter.

What would settle it

Take a neighbor encoder whose architecture was not seen in Phase I (for example, a transformer-based LiDAR backbone rather than a resolution variant of PointPillar, VoxelNet, or SECOND), fine-tune only its specific prompt, and measure AP@0.7 on OPV2V against retraining the full interpreter on that agent type; if the prompt-only result falls materially below the retrained interpreter while resolution-variant prompts do not, the extensibility claim is refuted.

Watch

Extended reading notes

Core claim

At its center is the claim that semantic alignment across immutable heterogeneous encoders can be factored into a shared alignment mechanism plus a tiny per-agent identity. The shared mechanism is learned once: a channel selection module matches neighbor feature channels to ego channels by cosine similarity, and a spatial attention module aligns spatial semantics by fused axial attention. The per-agent identity lives in learnable prompts: a general prompt is trained to capture common semantics via adversarial and style losses, while each specific prompt is trained to preserve agent-specific semantics via a single-agent detection loss. In the generalization phase, only the new agent's specific prompt and a resizer are optimized, and the paper shows on OPV2V (with supplementary results on V2XSet and DAIR-V2X) that this prompt-only adaptation reaches or exceeds the accuracy of interpreters that are fully retrained per agent pair.

Load-bearing premise

The shared interpreter network trained in Phase I generalizes to genuinely new agent types after only prompt fine-tuning; the Phase II tests use resolution variants of the same backbone families seen in Phase I, so the experiments do not yet establish transfer across architecturally novel encoders.

Editorial extensions

If this is right

  • A fleet can integrate a new vehicle model by storing one extra prompt per agent type instead of one interpreter per ego-neighbor pair, reducing storage and adaptation cost.
  • Because interpretation is one-stage, it avoids the cumulative semantic loss that two-stage standard-space methods incur, so the accuracy gap over those methods should widen as the semantic gap between encoders grows.
  • Prompts for previously integrated agents remain untouched when a new agent is added, so the interpreter's behavior for existing collaboration partners is not disturbed.
  • The low-rank prompt decomposition offers a direct trade-off between on-board parameter budget and perception precision, making prompt-only adaptation feasible under memory constraints.

Reading between the lines

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

  • The paper does not test architecturally novel backbones, since Phase II agents (pp4, sd1, vn6) are resolution and layer-count variants of the same PointPillar, SECOND, and VoxelNet families used in Phase I; a fair reader should treat transfer to genuinely new architectures as an open prediction rather than an established result.
  • If prompts encode a compact agent identity, then interpolating or averaging the prompts of known agents may produce a usable prompt for an unseen encoder without any fine-tuning, which would be a cheap testable extension.
  • The adversarial training that strips agent-specific statistics from the general feature suggests the same interpreter could double as a privacy filter, removing identifying encoder statistics from shared features; that use is not evaluated here.
  • The storage model of one interpreter plus many prompts raises a capacity question: at some number of agent types, prompt interference may degrade accuracy, and the paper gives no evidence about that ceiling.
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.

Referee Report

4 major / 5 minor

Summary. The paper proposes PolyInter, a polymorphic feature interpreter for immutable heterogeneous collaborative perception. A single interpreter network is trained in Phase I on a small set of neighbor encoders, with a shared general prompt and per-agent specific prompts; in Phase II, new agents are added by fine-tuning only the specific prompt and a resizer. Experiments on OPV2V (and in the supplement V2XSet, DAIR-V2X, and multi-modal settings) compare against PnPDA and MPDA, reporting improved AP@0.5/0.7 and reduced trainable parameters.

