pith. sign in

arxiv: 2607.00811 · v1 · pith:POYIYXO6new · submitted 2026-07-01 · 💻 cs.LG

From Pixels to Temporal Correlations: Learning Informative Representations for Reinforcement Learning Pre-training

Jinwen Wang , Youfang Lin , Xiaobo Hu , Siyu Yang , Sheng Han , Shuo Wang , Kai Lv This is my paper

Pith reviewed 2026-07-02 15:56 UTC · model grok-4.3

classification 💻 cs.LG
keywords reinforcement learningpre-trainingcontrastive learningtemporal correlationsvideo representationssample efficiencymulti-scale learningpolicy transfer
0
0 comments X

The pith

Modeling multi-scale temporal correlations separately in a dedicated space produces balanced video representations that support better reinforcement learning policies.

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

Existing pre-training approaches on action-free videos for RL rely on single-step transition prediction or image reconstruction, which the paper states tend to preserve large stationary pixel information while neglecting smaller but crucial details. The authors introduce a temporal correlation space that distinguishes each video element and implement Multi-scale Temporal Contrastive Learning to handle correlations at separate scales. This is claimed to equalize attention across elements and yield representations that improve sample efficiency and asymptotic performance when transferred to downstream RL tasks. A reader would care because pre-training on abundant unlabeled video could reduce the need for costly environment interactions during policy learning.

Core claim

The paper claims that introducing a temporal correlation space to distinguish each element in videos, together with the Multi-scale Temporal Contrastive Learning method that models multi-scale temporal correlations separately, balances attention across elements and preserves small-scale information that single-step or reconstruction methods discard due to stationary bias, thereby producing more informative representations that support policy learning in various downstream tasks.

What carries the argument

The Multi-scale Temporal Contrastive Learning (MTCL) method, which models multi-scale temporal correlations separately inside a temporal correlation space to equalize attention to video elements.

If this is right

  • Representations from MTCL improve sample efficiency across various downstream RL tasks.
  • Representations from MTCL improve asymptotic performance across various downstream RL tasks.
  • The method supports effective policy learning in multiple different downstream tasks.
  • MTCL avoids the preference for large-proportion stationary information that affects single-step and reconstruction pre-training.

Where Pith is reading between the lines

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

  • The same separation of temporal scales could be tested on non-RL video tasks such as action recognition or future-frame prediction to check generality.
  • If the correlation space truly balances elements, ablating individual scales should produce measurable drops in downstream RL metrics that single-scale versions cannot recover.
  • The approach implies that contrastive objectives grounded in explicit temporal distinctions may transfer more reliably than reconstruction objectives when video data contains both static backgrounds and sparse motion.

Load-bearing premise

That separating multi-scale temporal correlations in a dedicated space will avoid the stationary bias of prior methods while still preserving all necessary information for downstream policy learning.

What would settle it

Run MTCL pre-training on the same video dataset as a single-step baseline, then fine-tune both on an identical RL task and measure whether MTCL shows no gain or a loss in sample efficiency or final performance.

Figures

Figures reproduced from arXiv: 2607.00811 by Jinwen Wang, Kai Lv, Sheng Han, Shuo Wang, Siyu Yang, Xiaobo Hu, Youfang Lin.

Figure 1
Figure 1. Figure 1: Illustration of our method. We convert videos from [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of our model. Building on an action-free world model framework, we independently model multi-scale [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Learning curves on DMControl Remastered. We present the learning curves of our method (MTCL) compared to [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Meta-World (left) and CARLA (right) results. (a) Learning curves of MTCL (ours) compared to baselines across six tasks in Meta-World, based on the average success rate over five runs. (b) Learning curves of MTCL (ours) compared to baselines in the “ClearNoon” and “WetSunset” weather conditions on CARLA, measured by average episode return over five runs. 5.1 Evaluation on DMControl Remastered DMC Remastered… view at source ↗
Figure 5
Figure 5. Figure 5: Ablation Study. We conduct ablation studies on the MML and SAL components, selecting one task each from DMCR, [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Variant Experiments. (a) We compare the effectiveness of multi-scale temporal correlation modeling with single-scale [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Visualization of Reconstruction. We select two video examples and visualize the reconstructed images [PITH_FULL_IMAGE:figures/full_fig_p008_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Video Prediction. We present the predicted future [PITH_FULL_IMAGE:figures/full_fig_p008_8.png] view at source ↗
read the original abstract

