Pith. sign in

REVIEW 3 major objections 6 minor 69 references

ProWorld: Progress-Aware Hyperbolic World Models for Long-Horizon Visual Goal Reaching

T0 review · 3 major / 6 minor · reviewed 2026-08-04 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read ProWorld claims that organizing visual latent dynamics by goal-conditioned progress order in hyperbolic space lets a JEPA-style world model reach distant visual goals substantially more reliably than local next-step prediction.

desk verdict A solid method paper whose headline 9.67-point SR gain is probably inflated by a transient-success metric that the planner's own best-intermediate cost directly exploits; worth refereeing. read the letter →

arxiv 2608.01926 v1 pith:5IPYJEIX submitted 2026-08-03 cs.AI

Zihan Liu , Yuzhe Zhuang , Yuanzu Li , Wanshuang Gou , Jiahong Liu , Min Zhou , Menglin Yang This is my paper
classification cs.AI
keywords worldmodelsgoal-conditionedplanninghyperbolicgeometryLorentzmodellatentdynamicslong-horizonentailmentconeprogressorder
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

ProWorld sets out to show that a visual world model can plan long-horizon goal-reaching far better if its latent space is organized by how much each state advances the goal, not just by local next-step accuracy. The paper argues that JEPA-style models—which learn to predict the very next latent state—suffer from two long-horizon failures: multi-step rollouts drift, and locally similar future states can correspond to very different goal progress. To fix this, it introduces goal-conditioned progress order, an asymmetric, coarse-to-fine relation, and embeds states in hyperbolic (Lorentz) space so that early states occupy broad regions and later states concentrate near the goal. A progress-aware planning objective scores candidate action sequences by terminal, best, and mean goal distance. Across four visual goal-reaching tasks, the model reports an average absolute success-rate gain of 9.67 percentage points over a strong local-prediction baseline, with the biggest gains on the longest-horizon task.

What carries the argument

Goal-conditioned progress order, realized through a goal-anchored adaptive entailment cone in the Lorentz model of hyperbolic space. The relation s_i ≺_g s_j says that state j occurs later than i in a goal-conditioned trajectory and has progress potential at least ε higher toward goal g; the paper uses temporally separated pairs as weak supervision. The cone constraint pushes each child state into an adaptive cone pointing from parent to goal, with aperture widening as the parent-goal distance grows, and adds a margin requiring the child to be closer to the goal than the parent. A Lorentz contrastive loss and a planning cost using terminal, best, and mean goal distances complete the mechanis

What would settle it

Run ProWorld on a task whose successful routes require temporarily moving away from the goal before approaching it (e.g., a maze with a detour). If temporal-order supervision mislabels progress, the learned entailment cones should point in the wrong direction on the detour segment, success rate should fall toward the no-cone ablation, and cone-loss violations should concentrate on that segment.

Watch

Extended reading notes

Core claim

The paper claims that long-horizon visual goal planning fails for JEPA-style world models trained only on local next-step consistency, because multi-step rollouts drift and locally similar futures can have different goal progress. It introduces goal-conditioned progress order—a relative ordering that is asymmetric and coarse-to-fine—and shows it can be realized geometrically by embedding states in the Lorentz model of hyperbolic space. Training then imposes three objectives: hyperbolic future prediction, contrastive discrimination of future states, and a goal-anchored entailment cone that keeps later states inside the parent-to-goal direction and closer to the goal. At inference, a progress-

Load-bearing premise

The load-bearing premise is that temporal order within a goal-conditioned hindsight trajectory is a reliable weak label for goal progress—an assumption that fails for trajectories with detours, backtracking, or subgoal switches, as the paper's limitations section concedes.

Editorial extensions

If this is right

  • On long-horizon visual goal-reaching, models trained only on local transition consistency will drift; the largest reported gains occur on the longest-horizon task, AntMaze-L.
  • Treating temporal order in hindsight trajectories as weak progress supervision is sufficient to learn useful progress structure without success labels.
  • The progress-aware planning cost—terminal, best-intermediate, and mean goal distance—is necessary; terminal-only planning drops success from 78% to 62% in the Cube-S ablation.
  • The entailment cone relies on correct temporal order: reversing or randomizing parent-child pairs drops success from 78% to 44% and 52%, respectively.
  • Hyperbolic geometry contributes beyond Euclidean latent spaces: replacing the Lorentz model with Euclidean space drops success from 78% to 68%.
  • A testable extension: use the same progress-order supervision in model-based RL or video prediction, where 'progress' need not be goal-reaching but can be any trajectory-level completion signal.
  • The ablation order suggests future-state discriminability (contrastive loss) may matter more than hyperbolic geometry per se; a Euclidean contrastive-only variant would test whether the gain is geometric or discriminative.
  • Because the temporal offset Δ=2 is optimal and Δ=1 is worse, the method assumes a minimum separation to read progress; domains with dense, noisy transitions may need per-domain Δ tuning or learned progress estimation.
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

