pith. machine review for the scientific record. sign in

arxiv: 2604.18210 · v1 · submitted 2026-04-20 · 💻 cs.AI · cs.LG· cs.MA

Recognition: unknown

TacticGen: Grounding Adaptable and Scalable Generation of Football Tactics

Sheng Xu , Guiliang Liu , Tarak Kharrat , Yudong Luo , Mohamed Aloulou , Javier L\'opez Pe\~na , Konstantin Sofeikov , Adam Reid
show 7 more authors
Paul Roberts Steven Spencer Joe Carnall Ian McHale Oliver Schulte Hongyuan Zha Wei-Shi Zheng
Authors on Pith no claims yet

Pith reviewed 2026-05-10 04:08 UTC · model grok-4.3

classification 💻 cs.AI cs.LGcs.MA
keywords football tacticsmulti-agent diffusiontrajectory predictionclassifier guidancegenerative modelsports analyticstactic generationplayer trajectories
0
0 comments X

The pith

TacticGen steers a multi-agent diffusion model at inference time to generate football tactics matching user-specified strategic goals.

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

The paper presents TacticGen as a way to move beyond forecasting likely player paths to actively designing tactics that serve explicit objectives in football. It trains a diffusion-based transformer on millions of real match events and tracking frames to model cooperative and competitive player movements. The model first reaches state-of-the-art accuracy on trajectory prediction, then adds a classifier-guidance step that lets objectives be supplied as rules, text, or other networks. If the approach holds, coaches gain a tool that produces concrete movement sequences tailored to a given game situation rather than generic predictions. Expert review is offered as evidence that the outputs remain realistic enough for professional use.

Core claim

TacticGen formulates tactics as sequences of multi-agent movements and interactions conditioned on game context. It employs a multi-agent diffusion transformer with agent-wise self-attention and context-aware cross-attention to capture cooperative and competitive dynamics among players and the ball. Trained with over 3.3 million events and 100 million tracking frames from top-tier leagues, TacticGen achieves state-of-the-art precision in predicting player trajectories. Building on it, TacticGen enables adaptable tactic generation tailored to diverse inference-time objectives through classifier guidance mechanism, specified via rules, natural language, or neural models, with expert case study

What carries the argument

Multi-agent diffusion transformer that uses agent-wise self-attention to model interactions and context-aware cross-attention to condition on game state, allowing both accurate trajectory prediction and guided generation.

If this is right

  • The same trained model can predict player trajectories more accurately than prior methods on large-scale league data.
  • Tactics can be generated on demand by supplying objectives through simple rules, natural language prompts, or auxiliary neural classifiers.
  • The generative process scales with additional data or compute while preserving the same architecture.
  • Expert reviewers judge the resulting tactics as both realistic and strategically useful for planning.

Where Pith is reading between the lines

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

  • The guidance mechanism could be extended to other team sports that share similar multi-agent tracking data.
  • Real-time match feeds might allow the model to suggest mid-game tactical adjustments.
  • Explicit physics-based constraints could be added during sampling to reduce the chance of implausible paths.

Load-bearing premise

Classifier guidance applied at inference time will produce movement sequences that stay physically realistic and strategically coherent without creating invalid actions or artifacts.

What would settle it

A side-by-side comparison in which professional analysts or automated rule checkers flag generated trajectories for violations of physical limits such as maximum player speed or ball-interception rules at rates significantly higher than real match data.

Figures

Figures reproduced from arXiv: 2604.18210 by Adam Reid, Guiliang Liu, Hongyuan Zha, Ian McHale, Javier L\'opez Pe\~na, Joe Carnall, Konstantin Sofeikov, Mohamed Aloulou, Oliver Schulte, Paul Roberts, Sheng Xu, Steven Spencer, Tarak Kharrat, Wei-Shi Zheng, Yudong Luo.