Unsupervised pre-training on large-scale datasets has demonstrated significant potential for improving the sample efficiency and performance of Reinforcement Learning (RL). Given the large-scale action-free internet videos, existing methods utilize single-step transition prediction and image reconstruction to learn representations. However, these methods prefer to preserve large-proportion stationary information in the pixel space, neglecting small but crucial information. To preserve enough information in the representation, it is essential to pay equal attention to each element in videos. Specifically, we propose a temporal correlation space to distinguish each element. For implementation, we introduce the Multi-scale Temporal Contrastive Learning (MTCL) method to model multi-scale temporal correlations separately. This approach can balance the attention of different elements and yield more informative representations, effectively supporting policy learning in various downstream tasks. Experimental results demonstrate that our method improves sample efficiency and asymptotic performance across various downstream tasks.

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

1 major / 0 minor

Summary. The paper proposes Multi-scale Temporal Contrastive Learning (MTCL) operating in a temporal correlation space to learn representations from action-free internet videos for RL pre-training. It argues that single-step transition prediction and image reconstruction methods over-emphasize stationary pixel information and neglect small but crucial details; MTCL models multi-scale temporal correlations separately to balance attention across elements and produce more informative representations that improve sample efficiency and asymptotic performance on downstream RL tasks.

Significance. If the empirical claims hold with proper controls, the work would provide a concrete alternative to reconstruction- and single-step-based pre-training for RL by explicitly targeting multi-scale temporal structure, potentially yielding representations that better support policy learning across tasks. No machine-checked proofs, reproducible code releases, or parameter-free derivations are described.

major comments (1)
  1. [Abstract] Abstract: the central empirical claim (improved sample efficiency and asymptotic performance across downstream tasks) is asserted without any methods details, baselines, error bars, dataset descriptions, or result tables, so the support for the claim cannot be assessed from the provided text.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their feedback. We address the single major comment below.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central empirical claim (improved sample efficiency and asymptotic performance across downstream tasks) is asserted without any methods details, baselines, error bars, dataset descriptions, or result tables, so the support for the claim cannot be assessed from the provided text.

    Authors: We agree that the abstract is necessarily concise and therefore omits methodological specifics, baseline names, quantitative results with error bars, and dataset details. The full manuscript supplies these in the Methods (MTCL formulation and multi-scale contrastive objectives) and Experiments sections (baselines, datasets of action-free internet videos, multiple random seeds with error bars, and tables showing gains in sample efficiency and asymptotic performance). We will revise the abstract to incorporate a brief description of the MTCL approach and a high-level summary of the empirical outcomes while remaining within length constraints. revision: yes

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper's central contribution is an empirical unsupervised pre-training method (MTCL) that constructs a temporal correlation space and models multi-scale correlations to learn representations for RL. No equations, derivations, or first-principles claims are present that reduce any result to a fitted parameter or self-referential definition by construction. The abstract and described approach motivate the method via avoidance of stationary bias but do not invoke self-citations as load-bearing uniqueness theorems, smuggle ansatzes, or rename known results as new derivations. Claims of improved sample efficiency rest on experimental results rather than tautological reductions, making the derivation chain self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

The central claim rests on the domain assumption that multi-scale temporal correlations capture information missed by prior methods; no free parameters or invented physical entities are mentioned in the abstract.

axioms (2)
  • domain assumption Existing single-step transition prediction and image reconstruction methods preferentially preserve large-proportion stationary information while neglecting small but crucial information.
    Stated directly in the abstract as the motivation for introducing a temporal correlation space.
  • domain assumption Paying equal attention to each element in videos via multi-scale temporal correlations will produce more informative representations for downstream RL.
    Core premise of the MTCL proposal in the abstract.
invented entities (1)
  • Multi-scale Temporal Contrastive Learning (MTCL) no independent evidence
    purpose: To model multi-scale temporal correlations separately and balance attention across video elements.
    Introduced as the proposed implementation; no independent evidence outside the paper is referenced in the abstract.

pith-pipeline@v0.9.1-grok · 5691 in / 1429 out tokens · 33831 ms · 2026-07-02T15:56:23.143843+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

45 extracted references · 17 canonical work pages · 3 internal anchors

  1. [1]

    Variational Inference: A Review for Statisticians

    David M. Blei, Alp Kucukelbir, and Jon D. McAuliffe. 2016. Variational Inference: A Review for Statisticians.CoRRabs/1601.00670 (2016). arXiv:1601.00670 http: //arxiv.org/abs/1601.00670

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  2. [2]

    Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winter, Christopher Hesse, Mark Chen, Eric Sigler, Mateusz Litwin...