Significance. If the extensibility claim holds, PolyInter is a practically valuable contribution: it reduces the cost of adding new agent types to a collaborative perception system and avoids the cumulative loss of two-stage interpreters. The paper provides reproducible code, external datasets, ablations of each module, and parameter-count analysis. The central open question is whether prompt-only fine-tuning transfers to architecturally novel encoders, which the current experiments do not establish.

major comments (4)
  1. [Sec. 4.1, Table 2] The claim of extensibility to new agent types is not supported by the experimental design. In Phase I combination pp8-vn4-sd2, the Phase II agents pp4, sd1 and vn6 are resolution variants of the PointPillar, SECOND and VoxelNet families already present in Phase I (Table 1); in combination pp8-pp4-vn4, pp4 is literally a Phase I neighbor, so the '+' row for pp8-pp4 in Table 2 measures retained performance on a seen encoder rather than generalization. To substantiate the headline claim, the authors should evaluate on architecturally novel encoders not seen in Phase I, or explicitly restrict the claim to within-family adaptation.
  2. [Sec. 3.2, Eqs. (1)-(4)] Because the channel selection module computes the similarity matrix M using W_q and W_k learned only on Phase I encoders and then reorganizes both the neighbor feature and the specific prompt with the same M (Eqs. (2)-(4)), the specific prompt cannot compensate for a semantically mismatched channel selection on a truly new architecture. The current experiments never expose this risk, since all Phase II LiDAR encoders belong to the three Phase I families. Please provide an experiment with a distinct architecture (e.g., a different 3D backbone family or a camera-only encoder) or an analysis demonstrating that the channel correspondence learned in Phase I transfers.
  3. [Abstract and Sec. 4.2] The reported improvements '7.9% and 11.1%' (Abstract) and '9.2%' (Sec. 4.2) do not match the AP values in Table 2. For example, in pp8-pp4* with F-cooper, AP@0.7 is 60.9 vs 51.1 (PnPDA) and 53.1 (MPDA), giving improvements of 9.8 and 7.8 points, not 9.2; the largest AP@0.7 gain in the table is 24.5 points (pp8-sd1* with F-cooper vs MPDA). Please specify exactly how each claimed number is computed (which scenarios, which baseline, absolute or relative), and report variance across runs or seeds.
  4. [Sec. 4.2] The comparison set includes only PnPDA and MPDA, although the related work cites other heterogeneous collaborative perception methods (e.g., HEAL [23], Hetecooper [28], V2X-ViT [34]). If these are not directly comparable because they are not immutable-feature interpreters, state this explicitly; otherwise the 'SOTA interpreter' claim is not supported. Adding at least one recent comparable method would strengthen the empirical claim.
minor comments (5)
  1. [Table 2] The caption should clarify that '+' rows use the pp8-pp4-vn4 Phase I combination, in which pp4 is a Phase I neighbor, and that '*' rows use pp8-vn4-sd2; this distinction is easy to miss.
  2. [Eq. (8)] In the adversarial loss equation, both expectation terms are written with F_g as the argument; if one term is intended to be evaluated on F_ego, this is a typo and should be corrected.
  3. [Sec. 4.3] The statement that 0.16M parameters is '1.4% of the respective interpreters' would be clearer if the total interpreter parameter count were reported in the main text.
  4. [Table 1] The column header 'V oxel' and the formatting of 'Half Lidar Range (x,y)' appear inconsistent; the table should be cleaned up.
  5. [Figure 4] Please clarify whether the parameter counts for PnPDA and MPDA include the fusion module and the standard-semantic-space interpreters, so the comparison is apples-to-apples.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the method is trained and evaluated on external benchmarks, and the Phase II generalization evidence, while narrow, is not definitionally forced.

full rationale

PolyInter is an empirical system trained end-to-end with supervised losses (Lcollab, Lsingle, adversarial/style regularizers) and evaluated on fixed public benchmarks (OPV2V, V2XSet, DAIR-V2X) with standard AP@IoU metrics. The Phase II 'extension' is not derived from a fitted constant; it is a genuinely separate fine-tuning phase, and the 11.1% AP gain is a measured comparison against PnPDA and MPDA baselines. The paper does contain self-citations: PnPDA shares several authors and is used as a comparison baseline, but the central claim does not reduce to those citations; PnPDA is an externally published ECCV baseline rather than a premise used to define the method. The strongest limitation is experimental scope: all Phase II LiDAR agents are resolution variants of the Phase I backbone families, and the '+' condition pp8-pp4+ retests a Phase I neighbor. That weakens the generalization claim, but it is a threat to external validity, not a circular derivation: the numbers are not constructed from the claim, and no equation or fitted parameter is renamed as a prediction. No self-definitional, fitted-input, or self-citation-load-bearing step can be quoted from the paper, so no circularity is found.