Figure 1
Figure 1. Figure 1: Generating football tactics by modeling coordinated player movements. a This paper aims to move the field of football analytics from predictive modeling to generative tactical design, requiring models that adapt to diverse objectives. b Conditioned on the football scene context and diverse user objectives, TacticGen generates versatile tactics represented through subsequent player movements, characterized … view at source ↗
Figure 2
Figure 2. Figure 2: The proposed TacticGen framework. a Overview of the training and inference processes. During training, the ground-truth trajectories of Natt attackers, Ndef defenders, and the ball are altered by adding noise, and TacticGen learns to recover the corresponding denoised trajectories, conditioned on the context (past trajectories), the event type (e.g., pass, block, clearance), and the diffusion step. During … view at source ↗
Figure 3
Figure 3. Figure 3: Ground truth trajectories for a pass event and the corresponding best [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: 20 predicted trajectory samples by TacticGen variants for a pass event. [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: Visualizations of trajectories (top) and pitch control values (PCV) at [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 5
Figure 5. Figure 5: Trajectories generated by TacticGen for a pass event. Guidance scores [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: Trajectories generated by TacticGen for a pass event under different [PITH_FULL_IMAGE:figures/full_fig_p010_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Log-scale value heatmap [PITH_FULL_IMAGE:figures/full_fig_p010_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Visualizations of trajectories generated by TacticGen for a pass [PITH_FULL_IMAGE:figures/full_fig_p010_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Scaling performance of TacticGen across different model sizes. [PITH_FULL_IMAGE:figures/full_fig_p011_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Results of the case study on realism assessment. [PITH_FULL_IMAGE:figures/full_fig_p011_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Results of the case study on utility assessment. [PITH_FULL_IMAGE:figures/full_fig_p012_12.png] view at source ↗
read the original abstract

Success in association football relies on both individual skill and coordinated tactics. While recent advancements in spatio-temporal data and deep learning have enabled predictive analyses like trajectory forecasting, the development of tactical design remains limited. Bridging this gap is essential, as prediction reveals what is likely to occur, whereas tactic generation determines what should occur to achieve strategic objectives. In this work, we present TacticGen, a generative model for adaptable and scalable tactic generation. TacticGen formulates tactics as sequences of multi-agent movements and interactions conditioned on the game context. It employs a multi-agent diffusion transformer with agent-wise self-attention and context-aware cross-attention to capture cooperative and competitive dynamics among players and the ball. Trained with over 3.3 million events and 100 million tracking frames from top-tier leagues, TacticGen achieves state-of-the-art precision in predicting player trajectories. Building on it, TacticGen enables adaptable tactic generation tailored to diverse inference-time objectives through classifier guidance mechanism, specified via rules, natural language, or neural models. Its modeling performance is also inherently scalable. A case study with football experts confirms that TacticGen generates realistic, strategically valuable tactics, demonstrating its practical utility for tactical planning in professional football. The project page is available at: https://shengxu.net/TacticGen/.

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 / 1 minor

Summary. The paper presents TacticGen, a multi-agent diffusion transformer that models football tactics as sequences of player and ball movements conditioned on game context. It uses agent-wise self-attention and context-aware cross-attention, trains on 3.3M events and 100M tracking frames from top leagues, reports SOTA trajectory prediction accuracy, and introduces classifier guidance (via rules, language, or neural models) at inference time to enable adaptable tactic generation for diverse objectives. A case study with football experts is cited to confirm that generated tactics are realistic and strategically valuable.

Significance. If the classifier guidance mechanism can be shown to produce constraint-respecting, objective-achieving outputs without artifacts, the work would provide a scalable generative framework that bridges trajectory prediction and controllable tactical design in sports analytics, with potential applications in coaching tools and simulation.

major comments (3)
  1. [Abstract] Abstract: the claim of 'state-of-the-art precision in predicting player trajectories' is asserted without reference to specific baselines, error metrics (e.g., ADE/FDE, collision rates), ablation studies, or quantitative tables; this is load-bearing for the subsequent claim that the base model grounds the guided generation.
  2. [Abstract] Abstract, guidance paragraph: the adaptability claim rests on classifier guidance producing realistic, constraint-respecting tactics at inference time, yet the only supporting evidence cited is an expert case study; no quantitative results are provided for guided outputs (e.g., physical feasibility scores, success rate on user-specified objectives, violation rates vs. unguided sampling, or ablations isolating the guidance step).
  3. [Abstract] Abstract: the statement that 'its modeling performance is also inherently scalable' is not accompanied by scaling experiments, parameter counts, or inference-time measurements that would substantiate scalability beyond the training dataset size.
minor comments (1)
  1. [Abstract] The project page URL is given but no supplementary material (code, model weights, or additional quantitative results) is referenced in the abstract.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback on the abstract. The comments correctly identify areas where the abstract could better reference the supporting evidence in the manuscript body. We will revise the abstract accordingly while preserving its concise nature. Point-by-point responses follow.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim of 'state-of-the-art precision in predicting player trajectories' is asserted without reference to specific baselines, error metrics (e.g., ADE/FDE, collision rates), ablation studies, or quantitative tables; this is load-bearing for the subsequent claim that the base model grounds the guided generation.