    2020

  3. [3]

    Chang Chen, Yi-Fu Wu, Jaesik Yoon, and Sungjin Ahn. 2022. TransDreamer: Reinforcement Learning with Transformer World Models.CoRRabs/2202.09481 (2022). arXiv:2202.09481 https://arxiv.org/abs/2202.09481

    work page arXiv 2022

  4. [4]

    Swaroop Guntupalli, Miguel Lázaro-Gredilla, and Kevin Patrick Mur- phy

    Antoine Dedieu, Joseph Ortiz, Xinghua Lou, Carter Wendelken, Wolfgang Lehrach, J. Swaroop Guntupalli, Miguel Lázaro-Gredilla, and Kevin Patrick Mur- phy. 2025. Improving Transformer World Models for Data-Efficient RL.CoRR abs/2502.01591 (2025). arXiv:2502.01591 doi:10.48550/ARXIV.2502.01591

    work page doi:10.48550/arxiv.2502.01591 2025

  5. [5]

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. InProceedings of the 2019 Conference of the North American Chapter of the Associ- ation for Computational Linguistics: Human Language Technologies, NAACL-HLT 2019, Minneapolis, MN, USA, June 2-7, 2019, V...

    work page doi:10.18653/v1/n19-1423 2019

  6. [6]

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xi- aohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, and Neil Houlsby. 2021. An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale. In9th Interna- tional Conference on Learning Representations, IC...

    2021

  7. [7]

    López, and Vladlen Koltun

    Alexey Dosovitskiy, Germán Ros, Felipe Codevilla, Antonio M. López, and Vladlen Koltun. 2017. CARLA: An Open Urban Driving Simulator. In1st An- nual Conference on Robot Learning, CoRL 2017, Mountain View, California, USA, November 13-15, 2017, Proceedings (Proceedings of Machine Learning Research, Vol. 78). PMLR, 1–16. http://proceedings.mlr.press/v78/dos...

    2017

  8. [8]

    Dibya Ghosh, Chethan Anand Bhateja, and Sergey Levine. 2023. Reinforcement Learning from Passive Data via Latent Intentions. InInternational Conference on Machine Learning, ICML 2023, 23-29 July 2023, Honolulu, Hawaii, USA (Proceed- ings of Machine Learning Research, Vol. 202), Andreas Krause, Emma Brunskill, Kyunghyun Cho, Barbara Engelhardt, Sivan Sabat...

    2023

  9. [9]

    Something Something

    Raghav Goyal, Samira Ebrahimi Kahou, Vincent Michalski, Joanna Materzynska, Susanne Westphal, Heuna Kim, Valentin Haenel, Ingo Fründ, Peter Yianilos, Moritz Mueller-Freitag, Florian Hoppe, Christian Thurau, Ingo Bax, and Roland Memisevic. 2017. The "Something Something" Video Database for Learning and Evaluating Visual Common Sense. InIEEE International C...

    work page doi:10.1109/iccv.2017.622 2017

  10. [10]

    Jake Grigsby and Yanjun Qi. 2020. Measuring Visual Generalization in Con- tinuous Control from Pixels.CoRRabs/2010.06740 (2020). arXiv:2010.06740 https://arxiv.org/abs/2010.06740

    work page arXiv 2020

  11. [11]

    Lillicrap, Jimmy Ba, and Mohammad Norouzi

    Danijar Hafner, Timothy P. Lillicrap, Jimmy Ba, and Mohammad Norouzi. 2020. Dream to Control: Learning Behaviors by Latent Imagination. In8th International Conference on Learning Representations, ICLR 2020, Addis Ababa, Ethiopia, April 26-30, 2020. OpenReview.net. https://openreview.net/forum?id=S1lOTC4tDS

    2020

  12. [12]

    Lillicrap, Ian Fischer, Ruben Villegas, David Ha, Honglak Lee, and James Davidson

    Danijar Hafner, Timothy P. Lillicrap, Ian Fischer, Ruben Villegas, David Ha, Honglak Lee, and James Davidson. 2019. Learning Latent Dynamics for Plan- ning from Pixels. InProceedings of the 36th International Conference on Machine Learning, ICML 2019, 9-15 June 2019, Long Beach, California, USA (Proceedings of Machine Learning Research, Vol. 97), Kamalika...

