Pith. sign in

REVIEW 4 major objections 6 minor 44 references

Towards Generalizable Trajectory Prediction Using Dual-Level Representation Learning And Adaptive Prompting

T0 review · 4 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

Pith's one-line read Vehicle trajectory prediction can be made more generalizable by pretraining a single Perceiver-based architecture with self-distillation and masked reconstruction, then adapting it with prompt vectors; the authors report state-of-the-art…

desk verdict A plausible SSL+Perceiver assembly for trajectory prediction whose unqualified SOTA claim is undercut by its own Argoverse 2 numbers and unverified baseline comparability. read the letter →

arxiv 2501.04815 v1 pith:2C7D35OC submitted 2025-01-08 cs.CV

classification cs.CV
keywords trajectorypredictionself-supervisedlearningmaskedreconstructionself-distillationregisterqueriesprompttuningmultimodalautonomousdriving
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

Vehicle trajectory prediction models often overfit to one dataset and require expensive post-processing to pick among many candidate futures. This paper argues that a single Perceiver-based architecture can avoid those problems when it is pretrained with two complementary self-supervised tasks at once—self-distillation for global scene context and masked reconstruction for fine-grained detail—and then adapted with small prompt vectors instead of full fine-tuning. The resulting model, PerReg+, is claimed to set the best published numbers on nuScenes, Argoverse 2, and Waymo Open Motion Dataset, with pretraining reducing B-FDE by 6.8% on smaller datasets and by 11.8% in a cross-domain test. If the claim holds, trajectory prediction can be made more generalizable and cheaper to adapt without dataset-specific architectural surgery.

What carries the argument

The load-bearing object is the Perceiver IO encoder–decoder with a fixed-size latent array $Z_{\mathrm{latent}} \in \mathbb{R}^{L \times D}$, augmented by three decoder query types: mode queries for the $M$ predicted trajectories, register queries that act as structured memory, and segment reconstruction queries for agent histories, futures, and lane polylines. Pretraining couples two mechanisms: masked self-distillation, where a teacher encoder fed unmasked inputs (including future trajectories) is aligned with a student encoder fed masked inputs through a clustering head and cross-entropy loss, and masked reconstruction, where fine-grained masking with random query dropping forces the decoder to complete whole segments. Fine-tuning freezes the architecture and optimizes only a prompt pool $P=[p_1,\dots,p_K]$, with the pretrained clustering head selecting the prompt for each scene.

What would settle it

Re-run AutoBot, MTR, and Forecast-MAE inside the exact same code path and preprocessing—same 100-meter map radius, vehicle-only filtering, and same metric script—and compare B-FDE; if the rerun baselines match or beat PerReg+, the state-of-the-art claim collapses.

Watch

Extended reading notes

Core claim

The paper's central claim is that a trajectory predictor does not need a bespoke architecture per dataset or clustering and suppression to handle multimodal futures. PerReg+ is a Perceiver IO model whose encoder is pretrained by having a teacher encoder process unmasked inputs, including future trajectories, while a student encoder processes heavily masked inputs; a cross-entropy loss aligns the two, and a decoder reconstructs entire past and future trajectory segments and lane segments from masked tokens. During fine-tuning the main architecture is frozen and only a small prompt pool, selected by the pretrained clustering head, is optimized. On the three benchmarks the pretrained variant is reported to outperform AutoBot, MTR, and Forecast-MAE on most metrics, and transferring from Waymo Open Motion Dataset to nuScenes reduces B-FDE by 11.8% relative to its non-pretrained counterpart.

Load-bearing premise

The measured advantage depends on the baseline numbers being protocol-equivalent: two baselines are taken from the benchmark paper rather than rerun, the third was adapted by the authors, and no error bars are given, so a different map range, sample filter, or metric implementation could shrink or erase the reported gaps.