    Authors: The full manuscript substantiates this claim in Section 4.1 and Table 1, which report ADE/FDE, collision rates, and comparisons against baselines including Constant Velocity, Social-LSTM, and Trajectron++. Ablations on attention mechanisms are in Section 4.3. We will revise the abstract to explicitly reference these results (e.g., 'achieves state-of-the-art precision in predicting player trajectories (Table 1)') to make the grounding clear. revision: yes

  2. Referee: [Abstract] Abstract, guidance paragraph: the adaptability claim rests on classifier guidance producing realistic, constraint-respecting tactics at inference time, yet the only supporting evidence cited is an expert case study; no quantitative results are provided for guided outputs (e.g., physical feasibility scores, success rate on user-specified objectives, violation rates vs. unguided sampling, or ablations isolating the guidance step).

    Authors: The manuscript presents quantitative results for the base trajectory model and qualitative validation of guided outputs via the expert case study in Section 5. We agree that quantitative metrics for guided generation (e.g., objective success rates and feasibility) would strengthen the adaptability claim. We will add a short quantitative analysis of guided vs. unguided outputs in the revised manuscript. revision: yes

  3. Referee: [Abstract] Abstract: the statement that 'its modeling performance is also inherently scalable' is not accompanied by scaling experiments, parameter counts, or inference-time measurements that would substantiate scalability beyond the training dataset size.

    Authors: The claim is grounded in the architecture's design for variable agent counts and its successful training on 100M frames, as described in Sections 3 and 4. We acknowledge the absence of dedicated scaling curves. We will revise the abstract to qualify the statement (e.g., 'demonstrates scalability via training on large-scale data') and include model size and inference-time details from our experiments. revision: partial

Circularity Check

0 steps flagged

No significant circularity; derivation grounded in external data and standard inference techniques

full rationale

The paper trains a multi-agent diffusion transformer on independent external datasets (3.3M events, 100M frames from top-tier leagues) to achieve SOTA trajectory prediction. Adaptable generation is performed via classifier guidance at inference time (specified by rules, language, or neural models), which is a standard conditioning method that does not reduce to the training inputs or fitted parameters by construction. The expert case study provides qualitative validation rather than a quantitative fit. No self-citations, self-definitional equations, or renamings of known results appear as load-bearing steps in the abstract or described chain; the central claims remain independently testable against held-out data and physical constraints.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on standard assumptions of diffusion models and transformer attention being sufficient to capture cooperative/competitive dynamics, plus the unstated assumption that expert qualitative judgment is a reliable proxy for strategic value. No invented physical entities; free parameters are typical diffusion and attention hyperparameters not enumerated in the abstract.

axioms (1)
  • domain assumption Multi-agent self- and cross-attention can adequately model player interactions and game context for both prediction and guided generation.
    Invoked in the model description without further justification or ablation in the abstract.

pith-pipeline@v0.9.0 · 5593 in / 1403 out tokens · 24418 ms · 2026-05-10T04:08:23.490537+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

82 extracted references · 13 canonical work pages · 8 internal anchors

  1. [1]

    The most popular sports in the world,

    A. Singh, “The most popular sports in the world,” 2025, accessed: 2025-07-14. [Online]. Available: https://www.worldatlas.com/articles/ what-are-the-most-popular-sports-in-the-world.html

    2025

  2. [2]

    Match analysis in football: a systematic review,

    H. Sarmento, R. Marcelino, M. T. Anguera, J. CampaniC ¸ o, N. Matos, and J. C. Leit ˜Ao, “Match analysis in football: a systematic review,” Journal of Sports Sciences, vol. 32, no. 20, pp. 1831–1843, 2014

    2014

  3. [3]

    Towards a foundation large events model for soccer,

    T. Mendes-Neves, L. Meireles, and J. Mendes-Moreira, “Towards a foundation large events model for soccer,”Machine Learning, vol. 113, no. 11, pp. 8687–8709, 2024

    2024

  4. [4]

    Artificial intelligence for team sports: a survey,

    R. Beal, T. J. Norman, and S. D. Ramchurn, “Artificial intelligence for team sports: a survey,”The Knowledge Engineering Review, vol. 34, p. e28, 2019

    2019

  5. [5]

    A goal scoring probability model for shots based on synchronized positional and event data in football (soccer),

    G. Anzer and P. Bauer, “A goal scoring probability model for shots based on synchronized positional and event data in football (soccer),” Frontiers in sports and active living, vol. 3, p. 624475, 2021

    2021

  6. [6]

    Accuracy and explainability of statistical and machine learning xg models in football,

    M. Cefis and M. Carpita, “Accuracy and explainability of statistical and machine learning xg models in football,”Statistics, vol. 59, no. 2, pp. 426–445, 2025