3 major / 6 minor

Summary. ProWorld proposes a JEPA-style visual world model that embeds latents in Lorentz hyperbolic space and organizes them by a goal-conditioned progress order. Training combines hyperbolic prediction, contrastive future discrimination, and a goal-anchored adaptive entailment cone; planning uses a CEM over rollouts scored by terminal, best-intermediate, and mean hyperbolic goal distances. The paper reports an average absolute success-rate gain of 9.67 over LeWM on PushT, Cube-S, AntMaze-L, and Scene, supported by Cube-S ablations, sensitivity analyses, and geometric diagnostics.

Significance. If the reported gains survive a standard success criterion, ProWorld would be a solid empirical contribution: it adds goal-progress structure to latent world models with negligible overhead (+0.83% parameters), and the ablations are internally consistent. The explicit formalization of progress order, the adaptive cone design, and the decomposed planning cost are useful ideas. However, the headline claim is currently confounded by the transient-success evaluation criterion and by the absence of significance testing, so the central claim is not yet conclusively established.

major comments (3)
  1. [Appendix, Eq. (2); Eq. (23); Table 6; Fig. 13] The success metric counts an episode as successful if the environment termination flag fires at any step within the evaluation budget, regardless of the final state. The planning cost Eq. (23) includes C_best, the minimum goal distance along the rollout, and Table 6 shows that removing C_best lowers Cube-S SR from 78% to 64%. Fig. 13 also acknowledges post-success drift. A terminal-distance-only planner such as LeWM is systematically disadvantaged by this transient-success criterion, because it is never credited for passing near the goal. Please re-evaluate with a final-state success criterion (or stopping at first termination) and report whether the 9.67-point gain over LeWM and the C_best benefit persist.
  2. [Table 1; Table 4] No significance testing is reported for any comparison. With three seeds, the Scene result (69.33±0.94) versus EB-JEPA (67.33±0.94) has overlapping standard deviations, and the Table 4 MSE/Pearson diagnostics show no meaningful difference from LeWM. The paper's claim of consistent gains needs paired tests or confidence intervals across seeds, at least for the headline LeWM comparison and for the C_best ablation.
  3. [Definition 1, Eq. (19), Limitations] The central training signal assumes that temporal separation in a hindsight trajectory implies progress toward the goal. The Limitations section itself concedes that this is coarse for backtracking, detours, or sub-goal switching. Since Cube-S and Scene are play-style datasets collected by non-goal-directed policies, it would strengthen the paper to quantify the fraction of training pairs that actually reduce physical or hyperbolic goal distance, or to test sensitivity to pair-level label noise. Without this, the learned progress structure is plausible but not directly evidenced.
minor comments (6)
  1. [Appendix/Notation] Equation numbers restart in the appendix (e.g., Eq. (1) appears both in the main text and in the appendix). Use distinct numbering or labels to avoid confusion.
  2. [Table 1 vs Table 2] Cube-S ProWorld is reported as 77.33±1.15 in Table 1 but 78% in the ablation tables. Clarify whether ablations use a single seed, a rounded mean, or a different evaluation configuration.
  3. [Table 1, 'Rel. Gain' row] The caption says the last row is relative improvement over the second-best baseline, but for AntMaze-L the second-best is GCIQL (23.33), not LeWM. A footnote should make this explicit to avoid misreading.
  4. [Fig. 7 and Fig. 11] The variable 'Progress' in the correlation matrices is not defined in the main text or appendix. Specify how progress is computed (temporal index, goal-distance reduction, or manual labeling).
  5. [Appendix, 'Success Criterion'] The statement that an episode is counted as successful 'even if the state subsequently deviates from the goal' conflicts with standard termination semantics, where the episode ends at the termination flag. Clarify whether evaluation actually continues after termination and how this interacts with the reported SR.
  6. [Abstract] 'Average absolute success-rate gain' should be phrased as 'average absolute percentage-point gain' to avoid ambiguity with relative gains.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: progress-order supervision is an explicit weak-label assumption, and success is scored by external environment termination; the C_best/transient-success interaction is a correctness concern, not a circularity.