Editorial extensions

If this is right

  • Pretraining helps most where data is scarce: smaller datasets gain about 6.8% in B-FDE, so the recipe is a candidate for new cities or rare scenarios.
  • Retaining the pretrained decoder and optimizing only prompts makes adapting to a new dataset cheap in parameters and compute.
  • Register queries plus mode queries remove the need for trajectory clustering and non-maximum suppression, simplifying deployment.
  • Multi-dataset training improves or maintains accuracy on all three benchmarks, suggesting one model can serve several geographies.
  • Cross-domain transfer from a large dataset to a smaller one cuts B-FDE by 11.8% relative to no pretraining, supporting large-scale pretraining as a general route to generalization.

Reading between the lines

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

  • Editorial extension: the same dual-level pretraining should improve data efficiency for other road users, such as pedestrians and cyclists, if the learned scene representation is genuinely general.
  • Editorial extension: prompt selection by clustering could be turned into a continual-learning mechanism—add a new prompt for a new city while freezing old prompts, then check whether old-domain accuracy degrades.
  • Editorial extension: the 11.8% cross-domain gain compares a model with pretraining against the same model without it; the practical question is whether the gain survives protocol-matched baselines and error bars.
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 / 6 minor

Summary. The paper proposes PerReg+, a trajectory prediction model built on the Perceiver IO architecture. It combines self-distillation and masked reconstruction for representation learning, adds register queries to the decoder for multimodal prediction, and uses prompt tuning during fine-tuning. Experiments are conducted on nuScenes, Argoverse 2, and WOMD within the UniTraj framework, comparing single-dataset and multi-dataset training, and including out-of-domain transfer from WOMD to nuScenes. The abstract claims state-of-the-art results on all three benchmarks and reports a 6.8% pretraining improvement and an 11.8% cross-domain B-FDE reduction.

Significance. If the empirical results held, the paper would offer a useful recipe for pretraining and fine-tuning trajectory prediction models that avoids clustering/NMS and adapts efficiently to new datasets. The ablations, scalability experiments, and cross-domain evaluation are valuable. However, the headline SOTA claim is not supported by the paper's own Table 1 for Argoverse 2, and the abstract's pretraining improvement figure (6.8%) is inconsistent with Section 4.2 and the numbers in Table 1. The protocol under which baselines were evaluated is only partially described, and no error bars are provided. Because the central claims are empirical, these issues need to be resolved before the results can be accepted as stated.