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

Free-parameter and axiom entries reflect the core design choices. The paper introduces no invented physical entities; prompts, channel selection, and spatial attention are learned components, not new entities requiring external falsifiable evidence.

free parameters (5)
  • omega (style loss weight) = 0.5
    Set by hand in Section 4.1; balances style loss terms in Ls and Lg; no sensitivity analysis is reported.
  • lambda_s (single-agent loss weight) = 1
    Hand-set in Section 4.1; weights Lsingle in the total loss.
  • lambda_g (general loss weight) = 1
    Hand-set in Section 4.1; weights Ladversary plus style loss in the total loss.
  • Low-rank factors R and T for specific prompt = R varies (1-20), T varies (1-4)
    Selected via parameter sweep in Figure 5; the 1.4% parameter claim depends on this choice, with no separate validation protocol described.
  • N (number of point cloud samples for prompt initialization) = not stated
    Sampling-based prompt initialization in Section 3.2 requires N samples, but the paper never gives the value of N.
assumptions (5)
  • domain assumption Cosine similarity between ego and neighbor feature channels defines a meaningful correspondence for semantic alignment.
    Used in channel selection module, Section 3.2 Eq. (1)-(2); if channel semantics are not monotonically related across encoders, the softmax matching may be meaningless.
  • domain assumption A single shared interpreter network, frozen after Phase I, can align features of previously unseen agent types.
    Core of Phase II generalization, Section 3.4; the paper provides no theoretical guarantee and tests only variants of the same three architecture families.
  • domain assumption The ego agent's semantic space is a sufficient and lossless target for interpretation.
    One-stage interpretation always maps neighbors into ego space (Section 3.1); multi-agent fusion could lose information that a neutral space would preserve.
  • standard math Standard attention and layer normalization behave as expected for BEV features.
    SoftMax, LayerNorm, and fused axial attention are used as black boxes in Section 3.2.
  • domain assumption OPV2V and the selected encoder variants represent the range of immutable heterogeneous collaboration.
    All main results use OPV2V; DAIR-V2X and V2XSet are in the appendix; encoder variants are limited to three LiDAR backbones.

how reviews work

0 comments
Cite this review

Pith. "Pith review of One is Plenty: A Polymorphic Feature Interpreter for Immutable Heterogeneous Collaborative Perception." pith.science (2026). https://pith.science/paper/SV2G4ZWE

@misc{pith2026241116799,
  author       = {Pith},
  title        = {Pith review of: One is Plenty: A Polymorphic Feature Interpreter for Immutable Heterogeneous Collaborative Perception},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SV2G4ZWE}},
  note         = {Machine review of arXiv:2411.16799}
}
read the original abstract

Collaborative perception in autonomous driving significantly enhances the perception capabilities of individual agents. Immutable heterogeneity, where agents have different and fixed perception networks, presents a major challenge due to the semantic gap in exchanged intermediate features without modifying the perception networks. Most existing methods bridge the semantic gap through interpreters. However, they either require training a new interpreter for each new agent type, limiting extensibility, or rely on a two-stage interpretation via an intermediate standardized semantic space, causing cumulative semantic loss. To achieve both extensibility in immutable heterogeneous scenarios and low-loss feature interpretation, we propose PolyInter, a polymorphic feature interpreter. It provides an extension point where new agents integrate by overriding only their specific prompts, which are learnable parameters that guide interpretation, while reusing PolyInter's remaining parameters. By leveraging polymorphism, our design enables a single interpreter to accommodate diverse agents and interpret their features into the ego agent's semantic space. Experiments on the OPV2V dataset demonstrate that PolyInter improves collaborative perception precision by up to 11.1% compared to SOTA interpreters, while comparable results can be achieved by training only 1.4% of PolyInter's parameters when adapting to new agents. Code is available at https://github.com/yuchen-xia/PolyInter.