full rationale

The paper's central claim is empirical, not a first-principles derivation. Goal-conditioned progress order (Definition 1) is defined via an unobserved potential and then deliberately approximated by temporally separated hindsight pairs; this is a stated weak-label assumption, not a concealed definition of the evaluation outcome. The cone loss (Eq. 19) and planning cost (Eq. 23) both use hyperbolic distance to the goal, but this is a self-consistent objective design rather than a reduction of prediction to input: success is determined by the environment's native termination signal (Appendix Eq. 2), which is external to the learned latent metric. The C_best term (Eq. 22) may align with the any-step termination criterion and plausibly inflate the reported SR gain, and the Fig. 13 caption's 'post-success drift' is a genuine evaluation-protocol risk, but this is an objective/metric artifact, not a circular argument: C_best is a designed planning term, not a fitted parameter renamed as a prediction. Self-citations involving M. Yang (Bui et al. 2025, Madhu et al. 2026, Zhang et al. 2025) appear only as contextual related-work support for hyperbolic geometry and are not load-bearing for the method's correctness. The Limitations section explicitly acknowledges that temporal-order supervision may be coarse for backtracking, detours, and sub-goal switching; that is an acknowledged assumption rather than a circular step. No equation in the paper reduces to its own inputs by construction, and no load-bearing claim rests on a self-citation chain.

Assumptions & free parameters 9 free parameters · 5 assumptions · 1 invented entities

All four loss terms (pred, contrastive, cone, SIGReg) and the planning cost involve at least one hand-set weight or scale; the sensitivity tables show SR varies by 10-30 points across reasonable choices. The central empirical claim is therefore contingent on a specific hyperparameter configuration that was selected with knowledge of the benchmark outcomes.

free parameters (9)
  • Curvature c = 1.0 (default)
    Hand-set; sensitivity on Cube-S ranges 68-78 SR for c in 0.1-2.0 and learnable, so the default is tuned to the benchmark.
  • Temporal offset Delta = 2
    Tuned on Cube-S; sensitivity 62-78 SR for Delta in 1-10. Controls the span of progress supervision.
  • Planning horizon L = 5 for three tasks, 15 for AntMaze-L
    Tuned on Cube-S; sensitivity 60-78 SR for L in 3-20. Longer horizons accumulate rollout error.
  • Contrastive loss weight lambda_ctr = 0.03
    Joint sensitivity with lambda_cone ranges 40-78 SR; chosen from the 2D sweep (Table 12).
  • Cone loss weight lambda_cone = 0.02
    Joint sensitivity with lambda_ctr (Table 12); moderate values best, 0.05-0.10 degrade SR to 60-72.
  • Planning cost weights beta_b, beta_m = beta_b=0.35, beta_m=0.15 (beta_T=1.0)
    Ablation shows removing either term drops SR from 78 to 64-72; weights are hand-set, not derived.
  • Cone aperture bounds and margin = A_min=0.10, A_max=0.95, m_g=0.03
    Hand-set; fixed-aperture variant drops SR to 70, so the adaptive choice is empirically motivated.
  • Tangent norm bound rho_max = 4.0
    Chosen for numerical stability of the exponential map, per Appendix.
  • Contrastive temperature tau = 0.15
    Set by convention; no sensitivity analysis is reported.
assumptions (5)
  • standard math Lorentz model definitions and exponential/logarithmic map formulas (Eqs. 1-3, 14-16)
    Background geometry is standard and not proved.
  • domain assumption JEPA-style latent prediction is an appropriate substrate for visual goal-conditioned planning
    The entire method builds on the latent world model paradigm from LeWM without justifying it.
  • ad hoc to paper Existence of an unobserved progress potential rho_g and that temporal order in hindsight trajectories approximates it
    Definition 1 postulates rho_g; the Method substitutes temporal separation (Delta) for rho_g. This is the paper's core modeling assumption, acknowledged to fail for detours/backtracking in the Limitations.
  • ad hoc to paper Hyperbolic geometry's coarse-to-fine property captures goal-conditioned progress structure
    Motivated in Fig. 1 and the Introduction but never proved; only the trivial asymmetry of the relation is shown in the Appendix.
  • domain assumption Environment's native terminated signal within a fixed budget is a valid success criterion
    Evaluation counts any termination within budget as success (Evaluation Metrics), a protocol applied to all methods.
invented entities (1)
  • Goal-conditioned progress potential rho_g(s)
    purpose: Defines the ordering s_i <_g s_j that organizes latent-space dynamics; it is never observed directly.
    rho_g is a postulated scalar field; the only handle is temporal order in hindsight trajectories, which the paper itself notes is coarse for detours and subgoal switching.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ProWorld: Progress-Aware Hyperbolic World Models for Long-Horizon Visual Goal Reaching." pith.science (2026). https://pith.science/paper/5IPYJEIX