major comments (4)
  1. [Table 1, Argoverse 2 rows] The abstract claims that PerReg+ 'sets a new state-of-the-art performance on nuScenes, Argoverse 2, and Waymo Open Motion Dataset (WOMD).' In Table 1, single-dataset training on Argoverse 2 shows Forecast-MAE+ achieving B-FDE 2.05, minADE 0.74, minFDE 1.43, and MR 0.19, while PerReg+ achieves 2.07, 0.77, 1.46, and 0.21. PerReg+ is worse on every metric. In multi-dataset training, MTR achieves B-FDE 1.99 versus PerReg+'s 2.02, and other baselines also beat PerReg+ on some metrics. The SOTA claim for Argoverse 2 is therefore contradicted by the paper's own table. Please either correct the claim or re-evaluate to identify a setting in which PerReg+ is actually state-of-the-art.
  2. [Abstract vs. Section 4.2 and Table 1] The abstract states that pretraining 'reduces the error by 6.8% on smaller datasets.' Section 4.2 reports improvements of 11% on nuScenes, 13% on Argoverse 2, and 2.4% on WOMD. From Table 1, the B-FDE reductions for PerReg to PerReg+ are (3.06 to 2.62) = 14.4%, (2.38 to 2.07) = 13.0%, and (2.10 to 2.05) = 2.4%. None of these is 6.8%. The paper needs to reconcile the abstract, Section 4.2, and Table 1, and to specify exactly which comparison produces the 6.8% figure.
  3. [Section 7 and Section 4.1] The SOTA claim depends on direct comparison with baselines, but Section 7 says MTR and AutoBot results are 'sourced directly from the UniTraj paper' and Forecast-MAE was 'adapted' to the UniTraj framework. Section 4.1 states the map range is 100m and only vehicle trajectories are used, yet it is not established that the UniTraj baselines were run under identical filtering, map range, prediction horizon (2s history / 6s future), and metric code. The paper also reports no error bars or multiple seeds. Please provide a detailed protocol-equivalence statement, or re-run the baselines under the same protocol, and restrict any SOTA claim to the settings that are actually controlled.
  4. [Section 4.4 and Table 3] The ablation text says Masked SD (+MSD) 'lowers B-FDE to 2.76', but Table 3 reports 2.64 for +MSD. Additionally, the caption of Table 3 describes '+SR' as 'self-distillation (SR)', while the text and table row use '+SR' to refer to segment-level reconstruction. These inconsistencies make it difficult to interpret which component produces which gain and should be corrected.
minor comments (6)
  1. [Abstract] 'Remarkable, our pretrained model...' should read 'Remarkably, our pretrained model...'.
  2. [Throughout] The model is called PerReg+ in the title and abstract, but Table 1 and Section 4.2 use 'PerReg (Ours)' with a '+' marker for pretrained variants. Please standardize the notation, for example by writing 'PerReg+' and 'PerReg (no pretraining)'.
  3. [Section 4.2] 'A V2' is written with a space in several places; use 'Argoverse 2' or 'AV2' consistently.
  4. [Section 4.4] The description of the baseline as 'all decoder queries are used for prediction and aggregated via NMS' is unclear. Please clarify whether NMS is used in the baseline and how the final PerReg+ model eliminates the need for clustering and suppression, since this is one of the paper's stated contributions.
  5. [Section 3.4 and supplementary Section 6.1] Section 3.4 defines three losses (Ldistill, LGMM, Lrecon) in Equation (8), but the supplementary describes six losses for DWA. Please explain how the six losses map to the three terms in Equation (8), including the role of the KoLeo regularization term.
  6. [Section 2.2] The statement that 'only Forecast-PEFT keeps the pretrained decoder during finetuning' is a strong claim; please support it with a broader citation search or soften it to 'to the best of our knowledge' in a way that is clearly scoped.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's claims are empirical benchmark results, not a derivation that reduces to its own inputs.

full rationale

This is an empirical systems paper; there is no closed-form derivation whose output could equal its input. The training objectives (Eqs. 5-8) are standard self-distillation, GMM prediction, and reconstruction losses, and the reported metrics (B-FDE, minADE, minFDE, MR) are external benchmark measurements on fixed datasets. The self-citations to UniTraj [20] and Forecast-PEFT [42] provide an evaluation framework and a related-work precedent, but neither encodes the target metrics nor forces the reported gains by construction; baseline numbers are quoted from UniTraj or re-run by the authors (Section 7), which raises protocol-comparability risk but is not circularity. Likewise, the Table 1 result on Argoverse 2 single-dataset training (PerReg+ 2.07 vs. Forecast-MAE+ 2.05 B-FDE) weakens the unqualified 'state-of-the-art' claim, but that is a correctness/precision concern, not a circular-derivation concern. No fitted parameter is renamed as a prediction, and no load-bearing argument reduces to a self-citation chain. Score 0.

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

The paper makes no analytic derivation; its central claims rest on benchmark numbers. The key entries are hand-picked pretraining hyperparameters and assumptions about transferability of self-distilled and clustered representations. No new physical or symbolic entities are introduced.