Figures

Figures reproduced from arXiv: 2411.16799 by the authors.

Figure 1
Figure 1. Comparison of different immutable heterogeneous collaborative strategies for extending collaboration with new neighbor agents. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. The overall architecture of PolyInter. PolyInter establishes a common structure that can be inherited by multiple agents, providing [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Visualization of BEV feature maps from two hetero [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Comparison of the number of trainable parameters with [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 6
Figure 6. Figure 6: Visualization of intermediate features before and after [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 1
Figure 1. Figure 1: Visualization of the ego feature, the general prompt, the specific prompts corresponding to different neighbor agents, and the [PITH_FULL_IMAGE:figures/full_fig_p013_1.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

52 extracted references · 49 canonical work pages

  1. [23]

    An extensible framework for open heterogeneous collaborative perception

    Yifan Lu, Yue Hu, Yiqi Zhong, Dequan Wang, Yanfeng Wang, and Siheng Chen. An extensible framework for open heterogeneous collaborative perception. In International Conference on Learning Representationss (ICLR), 2024. 3

    work page 2024

  2. [28]

    Hetecooper: Feature collaboration graph for heterogeneous collaborative percep- tion

    Congzhang Shao, Guiyang Luo, Quan Yuan, Yifu Chen, Yilin Liu, Kexin Gong, and Jinglin Li. Hetecooper: Feature collaboration graph for heterogeneous collaborative percep- tion. In European Conference on Computer Vision (ECCV), pages 162–178, 2025. 3

    work page 2025

  3. [34]

    V2x-vit: Vehicle-to-everything cooperative perception with vision transformer

    Runsheng Xu, Hao Xiang, Zhengzhong Tu, Xin Xia, Ming- Hsuan Yang, and Jiaqi Ma. V2x-vit: Vehicle-to-everything cooperative perception with vision transformer. InEuropean Conference on Computer Vision (ECCV) , pages 107–124,

    work page

  4. [1]

    Cooperative perception for 3d object detec- tion in driving scenarios using infrastructure sensors

    Eduardo Arnold, Mehrdad Dianati, Robert de Temple, and Saber Fallah. Cooperative perception for 3d object detec- tion in driving scenarios using infrastructure sensors. IEEE Trans. Intell. Transp. Syst., 23(3):1852–1864, 2022. 1

    work page 2022

  5. [2]

    Diprompt: Disentan- gled prompt tuning for multiple latent domain generaliza- tion in federated learning

    Sikai Bai, Jie Zhang, Song Guo, Shuaicheng Li, Jingcai Guo, Jun Hou, Tao Han, and Xiaocheng Lu. Diprompt: Disentan- gled prompt tuning for multiple latent domain generaliza- tion in federated learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 27284–27293, 2024. 2

    work page 2024

  6. [3]

    Visual prompting via image inpaint- ing

    Amir Bar, Yossi Gandelsman, Trevor Darrell, Amir Glober- son, and Alexei Efros. Visual prompting via image inpaint- ing. In Advances in Neural Information Processing Systems (NeurIPS), pages 25005–25017, 2022. 3

    work page 2022

  7. [4]

    Meyer, Yuning Chai, Dennis Park, and Yong Jae Lee

    Mu Cai, Haotian Liu, Siva Karthik Mustikovela, Gregory P. Meyer, Yuning Chai, Dennis Park, and Yong Jae Lee. Vip- llava: Making large multimodal models understand arbitrary visual prompts. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 12914–12923, 2024. 3

    work page 2024

  8. [5]

    Domain-controlled prompt learning

    Qinglong Cao, Zhengqin Xu, Yuntian Chen, Chao Ma, and Xiaokang Yang. Domain-controlled prompt learning. In AAAI Conference on Artificial Intelligence (AAAI) , pages 936–944, 2024. 3

    work page 2024

Show all 52 references
  1. [6]