@misc{pith2026260801926,
  author       = {Pith},
  title        = {Pith review of: ProWorld: Progress-Aware Hyperbolic World Models for Long-Horizon Visual Goal Reaching},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5IPYJEIX}},
  note         = {Machine review of arXiv:2608.01926}
}
read the original abstract

JEPA-style visual world models offer an effective paradigm for visual goal planning by predicting future latent representations. Existing methods typically learn local transition consistency through next-step representation prediction. However, in long-horizon tasks, accurate local prediction alone need not ensure sustained progress toward the goal. First, multi-step rollouts can remain locally plausible while drifting away from goal-relevant trajectories. Second, locally similar future states can correspond to substantially different long-term progress, making them difficult to distinguish in a latent space optimized mainly for local consistency. To address these challenges, we introduce goal-conditioned progress order, a relative ordering of states according to how they advance toward a given goal. This order exhibits an asymmetric, coarse-to-fine structure: early states retain broader future possibilities, while later states concentrate on more specific goal-relevant regions. Such a structure is well suited to hyperbolic geometry. Motivated by this observation, we propose ProWorld, a progress-aware hyperbolic visual world model. ProWorld leverages goal-conditioned progress order to organize visual latent-space dynamics, maintains directional progress within trajectories via hyperbolic entailment learning, and mitigates progress ambiguity among locally similar future states via hyperbolic future discrimination. Furthermore, we design a progress-aware planning objective that scores candidate rollouts by jointly considering proximity to the goal and sustained progress across intermediate states. Experiments on four visual goal-reaching tasks demonstrate that ProWorld achieves an average absolute success-rate gain of 9.67 over LeWM. The code will be released after the paper is accepted.

Figures

Figures reproduced from arXiv: 2608.01926 by the authors.

Figure 1
Figure 1. Core motivation of ProWorld. In Euclidean la [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of ProWorld, including progress-aware [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 5
Figure 5. PushT physical state grid and t-SNE projection of [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: Qualitative planning visualization on the PushT [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 6
Figure 6. Figure 6: Planning-time progress on PushT. ProWorld main [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Spearman correlation analysis of progress-related [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: Diagnostic parent-child radial displacement [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 10
Figure 10. Figure 10: Polar bar visualization of the ablation studies in [PITH_FULL_IMAGE:figures/full_fig_p018_10.png]
Figure 9
Figure 9. Figure 9: Visualization of supplementary hyperparameter [PITH_FULL_IMAGE:figures/full_fig_p018_9.png]
Figure 11
Figure 11. Figure 11: Spearman correlation analysis of progress-related geometric variables on Cube-S. The layout follows the PushT [PITH_FULL_IMAGE:figures/full_fig_p019_11.png]
Figure 12
Figure 12. Figure 12: Open-loop prediction visualizations on PushT. The left column shows the context input, i.e., the ground-truth historical [PITH_FULL_IMAGE:figures/full_fig_p022_12.png]
Figure 13
Figure 13. Figure 13: Open-loop prediction visualizations on Cube-S. The left block shows the context frames, and the right block shows [PITH_FULL_IMAGE:figures/full_fig_p023_13.png]
Figure 14
Figure 14. Figure 14: Training behavior of ProWorld. (a) shows the total objective, hyperbolic prediction, Lorentz contrastive learning, and [PITH_FULL_IMAGE:figures/full_fig_p024_14.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

69 extracted references · 2 linked inside Pith

  1. [1]

    2017 IEEE International Conference on Robotics and Automation (ICRA) , pages =

    Deep Visual Foresight for Planning Robot Motion , author =. 2017 IEEE International Conference on Robotics and Automation (ICRA) , pages =. 2017 , organization =