    2025

  7. [7]

    Dynamic expected threat (dxt) model: Addressing the deficit of realism in football action evaluation,

    K. Hassani, M. Ramdani, and M. Lotfi, “Dynamic expected threat (dxt) model: Addressing the deficit of realism in football action evaluation,” Applied Sciences, vol. 15, no. 8, p. 4151, 2025

    2025

  8. [8]

    Action valuation of on-and off-ball soccer players based on multi-agent deep reinforcement learning,

    H. Nakahara, K. Tsutsui, K. Takeda, and K. Fujii, “Action valuation of on-and off-ball soccer players based on multi-agent deep reinforcement learning,”IEEE Access, vol. 11, pp. 131 237–131 244, 2023

    2023

  9. [9]

    Action valuation in sports: A survey,

    A. Xarles, S. Escalera, T. B. Moeslund, and A. Clap ´es, “Action valuation in sports: A survey,” inProceedings of the Computer Vision and Pattern Recognition Conference, 2025, pp. 6132–6142

    2025

  10. [10]

    Score: A convolutional approach for football event forecast- ing,

    R. Alves, “Score: A convolutional approach for football event forecast- ing,”International Journal of Forecasting, 2025

    2025

  11. [11]

    Modeling matches as language: A generative transformer approach for counter- factual player valuation in football,

    M. Hong, M. Lee, G. Jo, H. Jo, P. Bauer, and S.-K. Ko, “Modeling matches as language: A generative transformer approach for counter- factual player valuation in football,”arXiv preprint arXiv:2603.15212, 2026

    work page arXiv 2026

  12. [12]

    Multiagent off-screen behavior prediction in football,

    S. Omidshafiei, D. Hennes, M. Garnelo, Z. Wang, A. Recasens, E. Tarassov, Y. Yang, R. Elie, J. T. Connor, P. Mulleret al., “Multiagent off-screen behavior prediction in football,”Scientific reports, vol. 12, no. 1, p. 8638, 2022

    2022

  13. [13]

    Agentformer: Agent-aware transformers for socio-temporal multi-agent forecasting,

    Y. Yuan, X. Weng, Y. Ou, and K. M. Kitani, “Agentformer: Agent-aware transformers for socio-temporal multi-agent forecasting,” inProceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 9813–9823

    2021

  14. [14]

    Sports-traj: A unified trajectory generation model for multi-agent movement in sports,

    Y. Xu and Y. Fu, “Sports-traj: A unified trajectory generation model for multi-agent movement in sports,” inThe Thirteenth International Conference on Learning Representations, 2025

    2025

  15. [15]

    Smgdiff: Soccer motion generation using diffusion probabilistic models,

    H. Yang, C. Li, Z. Wu, G. Li, J. Wang, J. Yu, Z. Su, and L. Xu, “Smgdiff: Soccer motion generation using diffusion probabilistic models,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2025, pp. 11 807–11 817

    2025

  16. [16]

    Tacticai: an ai assistant for football tactics,

    Z. Wang, P. Veli ˇckovi´c, D. Hennes, N. Toma ˇsev, L. Prince, M. Kaisers, Y. Bachrach, R. Elie, L. K. Wenliang, F. Piccininiet al., “Tacticai: an ai assistant for football tactics,”Nature communications, vol. 15, no. 1, p. 1906, 2024

    1906

  17. [17]

    Taceleven: generative tactic discovery for football open play,

    S. Zhao, H. Ma, Z. Pu, J. Huang, Y. Pan, S. Wang, and Z. Ming, “Taceleven: generative tactic discovery for football open play,”arXiv preprint arXiv:2511.13326, 2025

    work page arXiv 2025

  18. [18]

    GenTac: Generative Modeling and Forecasting of Soccer Tactics

    J. Rao, T. Gui, H. Wu, Y. Wang, and W. Xie, “Gentac: Generative model- ing and forecasting of soccer tactics,”arXiv preprint arXiv:2604.11786, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  19. [19]

    On the Opportunities and Risks of Foundation Models

    R. Bommasani, D. A. Hudson, E. Adeli, R. Altman, S. Arora, S. von Arx, M. S. Bernstein, J. Bohg, A. Bosselut, E. Brunskillet al., “On the opportunities and risks of foundation models,”arXiv preprint arXiv:2108.07258, 2021

    work page internal anchor Pith review arXiv 2021

  20. [20]

    Explaining neural scaling laws,

    Y. Bahri, E. Dyer, J. Kaplan, J. Lee, and U. Sharma, “Explaining neural scaling laws,”Proceedings of the National Academy of Sciences, vol. 121, no. 27, p. e2311878121, 2024