    Multi-prompt alignment for multi-source unsuper- vised domain adaptation

    Haoran Chen, Xintong Han, Zuxuan Wu, and Yu-Gang Jiang. Multi-prompt alignment for multi-source unsuper- vised domain adaptation. In Advances in Neural Information Processing Systems (NeurIPS), pages 74127–74139, 2023. 3

    2023

  2. [7]

    F-cooper: feature based cooperative perception for autonomous vehicle edge computing system using 3d point clouds

    Qi Chen, Xu Ma, Sihai Tang, Jingda Guo, Qing Yang, and Song Fu. F-cooper: feature based cooperative perception for autonomous vehicle edge computing system using 3d point clouds. In Proceedings of the 4th ACM/IEEE Symposium on Edge Computing (SEC), page 88–100, 2019. 7, 2, 3

    2019

  3. [8]

    Disentangled prompt rep- resentation for domain generalization

    De Cheng, Zhipeng Xu, Xinyang Jiang, Nannan Wang, Dongsheng Li, and Xinbo Gao. Disentangled prompt rep- resentation for domain generalization. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 23595–23604, 2024. 3

    2024

  4. [9]

    Decorate the newcomers: visual domain prompt for continual test time adaptation

    Yulu Gan, Yan Bai, Yihang Lou, Xianzheng Ma, Renrui Zhang, Nian Shi, and Lin Luo. Decorate the newcomers: visual domain prompt for continual test time adaptation. In AAAI Conference on Artificial Intelligence (AAAI) , pages 7595–7603, 2023. 3

    2023

  5. [10]

    Collaborative perception in autonomous driv- ing: Methods, datasets, and challenges

    Yushan Han, Hui Zhang, Huifang Li, Yi Jin, Congyan Lang, and Yidong Li. Collaborative perception in autonomous driv- ing: Methods, datasets, and challenges. IEEE Intell. Transp. Syst. Mag., 15(6):131–151, 2023. 2

    2023

  6. [11]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 770–778, 2016. 1, 2

    2016

  7. [12]

    LoRA: Low-rank adaptation of large language models

    Edward J Hu, yelong shen, Phillip Wallis, Zeyuan Allen- Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. LoRA: Low-rank adaptation of large language models. InIn- ternational Conference on Learning Representations (ICLR),

  8. [13]

    Where2comm: Communication-efficient collab- orative perception via spatial confidence maps

    Yue Hu, Shaoheng Fang, Zixing Lei, Yiqi Zhong, and Si- heng Chen. Where2comm: Communication-efficient collab- orative perception via spatial confidence maps. In Advances in Neural Information Processing Systems (NeurIPS), pages 4874–4886, 2022. 1, 3

    2022

  9. [14]

    Communication-efficient collaborative percep- tion via information filling with codebook

    Yue Hu, Juntong Peng, Sifei Liu, Junhao Ge, Si Liu, and Si- heng Chen. Communication-efficient collaborative percep- tion via information filling with codebook. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition (CVPR), pages 15481–15490, 2024. 3

    2024

  10. [15]

    Belongie, Bharath Hariharan, and Ser-Nam Lim

    Menglin Jia, Luming Tang, Bor-Chun Chen, Claire Cardie, Serge J. Belongie, Bharath Hariharan, and Ser-Nam Lim. Vi- sual prompt tuning. In European Conference on Computer Vision (ECCV), pages 709–727, 2022. 2, 3

    2022

  11. [16]

    Lang, Sourabh V ora, Holger Caesar, Lubing Zhou, Jiong Yang, and Oscar Beijbom

    Alex H. Lang, Sourabh V ora, Holger Caesar, Lubing Zhou, Jiong Yang, and Oscar Beijbom. Pointpillars: Fast encoders for object detection from point clouds. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019. 4, 6, 7, 1, 2, 3

    2019

  12. [17]

    Latency-aware collaborative perception

    Zixing Lei, Shunli Ren, Yue Hu, Wenjun Zhang, and Siheng Chen. Latency-aware collaborative perception. In European Conference on Computer Vision (ECCV) , page 316–332,

  13. [18]