    2017

  2. [2]

    Proceedings of the 36th International Conference on Machine Learning , pages =

    Learning Latent Dynamics for Planning from Pixels , author =. Proceedings of the 36th International Conference on Machine Learning , pages =. 2019 , series =

    2019

  3. [3]

    Advances in Neural Information Processing Systems , volume =

    Recurrent World Models Facilitate Policy Evolution , author =. Advances in Neural Information Processing Systems , volume =

  4. [4]

    2, 2022-06-27 , author=

    A path towards autonomous machine intelligence version 0.9. 2, 2022-06-27 , author=. Open Review , volume=

    2022

  5. [5]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages =

    Self-Supervised Learning from Images with a Joint-Embedding Predictive Architecture , author =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages =

  6. [6]

    2025 , eprint =

    LeJEPA: Provable and Scalable Self-Supervised Learning Without the Heuristics , author =. 2025 , eprint =

    2025

  7. [7]

    2026 , eprint =

    LeWorldModel: Stable End-to-End Joint-Embedding Predictive Architecture from Pixels , author =. 2026 , eprint =

    2026

  8. [8]

    International Conference on Learning Representations , year =

    OGBench: Benchmarking Offline Goal-Conditioned RL , author =. International Conference on Learning Representations , year =

Show all 69 references
  1. [9]

    Proceedings of the 35th International Conference on Machine Learning , pages =

    Learning Continuous Hierarchies in the Lorentz Model of Hyperbolic Geometry , author =. Proceedings of the 35th International Conference on Machine Learning , pages =. 2018 , series =

    2018

  2. [10]

    Proceedings of the 35th International Conference on Machine Learning , pages =

    Hyperbolic Entailment Cones for Learning Hierarchical Embeddings , author =. Proceedings of the 35th International Conference on Machine Learning , pages =. 2018 , series =

    2018

  3. [11]

    International Conference on Learning Representations , year =

    Dream to Control: Learning Behaviors by Latent Imagination , author =. International Conference on Learning Representations , year =

  4. [12]

    arXiv preprint arXiv:2301.04104 , year =

    Mastering Diverse Domains through World Models , author =. arXiv preprint arXiv:2301.04104 , year =

    arXiv

  5. [13]

    Nickel, Maximilian and Kiela, Douwe , booktitle =. Poincar

  6. [14]

    2026 , eprint =

    GeoWorld: Geometric World Models , author =. 2026 , eprint =

    2026

  7. [15]

    Robotics: Science and Systems , year =

    Diffusion Policy: Visuomotor Policy Learning via Action Diffusion , author =. Robotics: Science and Systems , year =

  8. [16]

    Proceedings of the 5th Conference on Robot Learning , pages =

    Implicit Behavioral Cloning , author =. Proceedings of the 5th Conference on Robot Learning , pages =. 2022 , series =

    2022

  9. [17]

    arXiv preprint arXiv:2602.11389 , year=

    Causal-JEPA: Learning World Models through Object-Level Latent Masking , author=. arXiv preprint arXiv:2602.11389 , year=

    arXiv

  10. [18]

    International Conference on Learning Representations , year=

    Offline Reinforcement Learning with Implicit Q-Learning , author=. International Conference on Learning Representations , year=

  11. [19]

    International Conference on Learning Representations , year=

    Learning to Reach Goals via Iterated Supervised Learning , author=. International Conference on Learning Representations , year=

  12. [20]

    2026 , eprint=

    A Lightweight Library for Energy-Based Joint-Embedding Predictive Architectures , author=. 2026 , eprint=

    2026

  13. [21]

    Hansen, Nicklas and Su, Hao and Wang, Xiaolong , booktitle=

  14. [22]

    2025 , url=

    Zhou, Gaoyue and Pan, Hengkai and LeCun, Yann and Pinto, Lerrel , booktitle=. 2025 , url=

    2025

  15. [23]

    2026 , eprint =

    Sub-JEPA: Subspace Gaussian Regularization for Stable End-to-End World Models , author =. 2026 , eprint =

    2026

  16. [24]

    ICLR Workshop on World Models for Robot Learning , year=

    Stress-Testing Offline Reward-Free Reinforcement Learning: A Case for Planning with Latent Dynamics Models , author=. ICLR Workshop on World Models for Robot Learning , year=