    2019

  13. [13]

    Lillicrap, Mohammad Norouzi, and Jimmy Ba

    Danijar Hafner, Timothy P. Lillicrap, Mohammad Norouzi, and Jimmy Ba. 2021. Mastering Atari with Discrete World Models. In9th International Conference on Learning Representations, ICLR 2021, Virtual Event, Austria, May 3-7, 2021. OpenReview.net. https://openreview.net/forum?id=0oabwyZbOu

    2021

  14. [14]

    Mastering Diverse Domains through World Models

    Danijar Hafner, Jurgis Pasukonis, Jimmy Ba, and Timothy P. Lillicrap. 2023. Mastering Diverse Domains through World Models.CoRRabs/2301.04104 (2023)

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  15. [15]

    Efros, Lerrel Pinto, and Xiaolong Wang

    Nicklas Hansen, Rishabh Jangir, Yu Sun, Guillem Alenyà, Pieter Abbeel, Alexei A. Efros, Lerrel Pinto, and Xiaolong Wang. 2021. Self-Supervised Policy Adaptation during Deployment. In9th International Conference on Learning Representations, ICLR 2021, Virtual Event, Austria, May 3-7, 2021. OpenReview.net. https:// openreview.net/forum?id=o_V-MjyyGV_

    2021

  16. [16]

    Nicklas Hansen, Hao Su, and Xiaolong Wang. 2021. Stabilizing Deep Q-Learning with ConvNets and Vision Transformers under Data Augmentation. InAdvances in Neural Information Processing Systems 34: Annual Conference on Neural In- formation Processing Systems 2021, NeurIPS 2021, December 6-14, 2021, virtual, Marc’Aurelio Ranzato, Alina Beygelzimer, Yann N. D...

    2021

  17. [17]

    Nicklas Hansen, Hao Su, and Xiaolong Wang. 2022. Temporal Difference Learning for Model Predictive Control. InInternational Conference on Machine Learning, ICML 2022, 17-23 July 2022, Baltimore, Maryland, USA (Proceedings of Machine Learning Research, Vol. 162), Kamalika Chaudhuri, Stefanie Jegelka, Le Song, Csaba Szepesvári, Gang Niu, and Sivan Sabato (E...

    2022

  18. [18]

    Thrush, R

    Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, and Ross B. Girshick. 2022. Masked Autoencoders Are Scalable Vision Learners. InIEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2022, New Orleans, LA, USA, June 18-24, 2022. IEEE, 15979–15988. doi:10.1109/CVPR52688.2022.01553

    work page doi:10.1109/cvpr52688.2022.01553 2022

  19. [19]

    Girshick

    Kaiming He, Haoqi Fan, Yuxin Wu, Saining Xie, and Ross B. Girshick. 2020. Momentum Contrast for Unsupervised Visual Representation Learning. In2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2020, Seattle, W A, USA, June 13-19, 2020. Computer Vision Foundation / IEEE, 9726–

    2020

  20. [20]

    doi:10.1109/CVPR42600.2020.00975

    work page doi:10.1109/cvpr42600.2020.00975 2020

  21. [21]

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. 2016. Deep Residual Learning for Image Recognition. In2016 IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2016, Las Vegas, NV, USA, June 27-30, 2016. IEEE Computer Society, 770–778. doi:10.1109/CVPR.2016.90

    work page doi:10.1109/cvpr.2016.90 2016

  22. [22]

    Wenlong Huang, Chen Wang, Ruohan Zhang, Yunzhu Li, Jiajun Wu, and Li Fei- Fei. 2023. VoxPoser: Composable 3D Value Maps for Robotic Manipulation with Language Models. InConference on Robot Learning, CoRL 2023, 6-9 November 2023, Atlanta, GA, USA (Proceedings of Machine Learning Research, Vol. 229), Jie Tan, Marc Toussaint, and Kourosh Darvish (Eds.). PMLR...