    V2X-DGW: domain generalization for multi-agent percep- tion under adverse weather conditions

    Baolu Li, Jinlong Li, Xinyu Liu, Runsheng Xu, Zhengzhong Tu, Jiacheng Guo, Xiaopeng Li, and Hongkai Yu. V2X-DGW: domain generalization for multi-agent percep- tion under adverse weather conditions. arXiv preprint arXiv:2403.11371, 2024. 1

    2024 arXiv

  14. [19]

    Learning domain-aware detection head with prompt tuning

    Haochen Li, Rui Zhang, Hantao Yao, Xinkai Song, Yifan Hao, Yongwei Zhao, Ling Li, and Yunji Chen. Learning domain-aware detection head with prompt tuning. In Ad- vances in Neural Information Processing Systems (NeurIPS), pages 4248–4262, 2023. 3

    2023

  15. [20]

    Towards vehicle-to-everything autonomous driving: A survey on collaborative perception

    Si Liu, Chen Gao, Yuan Chen, Xingyu Peng, Xianghao Kong, Kun Wang, Runsheng Xu, Wentao Jiang, Hao Xiang, Jiaqi Ma, and Miao Wang. Towards vehicle-to-everything autonomous driving: A survey on collaborative perception. arXiv preprint arXiv:2308.16714, 2023. 2

    2023 arXiv

  16. [21]

    When2com: Multi-agent perception via communi- cation graph grouping

    Yen-Cheng Liu, Junjiao Tian, Nathaniel Glaser, and Zsolt Kira. When2com: Multi-agent perception via communi- cation graph grouping. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 4105–4114, 2020. 1, 3

    2020

  17. [22]

    Robust collaborative 3d object detection in presence of pose errors

    Yifan Lu, Quanhao Li, Baoan Liu, Mehrdad Dianati, Chen Feng, Siheng Chen, and Yanfeng Wang. Robust collaborative 3d object detection in presence of pose errors. In IEEE In- ternational Conference on Robotics and Automation (ICRA), pages 4812–4818, 2023. 1

    2023

  18. [24]

    Complementarity-enhanced and redundancy-minimized col- laboration network for multi-agent perception

    Guiyang Luo, Hui Zhang, Quan Yuan, and Jinglin Li. Complementarity-enhanced and redundancy-minimized col- laboration network for multi-agent perception. In Proceed- ings of the 30th ACM International Conference on Multime- dia (ACM MM), page 3578–3586, 2022. 3

    2022

  19. [25]

    Plug and play: A representation en- hanced domain adapter for collaborative perception

    Tianyou Luo, Quan Yuan, Guiyang Luo, Yuchen Xia, Yujia Yang, and Jinglin Li. Plug and play: A representation en- hanced domain adapter for collaborative perception. In Eu- ropean Conference on Computer Vision (ECCV), pages 287– 303, 2025. 1, 2, 3, 6, 7

    2025

  20. [26]

    Prod: Prompting-to-disentangle domain knowledge for cross-domain few-shot image classification

    Tianyi Ma, Yifan Sun, Zongxin Yang, and Yi Yang. Prod: Prompting-to-disentangle domain knowledge for cross-domain few-shot image classification. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition (CVPR), pages 19754–19763, 2023. 2

    2023

  21. [27]

    Interruption-aware cooperative perception for V2X communication-aided au- tonomous driving

    Shunli Ren, Zixing Lei, Zi Wang, Mehrdad Dianati, Yafei Wang, Siheng Chen, and Wenjun Zhang. Interruption-aware cooperative perception for V2X communication-aided au- tonomous driving. IEEE Trans. Intell. Veh., 9(4):4698–4714,

  22. [29]

    Ad-clip: Adapting domains in prompt space using clip

    Mainak Singha, Harsh Pal, Ankit Jha, and Biplab Banerjee. Ad-clip: Adapting domains in prompt space using clip. In 2023 IEEE/CVF International Conference on Computer Vi- sion Workshops (ICCVW), pages 4357–4366, 2023. 3

    2023

  23. [30]