free parameters (5)
  • Masking ratios (history 90%, future 97%, map 75%) = 90/97/75
    Chosen by hand; no sensitivity analysis is reported beyond the reconstruction query drop ratio. These ratios control pretraining difficulty and are likely load-bearing for the reported gains.
  • Reconstruction query drop ratio = 40%
    Tuned via ablation on nuScenes (Figure 4); 40% gives the best B-FDE, so the headline results depend on this hand-picked value.
  • Prompt pool size K = unspecified
    The number of cluster-specific prompts is not reported, despite being a core component of the adaptive prompt tuning claim.
  • Number of register queries N_R = unspecified
    The register query count is central to the multimodality design but is never specified.
  • DWA bounds and bias weights = unspecified
    Dynamic Weighted Aggregation uses clipping bounds [w_min, w_max] and biases that are not quantified in the paper.
assumptions (4)
  • domain assumption The teacher encoder may use ground-truth future trajectories during pretraining without leaking this information to the student at inference time.
    Section 3.2: "By granting the teacher model access to future trajectory information..." This transfer of self-distilled representations to a future-unavailable inference setting is assumed to hold.
  • domain assumption The clustering head trained with distilled logits produces semantically stable clusters across datasets.
    Section 3.5 and Eq. (15): cluster selection via argmax of MLP logits is assumed to pick appropriate prompts on unseen data; no cluster-quality analysis is given.
  • domain assumption Fine-grained masking plus segment-level reconstruction is a valid surrogate task for trajectory prediction.
    Sections 3.2-3.3: the connection between masked reconstruction and downstream forecasting accuracy is assumed, supported only by ablations, not by theory.
  • standard math Multi-head attention, EMA teacher updates, and GMM maximum-likelihood losses are used as standard mathematical machinery.
    The paper relies on standard transformer and self-supervised learning machinery without proving it; this is conventional in the field.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Towards Generalizable Trajectory Prediction Using Dual-Level Representation Learning And Adaptive Prompting." pith.science (2026). https://pith.science/paper/2C7D35OC

@misc{pith2026250104815,
  author       = {Pith},
  title        = {Pith review of: Towards Generalizable Trajectory Prediction Using Dual-Level Representation Learning And Adaptive Prompting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2C7D35OC}},
  note         = {Machine review of arXiv:2501.04815}
}
read the original abstract

Existing vehicle trajectory prediction models struggle with generalizability, prediction uncertainties, and handling complex interactions. It is often due to limitations like complex architectures customized for a specific dataset and inefficient multimodal handling. We propose Perceiver with Register queries (PerReg+), a novel trajectory prediction framework that introduces: (1) Dual-Level Representation Learning via Self-Distillation (SD) and Masked Reconstruction (MR), capturing global context and fine-grained details. Additionally, our approach of reconstructing segmentlevel trajectories and lane segments from masked inputs with query drop, enables effective use of contextual information and improves generalization; (2) Enhanced Multimodality using register-based queries and pretraining, eliminating the need for clustering and suppression; and (3) Adaptive Prompt Tuning during fine-tuning, freezing the main architecture and optimizing a small number of prompts for efficient adaptation. PerReg+ sets a new state-of-the-art performance on nuScenes [1], Argoverse 2 [2], and Waymo Open Motion Dataset (WOMD) [3]. Remarkable, our pretrained model reduces the error by 6.8% on smaller datasets, and multi-dataset training enhances generalization. In cross-domain tests, PerReg+ reduces B-FDE by 11.8% compared to its non-pretrained variant.

Figures

Figures reproduced from arXiv: 2501.04815 by the authors.

Figure 1
Figure 1. Comparison between Masked Autoencoder (MAE) and our pretraining strategy with PerReg+. Our approach incorporates dual-level representation learning through Self-Distillation (SD) and Masked Reconstruction (MR). Unlike MAE, which reconstructs each masked point independently, our method uses segment-level queries to reconstruct entire trajectories (past or future) and lane segments, enabling the decoder to infer compl… view at source ↗
Figure 2
Figure 2. Overview of the Proposed Trajectory Prediction Model. The model combines agent history and road in￾formation into a unified scene representation, processed by the Perceiver IO architecture to predict future trajectories. In the SD process, teacher and student Perceiver encoders operate in coordination: the teacher encoder (updated via Exponential Moving Average) processes unmasked data with future trajectories, gene… view at source ↗
Figure 3
Figure 3. Comparative Predictions of PerReg in Different Pretraining Settings. We illustrate three trajectory pre￾diction settings for our model (PerReg) at a complex intersection (nuscenes dataset), showing its performance without pretraining, with single-dataset pretraining, and with multi-dataset pretraining. In each setting, the target vehicle (pur￾ple to blue predicted trajectory) is positioned in the leftmost lane with … view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Impact of Reconstruction Query Drop Ratio. B-FDE for PerReg+ on the nuScenes dataset across varying reconstruction query drop ratios. References [1] Holger Caesar, Varun Bankiti, Alex H Lang, Sourabh Vora, Venice Erin Liong, Qiang Xu, Anush Krishnan, Yu Pan, Giancarlo …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 36 canonical work pages