  17. [25]

    Advances in Neural Information Processing Systems , volume =

    Embed to Control: A Locally Linear Latent Dynamics Model for Control from Raw Images , author =. Advances in Neural Information Processing Systems , volume =

  18. [26]

    Advances in Neural Information Processing Systems , volume =

    Long-Horizon Visual Planning with Goal-Conditioned Hierarchical Predictors , author =. Advances in Neural Information Processing Systems , volume =

  19. [27]

    Advances in Neural Information Processing Systems , volume =

    When to Trust Your Model: Model-Based Policy Optimization , author =. Advances in Neural Information Processing Systems , volume =

  20. [28]

    Proceedings of the 2nd Conference on Learning for Dynamics and Control , pages =

    Objective Mismatch in Model-Based Reinforcement Learning , author =. Proceedings of the 2nd Conference on Learning for Dynamics and Control , pages =. 2020 , series =

    2020

  21. [29]

    Proceedings of the 37th International Conference on Machine Learning , pages =

    Goal-Aware Prediction: Learning to Model What Matters , author =. Proceedings of the 37th International Conference on Machine Learning , pages =. 2020 , series =

    2020

  22. [30]

    Proceedings of the 7th Conference on Robot Learning , pages =

    Expansive Latent Planning for Sparse Reward Offline Reinforcement Learning , author =. Proceedings of the 7th Conference on Robot Learning , pages =. 2023 , series =

    2023

  23. [31]

    Advances in Neural Information Processing Systems , volume =

    Contrastive Learning as Goal-Conditioned Reinforcement Learning , author =. Advances in Neural Information Processing Systems , volume =

  24. [32]

    Advances in Neural Information Processing Systems , volume =

    Hyperbolic Graph Convolutional Neural Networks , author =. Advances in Neural Information Processing Systems , volume =

  25. [33]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages =

    Hyperbolic Image Embeddings , author =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages =

  26. [34]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages =

    Hyperbolic Vision Transformers: Combining Improvements in Metric Learning , author =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages =

  27. [35]

    Advances in Neural Information Processing Systems , volume =

    Action-Conditional Video Prediction using Deep Networks in Atari Games , author =. Advances in Neural Information Processing Systems , volume =

  28. [36]

    International Conference on Learning Representations , year =

    Stochastic Variational Video Prediction , author =. International Conference on Learning Representations , year =

  29. [37]

    Proceedings of the 35th International Conference on Machine Learning , pages =

    Stochastic Video Generation with a Learned Prior , author =. Proceedings of the 35th International Conference on Machine Learning , pages =. 2018 , series =

    2018

  30. [38]

    International Conference on Learning Representations , year =

    Mastering Atari with Discrete World Models , author =. International Conference on Learning Representations , year =

  31. [39]

    Nature , volume =

    Mastering Atari, Go, Chess and Shogi by Planning with a Learned Model , author =. Nature , volume =

  32. [40]

    Proceedings of the 39th International Conference on Machine Learning , pages =

    Temporal Difference Learning for Model Predictive Control , author =. Proceedings of the 39th International Conference on Machine Learning , pages =. 2022 , series =

    2022

  33. [41]

    International Conference on Learning Representations , year =

    Transformers are Sample-Efficient World Models , author =. International Conference on Learning Representations , year =

  34. [42]

    Proceedings of the 6th Conference on Robot Learning , pages =

    Masked World Models for Visual Control , author =. Proceedings of the 6th Conference on Robot Learning , pages =. 2023 , series =

    2023

  35. [43]

    Proceedings of the 6th Conference on Robot Learning , pages =

    DayDreamer: World Models for Physical Robot Learning , author =. Proceedings of the 6th Conference on Robot Learning , pages =. 2023 , series =

    2023

  36. [44]

    Transactions on Machine Learning Research , year =

    Revisiting Feature Prediction for Learning Visual Representations from Video , author =. Transactions on Machine Learning Research , year =

  37. [45]

    Advances in Neural Information Processing Systems , volume =

    iVideoGPT: Interactive VideoGPTs are Scalable World Models , author =. Advances in Neural Information Processing Systems , volume =. 2024 , url =

    2024

  38. [46]

    Proceedings of the 32nd International Conference on Machine Learning , pages =

    Universal Value Function Approximators , author =. Proceedings of the 32nd International Conference on Machine Learning , pages =. 2015 , series =