    Mingxing Tan and Quoc V . Le. Efficientnet: Rethink- ing model scaling for convolutional neural networks. In Proceedings of Machine Learning Research (ICML) , pages 6105–6114, 2019. 1, 2

    2019

  24. [31]

    Collaborative 3d object detection for autonomous vehicles via learnable com- munications

    Junyong Wang, Yuan Zeng, and Yi Gong. Collaborative 3d object detection for autonomous vehicles via learnable com- munications. IEEE Trans. Intell. Transp. Syst., 24(9):9804– 9816, 2023. 1

    2023

  25. [32]

    Cbam: Convolutional block attention module

    Sanghyun Woo, Jongchan Park, Joon-Young Lee, and In So Kweon. Cbam: Convolutional block attention module. In European Conference on Computer Vision (ECCV) , page 3–19, 2018. 3

    2018

  26. [33]

    Hm-vit: Hetero- modal vehicle-to-vehicle cooperative perception with vision transformer

    Hao Xiang, Runsheng Xu, and Jiaqi Ma. Hm-vit: Hetero- modal vehicle-to-vehicle cooperative perception with vision transformer. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , pages 284–295,

  27. [35]

    OPV2V: an open benchmark dataset and fusion pipeline for perception with vehicle-to-vehicle communica- tion

    Runsheng Xu, Hao Xiang, Xin Xia, Xu Han, Jinlong Li, and Jiaqi Ma. OPV2V: an open benchmark dataset and fusion pipeline for perception with vehicle-to-vehicle communica- tion. In 2022 IEEE International Conference on Robotics and Automation (ICRA), pages 2583–2589, 2022. 6

    2022

  28. [36]

    Bridging the domain gap for multi-agent percep- tion

    Runsheng Xu, Jinlong Li, Xiaoyu Dong, Hongkai Yu, and Jiaqi Ma. Bridging the domain gap for multi-agent percep- tion. In 2023 IEEE International Conference on Robotics and Automation (ICRA), pages 6035–6042, 2023. 1, 3, 6, 7, 2

    2023

  29. [37]

    Cobevt: Cooperative bird’s eye view se- mantic segmentation with sparse transformers

    Runsheng Xu, Zhengzhong Tu, Hao Xiang, Wei Shao, Bolei Zhou, and Jiaqi Ma. Cobevt: Cooperative bird’s eye view se- mantic segmentation with sparse transformers. In Proceed- ings of The 6th Conference on Robot Learning (CoRL), pages 989–1000, 2023. 2, 5, 7, 3

    2023

  30. [38]

    SECOND: sparsely em- bedded convolutional detection.Sensors, 18(10):3337, 2018

    Yan Yan, Yuxing Mao, and Bo Li. SECOND: sparsely em- bedded convolutional detection.Sensors, 18(10):3337, 2018. 6, 7, 1, 2, 3

    2018

  31. [39]

    Spatio-temporal domain awareness for multi-agent collab- orative perception

    Kun Yang, Dingkang Yang, Jingyu Zhang, Mingcheng Li, Yang Liu, Jing Liu, Hanqi Wang, Peng Sun, and Liang Song. Spatio-temporal domain awareness for multi-agent collab- orative perception. In Proceedings of the IEEE/CVF In- ternational Conference on Computer Vision (ICCV) , pag...

    2023

  32. [40]

    Exploring sparse visual prompt for domain adaptive dense prediction

    Senqiao Yang, Jiarui Wu, Jiaming Liu, Xiaoqi Li, Qizhe Zhang, Mingjie Pan, Yulu Gan, Zehui Chen, and Shanghang Zhang. Exploring sparse visual prompt for domain adaptive dense prediction. In AAAI Conference on Artificial Intelli- gence (AAAI), pages 16334–16342, 2024. 3

    2024

  33. [41]

    V2vformer++: Multi-modal vehicle-to-vehicle cooperative perception via global-local transformer