    2024

  21. [21]

    Scalable diffusion models with transformers,

    W. Peebles and S. Xie, “Scalable diffusion models with transformers,” inProceedings of the IEEE/CVF international conference on computer vision, 2023, pp. 4195–4205

    2023

  22. [22]

    A survey of deep learning in sports applica- tions: Perception, comprehension, and decision,

    Z. Zhao, W. Chai, S. Hao, W. Hu, G. Wang, S. Cao, M. Song, J.-N. Hwang, and G. Wang, “A survey of deep learning in sports applica- tions: Perception, comprehension, and decision,”IEEE Transactions on Visualization and Computer Graphics, vol. 31, no. 10, pp. 9368–9386, 2025

    2025

  23. [23]

    Long short-term memory,

    S. Hochreiter and J. Schmidhuber, “Long short-term memory,”Neural computation, vol. 9, no. 8, pp. 1735–1780, 1997

    1997

  24. [24]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin, “Attention is all you need,”Advances in neural information processing systems, vol. 30, 2017

    2017

  25. [25]

    Dag-net: Double attentive graph neural network for trajectory forecasting,

    A. Monti, A. Bertugli, S. Calderara, and R. Cucchiara, “Dag-net: Double attentive graph neural network for trajectory forecasting,” in2020 25th international conference on pattern recognition (ICPR), 2021, pp. 2551– 2558

    2021

  26. [26]

    Transportmer: A holistic approach to trajectory understanding in multi- agent sports,

    G. Capellera, L. Ferraz, A. Rubio, A. Agudo, and F. Moreno-Noguer, “Transportmer: A holistic approach to trajectory understanding in multi- agent sports,” inProceedings of the asian conference on computer vision, 2024, pp. 1652–1670

    2024

  27. [27]

    Football trajectory prediction and real-time feedback mechanism based on temporal convolutional network,

    C. Zhang, X. Xi, X. Wang, and Z. Zhang, “Football trajectory prediction and real-time feedback mechanism based on temporal convolutional network,”Alexandria Engineering Journal, vol. 114, pp. 476–484, 2025

    2025

  28. [28]

    Footbots: A transformer-based architecture for motion prediction in soccer,

    G. Capellera, L. Ferraz, A. Rubio, A. Agudo, and F. Moreno-Noguer, “Footbots: A transformer-based architecture for motion prediction in soccer,” inIEEE International Conference on Image Processing, 2024, pp. 2313–2319

    2024

  29. [29]

    Tutorial on

    C. Doersch, “Tutorial on variational autoencoders,”arXiv preprint arXiv:1606.05908, 2016

    work page arXiv 2016

  30. [30]

    Denoising diffusion probabilistic models,

    J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” Advances in Neural Information Processing Systems, vol. 33, pp. 6840– 6851, 2020

    2020

  31. [31]

    Playbest: Professional basketball player behavior synthesis via planning with diffusion,

    X. Chen, W.-Y. Wang, Z. Hu, D. Reynoso, K. Jin, M. Liu, P. J. Brantingham, and W. Wang, “Playbest: Professional basketball player behavior synthesis via planning with diffusion,” inProceedings of the 33rd ACM International Conference on Information and Knowledge Management, 2024, pp. 4406–4413

    2024

  32. [32]

    Unified uncertainty- aware diffusion for multi-agent trajectory modeling,

    G. Capellera, A. Rubio, L. Ferraz, and A. Agudo, “Unified uncertainty- aware diffusion for multi-agent trajectory modeling,” inProceedings of the IEEE/CVF international conference on computer vision, 2025

    2025

  33. [33]

    A survey on trajectory-prediction methods for autonomous driving,

    Y. Huang, J. Du, Z. Yang, Z. Zhou, L. Zhang, and H. Chen, “A survey on trajectory-prediction methods for autonomous driving,”IEEE transactions on intelligent vehicles, vol. 7, no. 3, pp. 652–674, 2022

    2022

  34. [34]

    Deep unsupervised learning using nonequilibrium thermodynamics,

    J. Sohl-Dickstein, E. Weiss, N. Maheswaranathan, and S. Ganguli, “Deep unsupervised learning using nonequilibrium thermodynamics,” in International Conference on Machine Learning, 2015, pp. 2256–2265

    2015

  35. [35]

    Diffusion models beat gans on image synthesis,

    P. Dhariwal and A. Nichol, “Diffusion models beat gans on image synthesis,”Advances in neural information processing systems, vol. 34, pp. 8780–8794, 2021

    2021

  36. [36]