  1. [1]

    nuscenes: A multimodal dataset for autonomous driving

    Holger Caesar, Varun Bankiti, Alex H Lang, Sourabh V ora, Venice Erin Liong, Qiang Xu, Anush Krishnan, Yu Pan, Giancarlo Baldan, and Oscar Beijbom. nuscenes: A multimodal dataset for autonomous driving. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 11621–11631, 2020. 1, 2, 7

  2. [2]

    Argoverse 2: Next generation datasets for self-driving perception and forecasting

    Benjamin Wilson, William Qi, Tanmay Agarwal, John Lambert, Jagjeet Singh, Siddhesh Khandelwal, Bowen Pan, Ratnesh Kumar, Andrew Hartnett, Jhony Kaesemodel Pontes, et al. Argoverse 2: Next generation datasets for self-driving perception and forecasting. arXiv preprint arXiv:2301.00493, 2023. 1, 2, 7

  3. [3]

    Large scale interactive motion forecasting for autonomous driving: The waymo open motion dataset

    Scott Ettinger, Shuyang Cheng, Benjamin Caine, Chenxi Liu, Hang Zhao, Sabeek Pradhan, Yuning Chai, Ben Sapp, Charles R Qi, Yin Zhou, et al. Large scale interactive motion forecasting for autonomous driving: The waymo open motion dataset. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 9710–9719, 2021. 1, 2, 7, 3

  4. [4]

    Pretram: Self-supervised pre-training via connecting trajectory and map

    Chenfeng Xu, Tian Li, Chen Tang, Lingfeng Sun, Kurt Keutzer, Masayoshi Tomizuka, Alireza Fathi, and Wei Zhan. Pretram: Self-supervised pre-training via connecting trajectory and map. In ECCV, pages 34–50. Springer, 2022. 1, 3

  5. [5]

    Sept: Towards efficient scene representation learning for motion prediction

    Zhiqian Lan, Yuxuan Jiang, Yao Mu, Chen Chen, and Shengbo Eben Li. Sept: Towards efficient scene representation learning for motion prediction. arXiv preprint arXiv:2309.15289, 2023. 1, 3

  6. [6]

    Forecast-MAE: Self-supervised pre-training for motion forecasting with masked autoencoders

    Jie Cheng, Xiaodong Mei, and Ming Liu. Forecast-MAE: Self-supervised pre-training for motion forecasting with masked autoencoders. Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023. 1, 3, 6, 7, 8, 2

  7. [7]

    Dyset: A dynamic masked self-distillation approach for robust trajectory prediction

    Mozhgan Pourkeshavarz, Junrui Zhang, and Amir Rasouli. Dyset: A dynamic masked self-distillation approach for robust trajectory prediction. In Ale ˇs Leonardis, Elisa Ricci, Stefan Roth, Olga Russakovsky, Torsten Sattler, and G¨ul Varol, editors, Computer Vision – ECCV 2024, pages 324–342, Cham, 2025. Springer Nature Switzerland. 1, 3, 6

  8. [8]

    Smartpretrain: Model-agnostic and dataset-agnostic representation learning for motion prediction, 10 2024

    Yang Zhou, Hao Shao, Letian Wang, Steven Waslander, Hongsheng Li, and Yu Liu. Smartpretrain: Model-agnostic and dataset-agnostic representation learning for motion prediction, 10 2024. 1, 3

Show all 44 references
  1. [9]

    SSL-lanes: Self-supervised learning for motion fore- casting in autonomous driving

    Prarthana Bhattacharyya, Chengjie Huang, and Krzysztof Czarnecki. SSL-lanes: Self-supervised learning for motion fore- casting in autonomous driving. In 6th Annual Conference on Robot Learning, 2022. 1

  2. [10]