    2023

  23. [23]

    Stephen James, Zicong Ma, David Rovick Arrojo, and Andrew J. Davison. 2020. RLBench: The Robot Learning Benchmark & Learning Environment.IEEE Ro- botics Autom. Lett.5, 2 (2020), 3019–3026. doi:10.1109/LRA.2020.2974707

    work page doi:10.1109/lra.2020.2974707 2020

  24. [24]

    Sergey Levine, Chelsea Finn, Trevor Darrell, and Pieter Abbeel. 2016. End-to-End Training of Deep Visuomotor Policies.J. Mach. Learn. Res.17 (2016), 39:1–39:40. https://jmlr.org/papers/v17/15-522.html

    2016

  25. [25]

    Dipendra Misra, Akanksha Saran, Tengyang Xie, Alex Lamb, and John Langford

  26. [26]

    arXiv:2403.13765 doi:10.48550/ ARXIV.2403.13765

    Towards Principled Representation Learning from Videos for Reinforce- ment Learning.CoRRabs/2403.13765 (2024). arXiv:2403.13765 doi:10.48550/ ARXIV.2403.13765

    work page arXiv 2024

  27. [27]

    Suraj Nair, Aravind Rajeswaran, Vikash Kumar, Chelsea Finn, and Abhinav Gupta. 2022. R3M: A Universal Visual Representation for Robot Manipulation. InConference on Robot Learning, CoRL 2022, 14-18 December 2022, Auckland, New Zealand (Proceedings of Machine Learning Research, Vol. 205), Karen Liu, Dana Kulic, and Jeffrey Ichnowski (Eds.). PMLR, 892–909. h...

    2022

  28. [28]

    Masashi Okada and Tadahiro Taniguchi. 2021. Dreaming: Model-based Rein- forcement Learning by Latent Imagination without Reconstruction. InIEEE International Conference on Robotics and Automation, ICRA 2021, Xi’an, China, May 30 - June 5, 2021. IEEE, 4209–4215. doi:10.1109/ICRA48506.2021.9560734

    work page doi:10.1109/icra48506.2021.9560734 2021

  29. [29]

    Ilija Radosavovic, Tete Xiao, Stephen James, Pieter Abbeel, Jitendra Malik, and Trevor Darrell. 2022. Real-World Robot Learning with Masked Visual Pre-training. InConference on Robot Learning, CoRL 2022, 14-18 December 2022, Auckland, New Zealand (Proceedings of Machine Learning Research, Vol. 205), Karen Liu, Dana Kulic, and Jeffrey Ichnowski (Eds.). PML...

    2022

  30. [30]

    James, and Pieter Abbeel

    Younggyo Seo, Kimin Lee, Stephen L. James, and Pieter Abbeel. 2022. Rein- forcement Learning with Action-Free Pre-Training from Videos. InInternational Conference on Machine Learning, ICML 2022, 17-23 July 2022, Baltimore, Maryland, USA (Proceedings of Machine Learning Research, Vol. 162), Kamalika Chaudhuri, Stefanie Jegelka, Le Song, Csaba Szepesvári, G...

    2022

  31. [31]

    DeepMind Control Suite

    Yuval Tassa, Yotam Doron, Alistair Muldal, Tom Erez, Yazhe Li, Diego de Las Casas, David Budden, Abbas Abdolmaleki, Josh Merel, Andrew Lefrancq, Timothy P. Lillicrap, and Martin A. Riedmiller. 2018. DeepMind Control Suite. CoRRabs/1801.00690 (2018). arXiv:1801.00690 http://arxiv.org/abs/1801.00690

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  32. [32]

    Gomez, Lukasz Kaiser, and Illia Polosukhin

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, and Illia Polosukhin. 2017. Attention is All you Need. InAdvances in Neural Information Processing Systems 30: An- nual Conference on Neural Information Processing Systems 2017, December 4- 9, 2017, Long Beach, CA, USA, Isabelle Guyon, Ulrike von Luxbur...

    2017

  33. [33]

    Shuo Wang, Zhihao Wu, Xiaobo Hu, Youfang Lin, and Kai Lv. 2023. Skill-Based Hierarchical Reinforcement Learning for Target Visual Navigation.IEEE Trans. Multim.25 (2023), 8920–8932. doi:10.1109/TMM.2023.3243618 Jinwen Wang et al

    work page doi:10.1109/tmm.2023.3243618 2023

  34. [34]

    Shuo Wang, Zhihao Wu, Xiaobo Hu, Jinwen Wang, Youfang Lin, and Kai Lv

  35. [35]

    What Effects the Generalization in Visual Reinforcement Learning: Policy Consistency with Truncated Return Prediction. InThirty-Eighth AAAI Conference on Artificial Intelligence, AAAI 2024, Thirty-Sixth Conference on Innovative Appli- cations of Artificial Intelligence, IAAI 2024, Fourteenth Symposium on Educational Advances in Artificial Intelligence, EA...

    work page doi:10.1609/aaai.v38i6.28369 2024

  36. [36]

    Jialong Wu, Haoyu Ma, Chaoyi Deng, and Mingsheng Long. 2023. Pre- training Contextualized World Models with In-the-wild Videos for Re- inforcement Learning. InAdvances in Neural Information Processing Sys- tems 36: Annual Conference on Neural Information Processing Systems 2023, NeurIPS 2023, New Orleans, LA, USA, December 10 - 16, 2023, Alice Oh, Tristan...