    Diffusion-based planning for autonomous driving with flexible guidance,

    Y. Zheng, R. Liang, K. Zheng, J. Zheng, L. Mao, J. Li, W. Gu, R. Ai, S. E. Li, and Zhan, “Diffusion-based planning for autonomous driving with flexible guidance,” inInternational Conference on Learning Representations, 2025

    2025

  37. [37]

    Planning with diffusion for flexible behavior synthesis,

    M. Janner, Y. Du, J. B. Tenenbaum, and S. Levine, “Planning with diffusion for flexible behavior synthesis,” inInternational Conference on Machine Learning, 2022

    2022

  38. [38]

    Mlp-mixer: 14 An all-mlp architecture for vision,

    I. O. Tolstikhin, N. Houlsby, A. Kolesnikov, L. Beyer, X. Zhai, T. Un- terthiner, J. Yung, A. Steiner, D. Keysers, J. Uszkoreitet al., “Mlp-mixer: 14 An all-mlp architecture for vision,”Advances in neural information processing systems, vol. 34, pp. 24 261–24 272, 2021

    2021

  39. [39]

    Classifier-Free Diffusion Guidance

    J. Ho and T. Salimans, “Classifier-free diffusion guidance,”arXiv preprint arXiv:2207.12598, 2022

    work page internal anchor Pith review arXiv 2022

  40. [40]

    Contrastive energy prediction for exact energy-guided diffusion sampling in offline rein- forcement learning,

    C. Lu, H. Chen, J. Chen, H. Su, C. Li, and J. Zhu, “Contrastive energy prediction for exact energy-guided diffusion sampling in offline rein- forcement learning,” inInternational Conference on Machine Learning, 2023, pp. 22 825–22 855

    2023

  41. [41]

    Diffusion posterior sampling for general noisy inverse problems,

    H. Chung, J. Kim, M. T. Mccann, M. L. Klasky, and J. C. Ye, “Diffusion posterior sampling for general noisy inverse problems,” inInternational Conference on Learning Representations, 2023

    2023

  42. [42]

    Rethinking diffusion posterior sampling: From conditional score estimator to maximizing a posterior,

    T. Xu, X. Cai, X. Zhang, X. Ge, D. He, M. Sun, J. Liu, Y.-Q. Zhang, J. Li, and Y. Wang, “Rethinking diffusion posterior sampling: From conditional score estimator to maximizing a posterior,” inInternational Conference on Learning Representations, 2025

    2025

  43. [43]

    GPT-4 Technical Report

    J. Achiam, S. Adler, S. Agarwal, L. Ahmad, I. Akkaya, F. L. Aleman, D. Almeida, J. Altenschmidt, S. Altman, S. Anadkatet al., “Gpt-4 technical report,”arXiv preprint arXiv:2303.08774, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  44. [44]

    OpenAI o1 System Card

    A. Jaech, A. Kalai, A. Lerer, A. Richardson, A. El-Kishky, A. Low, A. Helyar, A. Madry, A. Beutel, A. Carneyet al., “Openai o1 system card,”arXiv preprint arXiv:2412.16720, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  45. [45]

    OpenAI GPT-5 System Card

    A. Singh, A. Fry, A. Perelman, A. Tart, A. Ganesh, A. El-Kishky, A. McLaughlin, A. Low, A. Ostrow, A. Ananthramet al., “Openai gpt-5 system card,”arXiv preprint arXiv:2601.03267, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  46. [46]

    R. S. Sutton, A. G. Bartoet al.,Reinforcement learning: An introduction. Cambridge, MA: MIT press Cambridge, 1998, vol. 1

    1998

  47. [47]

    Diffusion models for reinforcement learning: A survey.arXiv preprint arXiv:2311.01223, 2023

    Z. Zhu, H. Zhao, H. He, Y. Zhong, S. Zhang, H. Guo, T. Chen, and W. Zhang, “Diffusion models for reinforcement learning: A survey,” arXiv preprint arXiv:2311.01223, 2023

    work page arXiv 2023

  48. [48]

    Deep reinforcement learning: A brief survey,

    K. Arulkumaran, M. P. Deisenroth, M. Brundage, and A. A. Bharath, “Deep reinforcement learning: A brief survey,”IEEE signal processing magazine, vol. 34, no. 6, pp. 26–38, 2017

    2017

  49. [49]

    Synchronise event and tracking data using dynamic programming,

    M. Kwiatkowski, “Synchronise event and tracking data using dynamic programming,” https://github.com/huffyhenry/sync.soccer, 2020, gitHub repository

    2020

  50. [50]