    Social nce: Contrastive learning of socially-aware motion representations

    Yuejiang Liu, Qi Yan, and Alexandre Alahi. Social nce: Contrastive learning of socially-aware motion representations. In ICCV, pages 15118–15129, 2021. 1, 3

  3. [11]

    Perceiver: General percep- tion with iterative attention

    Andrew Jaegle, Felix Gimeno, Andy Brock, Oriol Vinyals, Andrew Zisserman, and Joao Carreira. Perceiver: General percep- tion with iterative attention. In International conference on machine learning, pages 4651–4664. PMLR, 2021. 1

  4. [12]

    Perceiver io: A general architecture for structured inputs & outputs

    Andrew Jaegle, Sebastian Borgeaud, Jean-Baptiste Alayrac, Carl Doersch, Catalin Ionescu, David Ding, Skanda Koppula, Daniel Zoran, Andrew Brock, Evan Shelhamer, et al. Perceiver io: A general architecture for structured inputs & outputs. arXiv preprint arXiv:2107.14795, 2021. 1

  5. [13]

    Hivt: Hierarchical vector transformer for multi-agent motion prediction

    Zikang Zhou, Luyao Ye, Jianping Wang, Kui Wu, and Kejie Lu. Hivt: Hierarchical vector transformer for multi-agent motion prediction. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 8823–8833,

  6. [14]

    Motion transformer with global intention localization and local movement refinement

    Shaoshuai Shi, Li Jiang, Dengxin Dai, and Bernt Schiele. Motion transformer with global intention localization and local movement refinement. Advances in Neural Information Processing Systems, 35:6531–6543, 2022. 1, 3, 7, 8, 2

  7. [15]

    Hpnet: Dynamic trajectory forecasting with historical prediction attention

    Xiaolong Tang, Meina Kan, Shiguang Shan, Zhilong Ji, Jinfeng Bai, and Xilin Chen. Hpnet: Dynamic trajectory forecasting with historical prediction attention. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 15261–15270, 2024. 1, 3

  8. [16]

    Mtr++: Multi-agent motion prediction with symmetric scene modeling and guided intention querying

    Shaoshuai Shi, Li Jiang, Dengxin Dai, and Bernt Schiele. Mtr++: Multi-agent motion prediction with symmetric scene modeling and guided intention querying. IEEE Trans. Pattern Anal. Mach. Intell., 46(5):3955–3971, January 2024. 1

  9. [17]

    Wayformer: Motion forecasting via simple & efficient attention networks

    Nigamaa Nayakanti, Rami Al-Rfou, Aurick Zhou, Kratarth Goel, Khaled S Refaat, and Benjamin Sapp. Wayformer: Motion forecasting via simple & efficient attention networks. In IEEE International Conference on Robotics and Automation (ICRA), pages 2980–2987. IEEE, 2023. 1, 3, 7 10...

  10. [18]

    Motionlm: Multi-agent motion forecasting as language modeling

    Ari Seff, Brian Cera, Dian Chen, Mason Ng, Aurick Zhou, Nigamaa Nayakanti, Khaled S Refaat, Rami Al-Rfou, and Ben- jamin Sapp. Motionlm: Multi-agent motion forecasting as language modeling. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 8579–...

  11. [19]

    Bootstrap your own latent-a new approach to self-supervised learning

    Jean-Bastien Grill, Florian Strub, Florent Altch ´e, Corentin Tallec, Pierre Richemond, Elena Buchatskaya, Carl Doersch, Bernardo Avila Pires, Zhaohan Guo, Mohammad Gheshlaghi Azar, et al. Bootstrap your own latent-a new approach to self-supervised learning. Advances in neural...