    2023

  37. [37]

    Carbonell, Ruslan Salakhutdi- nov, and Quoc V

    Zhilin Yang, Zihang Dai, Yiming Yang, Jaime G. Carbonell, Ruslan Salakhutdi- nov, and Quoc V. Le. 2019. XLNet: Generalized Autoregressive Pretraining for Language Understanding. InAdvances in Neural Information Processing Systems 32: Annual Conference on Neural Information Processing Systems 2019, NeurIPS 2019, December 8-14, 2019, Vancouver, BC, Canada, ...

    2019

  38. [38]

    Weirui Ye, Shaohuai Liu, Thanard Kurutach, Pieter Abbeel, and Yang Gao. 2021. Mastering Atari Games with Limited Data. InAdvances in Neural Informa- tion Processing Systems 34: Annual Conference on Neural Information Process- ing Systems 2021, NeurIPS 2021, December 6-14, 2021, virtual, Marc’Aurelio Ran- zato, Alina Beygelzimer, Yann N. Dauphin, Percy Lia...

    2021

  39. [39]

    Weirui Ye, Yunsheng Zhang, Pieter Abbeel, and Yang Gao. 2023. Become a Proficient Player with Limited Data through Watching Pure Videos. InThe Eleventh International Conference on Learning Representations, ICLR 2023, Kigali, Rwanda, May 1-5, 2023. OpenReview.net. https://openreview.net/forum?id=Sy- o2N0hF4f

    2023

  40. [40]

    Tianhe Yu, Deirdre Quillen, Zhanpeng He, Ryan Julian, Karol Hausman, Chelsea Finn, and Sergey Levine. 2019. Meta-World: A Benchmark and Evaluation for Multi-Task and Meta Reinforcement Learning. In3rd Annual Conference on Robot Learning, CoRL 2019, Osaka, Japan, October 30 - November 1, 2019, Proceedings (Proceedings of Machine Learning Research, Vol. 100...

    2019

  41. [41]

    Zhecheng Yuan, Zhengrong Xue, Bo Yuan, Xueqian Wang, Yi Wu, Yang Gao, and Huazhe Xu. 2022. Pre-Trained Image Encoder for Generalizable Visual Reinforcement Learning. InAdvances in Neural Information Process- ing Systems 35: Annual Conference on Neural Information Processing Sys- tems 2022, NeurIPS 2022, New Orleans, LA, USA, November 28 - Decem- ber 9, 20...

    2022

  42. [42]

    Lixuan Zhang, Meina Kan, Shiguang Shan, and Xilin Chen. 2024. PreLAR: World Model Pre-training with Learnable Action Representation. InComputer Vision - ECCV 2024 - 18th European Conference, Milan, Italy, September 29-October 4, 2024, Proceedings, Part XXIII (Lecture Notes in Computer Science, Vol. 15081). Springer, 185–201

    2024

  43. [43]

    Bolei Zhou, Àgata Lapedriza, Aditya Khosla, Aude Oliva, and Antonio Torralba

  44. [44]

    Pattern Anal

    Places: A 10 Million Image Database for Scene Recognition.IEEE Trans. Pattern Anal. Mach. Intell.40, 6 (2018), 1452–1464. doi:10.1109/TPAMI.2017. 2723009

    work page doi:10.1109/tpami.2017 2018

  45. [45]

    Bohan Zhou, Ke Li, Jiechuan Jiang, and Zongqing Lu. 2023. Learning from Visual Observation via Offline Pretrained State-to-Go Transformer. InAdvances in Neu- ral Information Processing Systems 36: Annual Conference on Neural Information Processing Systems 2023, NeurIPS 2023, New Orleans, LA, USA, December 10 - 16, 2023, Alice Oh, Tristan Naumann, Amir Glo...

    2023