    2015

  39. [47]

    Advances in Neural Information Processing Systems , volume =

    Hindsight Experience Replay , author =. Advances in Neural Information Processing Systems , volume =

  40. [48]

    Advances in Neural Information Processing Systems , volume =

    Visual Reinforcement Learning with Imagined Goals , author =. Advances in Neural Information Processing Systems , volume =

  41. [49]

    Advances in Neural Information Processing Systems , volume =

    Search on the Replay Buffer: Bridging Planning and Reinforcement Learning , author =. Advances in Neural Information Processing Systems , volume =

  42. [50]

    Advances in Neural Information Processing Systems , volume =

    Planning with Goal-Conditioned Policies , author =. Advances in Neural Information Processing Systems , volume =

  43. [51]

    International Conference on Learning Representations , year =

    C-Learning: Learning to Achieve Goals via Recursive Classification , author =. International Conference on Learning Representations , year =

  44. [52]

    Proceedings of the 37th International Conference on Machine Learning , pages =

    Skew-Fit: State-Covering Self-Supervised Reinforcement Learning , author =. Proceedings of the 37th International Conference on Machine Learning , pages =. 2020 , series =

    2020

  45. [53]

    2023 , url =

    Ma, Yecheng Jason and Sodhani, Shagun and Jayaraman, Dinesh and Bastani, Osbert and Kumar, Vikash and Zhang, Amy , booktitle =. 2023 , url =

    2023

  46. [54]

    Proceedings of the 3rd Conference on Robot Learning , pages =

    Learning Latent Plans from Play , author =. Proceedings of the 3rd Conference on Robot Learning , pages =. 2020 , series =

    2020

  47. [55]

    Advances in Neural Information Processing Systems , volume =

    Hyperbolic Neural Networks , author =. Advances in Neural Information Processing Systems , volume =

  48. [56]

    Proceedings of the 35th International Conference on Machine Learning , pages =

    Representation Tradeoffs for Hyperbolic Embeddings , author =. Proceedings of the 35th International Conference on Machine Learning , pages =. 2018 , series =

    2018

  49. [57]

    and Tomioka, Ryota and Teh, Yee Whye , booktitle =

    Mathieu, Emile and Le Lan, Charline and Maddison, Chris J. and Tomioka, Ryota and Teh, Yee Whye , booktitle =. Continuous Hierarchical Representations with Poincar

  50. [58]

    Advances in Neural Information Processing Systems , volume =

    Hyperbolic Graph Neural Networks , author =. Advances in Neural Information Processing Systems , volume =

  51. [59]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages =

    Hyperbolic Visual Embedding Learning for Zero-Shot Recognition , author =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages =

  52. [60]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages =

    Hyperbolic Image Segmentation , author =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages =

  53. [61]

    Proceedings of the 40th International Conference on Machine Learning , pages =

    Hyperbolic Image-Text Representations , author =. Proceedings of the 40th International Conference on Machine Learning , pages =. 2023 , series =

    2023

  54. [62]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages =

    Hyperbolic Contrastive Learning for Visual Representations Beyond Objects , author =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages =

  55. [63]

    2025 , eprint =

    Learning Along the Arrow of Time: Hyperbolic Geometry for Backward-Compatible Representation Learning , author =. 2025 , eprint =

    2025

  56. [64]

    2026 , eprint =

    HypRAG: Hyperbolic Dense Retrieval for Retrieval Augmented Generation , author =. 2026 , eprint =

    2026

  57. [65]

    2022 , eprint =

    Bending the Future: Autoregressive Modeling of Temporal Knowledge Graphs in Curvature-Variable Hyperbolic Spaces , author =. 2022 , eprint =

    2022

  58. [66]

    2025 , eprint =

    HMamba: Hyperbolic Mamba for Sequential Recommendation , author =. 2025 , eprint =

    2025

  59. [67]

    2022 , eprint =

    Hyperbolic Deep Reinforcement Learning , author =. 2022 , eprint =

    2022

  60. [68]

    2026 , eprint =

    Understanding and Improving Hyperbolic Deep Reinforcement Learning , author =. 2026 , eprint =

    2026

  61. [69]

    2026 , eprint =

    HyperGuide: Hyperbolic Guidance for Efficient Multi-Step Reasoning in Large Language Models , author =. 2026 , eprint =

    2026

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.