    Multi-modal trajectory prediction of nba players,

    S. Hauri, N. Djuric, V. Radosavljevic, and S. Vucetic, “Multi-modal trajectory prediction of nba players,” inProceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2021, pp. 1640– 1649

    2021

  51. [51]

    Stochastic tra- jectory prediction via motion indeterminacy diffusion,

    T. Gu, G. Chen, J. Li, C. Lin, Y. Rao, J. Zhou, and J. Lu, “Stochastic tra- jectory prediction via motion indeterminacy diffusion,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 17 113–17 122

    2022

  52. [52]

    Leapfrog diffu- sion model for stochastic trajectory prediction,

    W. Mao, C. Xu, Q. Zhu, S. Chen, and Y. Wang, “Leapfrog diffu- sion model for stochastic trajectory prediction,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2023, pp. 5517–5526

    2023

  53. [53]

    Learning group interactions and se- mantic intentions for multi-object trajectory prediction,

    M. Qi, Y. Yang, and H. Ma, “Learning group interactions and se- mantic intentions for multi-object trajectory prediction,”arXiv preprint arXiv:2412.15673, 2024

    work page arXiv 2024

  54. [54]

    Multi-agent trajectory prediction with scalable diffusion transformer,

    S. Zhang, S. Kai, C. Chen, Y. Zhuang, Z. Zhu, M. Liu, and W. Zhang, “Multi-agent trajectory prediction with scalable diffusion transformer,” inProceedings of the International Conference on Distributed Artificial Intelligence, 2024

    2024

  55. [55]

    Joint metrics matter: A better standard for trajectory forecasting,

    E. Weng, H. Hoshino, D. Ramanan, and K. Kitani, “Joint metrics matter: A better standard for trajectory forecasting,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 20 315–20 326

    2023

  56. [56]

    Social lstm: Human trajectory prediction in crowded spaces,

    A. Alahi, K. Goel, V. Ramanathan, A. Robicquet, L. Fei-Fei, and S. Savarese, “Social lstm: Human trajectory prediction in crowded spaces,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 961–971

    2016

  57. [57]

    Stgat: Modeling spatial- temporal interactions for human trajectory prediction,

    Y. Huang, H. Bi, Z. Li, T. Mao, and Z. Wang, “Stgat: Modeling spatial- temporal interactions for human trajectory prediction,” inProceedings of the IEEE/CVF international conference on computer vision, 2019, pp. 6272–6281

    2019

  58. [58]

    Rating player actions in soccer,

    U. Dick, M. Tavakol, and U. Brefeld, “Rating player actions in soccer,” Frontiers in Sports and Active Living, vol. 3, p. 682986, 2021

    2021

  59. [59]

    Trajectron++: Dynamically-feasible trajectory forecasting with heterogeneous data,

    T. Salzmann, B. Ivanovic, P. Chakravarty, and M. Pavone, “Trajectron++: Dynamically-feasible trajectory forecasting with heterogeneous data,” in European Conference on Computer Vision. Springer, 2020, pp. 683– 700

    2020

  60. [60]

    Multimodal motion prediction with stacked transformers,

    Y. Liu, J. Zhang, L. Fang, Q. Jiang, and B. Zhou, “Multimodal motion prediction with stacked transformers,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2021, pp. 7577– 7586

    2021

  61. [61]

    Scene transformer: A unified architecture for predicting multiple agent trajectories,

    J. Ngiam, B. Caine, V. Vasudevan, Z. Zhang, H.-T. L. Chiang, J. Ling, R. Roelofs, A. Bewley, C. Liu, A. Venugopalet al., “Scene transformer: A unified architecture for predicting multiple agent trajectories,” inThe Thirteenth International Conference on Learning Representations, 2022

    2022

  62. [62]

    Diverse gener- ation for multi-agent sports games,

    R. A. Yeh, A. G. Schwing, J. Huang, and K. Murphy, “Diverse gener- ation for multi-agent sports games,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019, pp. 4610–4619

    2019

  63. [63]

    Madiff: Offline multi-agent learning with diffusion models,

    Z. Zhu, M. Liu, L. Mao, B. Kang, M. Xu, Y. Yu, S. Ermon, and W. Zhang, “Madiff: Offline multi-agent learning with diffusion models,” Advances in Neural Information Processing Systems, vol. 37, pp. 4177– 4206, 2024

    2024

  64. [64]

    Social gan: Socially acceptable trajectories with generative adversarial networks,

    A. Gupta, J. Johnson, L. Fei-Fei, S. Savarese, and A. Alahi, “Social gan: Socially acceptable trajectories with generative adversarial networks,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 2255–2264

    2018

  65. [65]