    Hongbo Yin, Daxin Tian, Chunmian Lin, Xuting Duan, Jian- shan Zhou, Dezong Zhao, and Dongpu Cao. V2vformer++: Multi-modal vehicle-to-vehicle cooperative perception via global-local transformer. IEEE Trans. Intell. Transp. Syst. , 25(2):2153–2166, 2024. 3

    2024

  34. [42]

    DAIR-V2X: A large-scale dataset for vehicle-infrastructure cooperative 3d object detection

    Haibao Yu, Yizhen Luo, Mao Shu, Yiyi Huo, Zebang Yang, Yifeng Shi, Zhenglong Guo, Hanyu Li, Xing Hu, Jirui Yuan, and Zaiqing Nie. DAIR-V2X: A large-scale dataset for vehicle-infrastructure cooperative 3d object detection. In IEEE/CVF Conference on Computer Vision and Pattern R...

    2022

  35. [43]

    Bm2cp: Effi- cient collaborative perception with lidar-camera modalities

    Binyu Zhao, Wei Zhang, and Zhaonian Zou. Bm2cp: Effi- cient collaborative perception with lidar-camera modalities. In Proceedings of The 7th Conference on Robot Learning (CoRL), pages 1022–1035, 2023. 3

    2023

  36. [44]

    V oxelnet: End-to-end learning for point cloud based 3d object detection

    Yin Zhou and Oncel Tuzel. V oxelnet: End-to-end learning for point cloud based 3d object detection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 4490–4499, 2018. 4, 6, 7, 1, 2, 3 One is Plenty: A Polymorphic Feature Interpr...

    2018

  37. [45]

    The resizer consists of a max-pooling layer and a 1 × 1 convolution

    Experimental Details When adapting to a new neighbor agent, trainable param- eters include a specific prompt for the neighbor agent and a resizer to align the size of the neighbor features with the ego features. The resizer consists of a max-pooling layer and a 1 × 1 convoluti...

  38. [46]

    Additional Ablation Study We perform ablation experiments on three loss components: style loss (regulating both shared and agent-specific seman- tics), adversary loss (regulating shared semantics), and sin- gle loss (regulating agent-specific semantics), with results detailed ...

  39. [47]

    Trainable parameter numbers of different encoders

    Performance of Multi-Modal Fusion We conduct multi-modal fusion experiments by integrating LiDAR and camera data, using two image encoders, Effi- cientNet [30] and ResNet [11], and conduct two sets of ex- Parameters (M) Encoder Feature Size Specific Prompt Resizer Total pp8 [1...

  40. [48]

    Performance Comparison on V2XSet An open dataset, V2XSet [34], is used in the comparative experiments

    New Datasets 4.1. Performance Comparison on V2XSet An open dataset, V2XSet [34], is used in the comparative experiments. Compared to the OPV2V dataset, V2XSet incorporates vehicle-to-everything cooperation and realis- tic noise simulation. The experimental results comparing Po...

  41. [49]

    We use pp8-pp4-vn4 combination in the base model training phase

    dataset. We use pp8-pp4-vn4 combination in the base model training phase. The experimental results comparing PolyInter with PnPDA [25] and MPDA [36] are presented in Table 5

  42. [50]

    Additional Experiments 5.1. Three-agent Collaborative Perception We compare the performance of PolyInter with PnPDA [25] and MPDA [36] in the immutable heterogeneous scenario of three-agent collaborative perception, as shown in Table

  43. [51]

    ego-neb1-neb2

    The three collaborating agents are set as three different agent types, in the format of “ego-neb1-neb2”. The selected scenarios include pp8-pp4-vn6, pp8-pp4-sd1, and pp8-vn6- sd1. In three-agent collaboration, the ego agent’s interpreter separately interprets the heterogeneous...

  44. [52]

    Taking pp8 as the ego agent, pp4, vn4, and sd1 were sequentially se- lected as neighbor agents

    Additional Qualitative Evaluation As shown in Figure 1, the specific prompts and features for different neighbor agents are visualized. Taking pp8 as the ego agent, pp4, vn4, and sd1 were sequentially se- lected as neighbor agents. Features of different heteroge- neous neighbo...

Pith tools

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