  12. [20]

    Unitraj: A unified framework for scalable vehicle trajectory prediction

    Lan Feng, Mohammadhossein Bahari, Kaouther Messaoud Ben Amor, ´Eloi Zablocki, Matthieu Cord, and Alexandre Alahi. Unitraj: A unified framework for scalable vehicle trajectory prediction. arXiv preprint arXiv:2403.15098, 2024. 2, 7

  13. [21]

    Vision transformers need registers

    Timoth ´ee Darcet, Maxime Oquab, Julien Mairal, and Piotr Bojanowski. Vision transformers need registers. In The Twelfth International Conference on Learning Representations, 2024. 2

  14. [22]

    Multimodal trajectory prediction conditioned on lane-graph traversals

    Nachiket Deo, Eric Wolff, and Oscar Beijbom. Multimodal trajectory prediction conditioned on lane-graph traversals. In 5th Annual Conference on Robot Learning, 2021. 3

  15. [23]

    Multimodal trajectory predictions for autonomous driving using deep convolutional networks

    Henggang Cui, Vladan Radosavljevic, Fang Chou, Tsung Lin, Thi Nguyen, Tzu Huang, Jeff Schneider, and Nemanja Djuric. Multimodal trajectory predictions for autonomous driving using deep convolutional networks. ICRA, 2019. 3

  16. [24]

    Latent variable sequential set transformers for joint multi-agent motion prediction

    Roger Girgis, Florian Golemo, Felipe Codevilla, Martin Weiss, Jim Aldon D’Souza, Samira Ebrahimi Kahou, Felix Heide, and Christopher Pal. Latent variable sequential set transformers for joint multi-agent motion prediction. In International Conference on Learning Representation...

  17. [25]

    Scene transformer: A unified architecture for predicting future trajectories of multiple agents

    Jiquan Ngiam, Vijay Vasudevan, Benjamin Caine, Zhengdong Zhang, Hao-Tien Lewis Chiang, Jeffrey Ling, Rebecca Roelofs, Alex Bewley, Chenxi Liu, Ashish Venugopal, et al. Scene transformer: A unified architecture for predicting future trajectories of multiple agents. In Internati...

  18. [26]

    Multipath: Multiple probabilistic anchor trajectory hypotheses for behavior prediction

    Yuning Chai, Benjamin Sapp, Mayank Bansal, and Dragomir Anguelov. Multipath: Multiple probabilistic anchor trajectory hypotheses for behavior prediction. pages 86–99, 2020. 3

  19. [27]

    Trajectory prediction for autonomous driving based on multi-head attention with joint agent-map representation

    Kaouther Messaoud, Nachiket Deo, Mohan M Trivedi, and Fawzi Nashashibi. Trajectory prediction for autonomous driving based on multi-head attention with joint agent-map representation. In 2021 IEEE Intelligent Vehicles Symposium (IV), pages 165–170. IEEE, 2021. 3

  20. [28]

    Real-time motion prediction via hetero- geneous polyline transformer with relative pose encoding

    Zhejun Zhang, Alexander Liniger, Christos Sakaridis, Fisher Yu, and Luc Van Gool. Real-time motion prediction via hetero- geneous polyline transformer with relative pose encoding. In Advances in Neural Information Processing Systems (NeurIPS),

  21. [29]

    Vectornet: Encoding hd maps and agent dynamics from vectorized representation

    Jiyang Gao, Chen Sun, Hang Zhao, Yi Shen, Dragomir Anguelov, Congcong Li, and Cordelia Schmid. Vectornet: Encoding hd maps and agent dynamics from vectorized representation. In CVPR, pages 11525–11533, 2020. 3

  22. [30]

    Lanercnn: Distributed representations for graph-centric motion forecasting

    Wenyuan Zeng, Ming Liang, Renjie Liao, and Raquel Urtasun. Lanercnn: Distributed representations for graph-centric motion forecasting. 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 532–539, 2021. 3

  23. [31]

    Adaptive trajectory prediction via transferable gnn

    Yi Xu, Lichen Wang, Yizhou Wang, and Yun Fu. Adaptive trajectory prediction via transferable gnn. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 6520–6531, 2022. 3

  24. [32]

    Multimodal motion prediction with stacked transformers

    Yicheng Liu, Jinghuai Zhang, Liangji Fang, Qinhong Jiang, and Bolei Zhou. Multimodal motion prediction with stacked transformers. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 7577–7586,

  25. [33]

    Latent variable sequential set transformers for joint multi-agent motion prediction

    Roger Girgis, Florian Golemo, Felipe Codevilla, Martin Weiss, Jim Aldon D’Souza, Samira Ebrahimi Kahou, Felix Heide, and Christopher Pal. Latent variable sequential set transformers for joint multi-agent motion prediction. In International Conference on Learning Representation...