    Social-stgcnn: A social spatio-temporal graph convolutional neural network for human trajectory prediction,

    A. Mohamed, K. Qian, M. Elhoseiny, and C. Claudel, “Social-stgcnn: A social spatio-temporal graph convolutional neural network for human trajectory prediction,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 14 424–14 432

    2020

  66. [66]

    It is not the journey but the destination: Endpoint conditioned trajectory prediction,

    K. Mangalam, H. Girase, S. Agarwal, K.-H. Lee, E. Adeli, J. Malik, and A. Gaidon, “It is not the journey but the destination: Endpoint conditioned trajectory prediction,” inEuropean Conference on Computer Vision, 2020, pp. 759–776

    2020

  67. [67]

    Spatio-temporal graph transformer networks for pedestrian trajectory prediction,

    C. Yu, X. Ma, J. Ren, H. Zhao, and S. Yi, “Spatio-temporal graph transformer networks for pedestrian trajectory prediction,” inEuropean Conference on Computer Vision, 2020, pp. 507–523

    2020

  68. [68]

    Remember intentions: Retrospective-memory-based trajectory prediction,

    C. Xu, W. Mao, W. Zhang, and S. Chen, “Remember intentions: Retrospective-memory-based trajectory prediction,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 6488–6497

    2022

  69. [69]

    Non-probability sampling net- work for stochastic human trajectory prediction,

    I. Bae, J.-H. Park, and H.-G. Jeon, “Non-probability sampling net- work for stochastic human trajectory prediction,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 6477–6487

    2022

  70. [70]

    Groupnet: Multiscale hypergraph neural networks for trajectory prediction with relational reasoning,

    C. Xu, M. Li, Z. Ni, Y. Zhang, and S. Chen, “Groupnet: Multiscale hypergraph neural networks for trajectory prediction with relational reasoning,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 6498–6507

    2022

  71. [71]

    Physics-based modeling of pass probabilities in soccer,

    W. Spearman, A. Basye, G. Dick, R. Hotovy, and P. Pop, “Physics-based modeling of pass probabilities in soccer,” inProceeding of the 11th MIT Sloan Sports Analytics Conference, vol. 1, 2017

    2017

  72. [72]

    Beyond expected goals,

    W. Spearman, “Beyond expected goals,” inProceedings of the 12th MIT sloan sports analytics conference, 2018, pp. 1–17

    2018

  73. [73]

    Wide open spaces: A statistical technique for measuring space creation in professional soccer,

    J. Fernandez and L. Bornn, “Wide open spaces: A statistical technique for measuring space creation in professional soccer,” inSloan sports analytics conference, 2018

    2018

  74. [74]

    Space and control in soccer,

    F. Martens, U. Dick, and U. Brefeld, “Space and control in soccer,” Frontiers in Sports and Active Living, vol. 3, p. 676179, 2021

    2021

  75. [75]

    Measuring the pitch control of professional football players using spatiotemporal tracking data,

    L. Higgins, T. Galla, B. Prestidge, and T. Wyatt, “Measuring the pitch control of professional football players using spatiotemporal tracking data,”Journal of Physics: Complexity, vol. 4, no. 2, 2023

    2023

  76. [76]

    Decoupled Weight Decay Regularization

    I. Loshchilov and F. Hutter, “Decoupled weight decay regularization,” arXiv preprint arXiv:1711.05101, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  77. [77]

    Gatraj: A graph-and attention-based multi-agent trajectory prediction model,

    H. Cheng, M. Liu, L. Chen, H. Broszio, M. Sester, and M. Y. Yang, “Gatraj: A graph-and attention-based multi-agent trajectory prediction model,”ISPRS Journal of Photogrammetry and Remote Sensing, vol. 205, pp. 163–175, 2023

    2023

  78. [78]

    M2i: From factored marginal trajectory prediction to interactive prediction,

    Q. Sun, X. Huang, J. Gu, B. C. Williams, and H. Zhao, “M2i: From factored marginal trajectory prediction to interactive prediction,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 6543–6552

    2022

  79. [79]

    Improved denoising diffusion probabilis- tic models,

    A. Q. Nichol and P. Dhariwal, “Improved denoising diffusion probabilis- tic models,” inInternational Conference on Machine Learning (ICML). PMLR, 2021, pp. 8162–8171

    2021

  80. [80]

    Denoising Diffusion Implicit Models

    J. Song, C. Meng, and S. Ermon, “Denoising diffusion implicit models,” arXiv preprint arXiv:2010.02502, 2020

    work page internal anchor Pith review Pith/arXiv arXiv 2010

Showing first 80 references.