  26. [34]

    Query-centric trajectory prediction

    Zikang Zhou, Jianping Wang, Yung-Hui Li, and Yu-Kai Huang. Query-centric trajectory prediction. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023. 3

  27. [35]

    Learning lane graph representations for motion forecasting

    Ming Liang, Bin Yang, Rui Hu, Yun Chen, Renjie Liao, Song Feng, and Raquel Urtasun. Learning lane graph representations for motion forecasting. In Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part II 16, pages 541–556. Spri...

  28. [36]

    Hdgt: Heterogeneous driving graph transformer for multi-agent trajectory prediction via scene encoding

    Xiaosong Jia, Penghao Wu, Li Chen, Yu Liu, Hongyang Li, and Junchi Yan. Hdgt: Heterogeneous driving graph transformer for multi-agent trajectory prediction via scene encoding. IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 2023. 3

  29. [37]

    Multipath++: Efficient information fusion and trajectory aggregation for behavior prediction

    Balakrishnan Varadarajan, Ahmed Hefny, Avikalp Srivastava, Khaled S Refaat, Nigamaa Nayakanti, Andre Cornman, Kan Chen, Bertrand Douillard, Chi Pang Lam, Dragomir Anguelov, et al. Multipath++: Efficient information fusion and trajectory aggregation for behavior prediction. In2...

  30. [38]

    Adapt: Efficient multi-agent trajectory prediction with adaptation

    G ¨orkay Aydemir, Adil Kaan Akan, and Fatma G ¨uney. Adapt: Efficient multi-agent trajectory prediction with adaptation. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 8295–8305, 2023. 3

  31. [39]

    Densetnt: End-to-end trajectory prediction from dense goal sets

    Junru Gu, Chen Sun, and Hang Zhao. Densetnt: End-to-end trajectory prediction from dense goal sets. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 15303–15312, 2021. 3 11 A PREPRINT - JANUARY 8, 2025

  32. [40]

    Traj- mae: Masked autoencoders for trajectory prediction

    Hao Chen, Jiaze Wang, Kun Shao, Furui Liu, Jianye Hao, Chenyong Guan, Guangyong Chen, and Pheng-Ann Heng. Traj- mae: Masked autoencoders for trajectory prediction. arXiv preprint arXiv:2303.06697, 2023. 3, 6

  33. [41]

    Encoder and decoder, not one less for pre-trained language model sponsored nmt

    Sufeng Duan and Hai Zhao. Encoder and decoder, not one less for pre-trained language model sponsored nmt. In Findings of the Association for Computational Linguistics: ACL 2023, pages 3602–3613, 2023. 3, 1

  34. [42]

    Forecast-peft: Parameter-efficient fine- tuning for pre-trained motion forecasting models

    Jifeng Wang, Kaouther Messaoud, Yuejiang Liu, Juergen Gall, and Alexandre Alahi. Forecast-peft: Parameter-efficient fine- tuning for pre-trained motion forecasting models. arXiv preprint arXiv:2407.19564, 2024. 3

  35. [43]

    Visual prompt tuning

    Menglin Jia, Luming Tang, Bor-Chun Chen, Claire Cardie, Serge Belongie, Bharath Hariharan, and Ser-Nam Lim. Visual prompt tuning. In ECCV, pages 709–727. Springer, 2022. 3

  36. [44]

    Parameter-efficient tuning on layer normalization for pre-trained language models

    Wang Qi, Yu-Ping Ruan, Yuan Zuo, and Taihao Li. Parameter-efficient tuning on layer normalization for pre-trained language models. arXiv preprint arXiv:2211.08682, 2022. 3 12 A PREPRINT - JANUARY 8, 2025 This supplementary document provides additional insights and experiments ...

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.