Pith

open record

sign in

arxiv: 2607.06401 · v1 · pith:WALIB4IW · submitted 2026-07-07 · cs.AI

A Definition and Roadmap for World Models

Xinyuan Chen , Haoyu Guo , Shi Guo , Bingqi Jiang , Chunhua Shen , Xing Shen , Tianfan Xue , Yufei Xue
show 5 more authors
Mulin Yu Weinan Zhang Bin Zhao Bowen Zhou Ming Zhou
This is my paper

Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel 2026-07-08 06:53 UTCglm-5.2pith:WALIB4IWrecord.jsonopen to challenge →

classification cs.AI
keywords world modelslossy compressionphysical representationembodied AImodel-based reinforcement learningPOMDPworld action modelssim-to-real transfer
0
0 comments X

The pith

A world model is a compression problem, not a generation problem

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

The paper argues that the term 'world model' has become overloaded across AI subfields—applied to video generators, reinforcement-learning dynamics models, and embodied planners without a shared definition. The authors propose a single formal definition: a world model is a compression of the physical world's state-transition processes, built under finite computational resources. Under this definition, the core task is not to generate realistic pixels or simulate futures, but to distill high-dimensional sensory data into a compact internal representation that preserves the physical, causal, and dynamical structure needed for reasoning, planning, and control. Generation and simulation are downstream decodings of this representation, not the objective itself. The paper frames understanding as primary and prediction as secondary—a model must first identify what is happening and why before it can predict what will happen. It situates this definition within a POMDP agent-environment loop, proposes a two-dimensional taxonomy (functional role × architectural substrate) that supersedes prior one-dimensional classifications, and lays out a staged roadmap: first unify modalities, then distill them into a single shared physical representation from which rendering, simulation, and planning are all decoded, and finally scale this into foundation-level interactive simulators. The paper also introduces an 'Inverted Pyramid' data pipeline that funnels internet-scale video down to compact task-aligned robot training data, arguing that data diversity—not architecture or compute—sets the ceiling on physical generalization.

Core claim

The central conceptual move is the redefinition of a world model as a compression of physical state transitions under finite resources, coupled with the principle that understanding should be primary while prediction serves it. This reframing dissolves the boundary between renderers, simulators, and planners: they become different decoding operations on a single compressed internal state. The paper calls this the 'one state, many decoders' principle and argues that the field's central open problem is discovering what compact internal structure can preserve sufficient physical and semantic information to support all downstream projections of an embodied intelligence.

What carries the argument

The load-bearing mechanism is the unified physical representation—a single, compact, learned internal state encoding geometry, motion, material properties, appearance, semantics, uncertainty, and interaction state jointly. From this state, rendering (producing visual observations), simulation (propagating physical dynamics under action), and planning (evaluating counterfactual futures to select actions) are recovered as distinct decoding operations. The POMDP loop provides the formal scaffolding: the agent maintains a Bayesian belief over latent states, the world model approximates both the transition kernel and observation model, and the planner optimizes over the resulting belief-continged

If this is right

  • If the compression-first definition is correct, then evaluating world models by visual fidelity (FID, FVD) is measuring the wrong thing; the right metrics should test whether the compressed state preserves decision-relevant physical structure—contact dynamics, causality, object permanence—sufficient for downstream control.
  • The 'one state, many decoders' principle implies that current systems maintaining separate representations for rendering (radiance fields), simulation (meshes/particles), and planning (occupancy grids) are architecturally suboptimal; a single shared substrate should eventually replace them.
  • The Inverted Pyramid pipeline implies that the path to generalist robot policies runs through internet-scale video, not through more robot data collection—data diversity from passive video, not interaction volume, sets the generalization ceiling.
  • The Trinity Architecture (Agent–Evaluator–World Model) implies a self-improving loop where the world model generates curricula at the edge of the agent's current capability, suggesting that world models are not just predictors but active drivers of embodied skill acquisition.
  • If understanding is primary and prediction is derivative, then architectures that skip explicit representation learning (pure pixel-to-pixel video generators) will plateau as physical reasoners regardless of scale, because they lack the compressed state that carries causal and dynamical structure.

Where Pith is reading between the lines

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

  • The compression framing implicitly assumes a favorable information-theoretic structure: that the physical regularities relevant for control occupy a low-dimensional manifold within high-dimensional sensory data. If the decision-relevant physical state for contact-rich manipulation is actually high-dimensional (e.g., requiring fine-grained deformation fields or friction distributions), then lossy c
  • The claim that data diversity sets the ceiling while architecture only affects efficiency echoes scaling-law arguments from language modeling, but physical data may not obey the same power-law regime: the long tail of rare physical events (edge cases in contact, fracture, anomalous dynamics) may be so structurally diverse that no finite corpus captures sufficient coverage, making the 'ceiling' a m
  • The Trinity Architecture's self-curriculum loop assumes the world model can reliably identify the edge of the agent's capability. If the model's uncertainty estimates are poorly calibrated—a known problem in deep generative models—the curriculum will either propose trivial tasks (stalling progress) or impossible ones (wasting interaction), turning the self-improvement loop into a random walk.

Load-bearing premise

The paper assumes that a single, unified, compact internal representation can be learned from internet-scale video data that preserves sufficient physical structure—geometry, contact, dynamics, causality—to support rendering, simulation, and planning simultaneously. This 'one state, many decoders' principle presupposes that the information loss inherent in lossy compression will not destroy decision-relevant physical details, which remains unproven for complex, contact-rich,

What would settle it

A contact-rich manipulation task where a visually plausible but physically incorrect compressed representation causes a robot to fail in ways that a less compressed but more physically faithful representation would not—demonstrating that the compression objective and the control objective are fundamentally in tension rather than aligned.

Figures

Figures reproduced from arXiv: 2607.06401 by Bingqi Jiang, Bin Zhao, Bowen Zhou, Chunhua Shen, Haoyu Guo, Ming Zhou, Mulin Yu, Shi Guo, Tianfan Xue, Weinan Zhang, Xing Shen, Xinyuan Chen, Yufei Xue.

Figure 1
Figure 1. Figure 1: This diagram illustrates a data pyramid inversion funnel pipeline for robot and phys [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: An illustration of the nature of the world model and its major properties. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: A schematic of the POMDP agent–environment loop using the notation of this subsection. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Functional roles of world models in the POMDP loop. [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: A Two-Dimensional Taxonomy: Functions and Architectures. [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Unification trend of world-model architectures. Generation–understanding unified mod [PITH_FULL_IMAGE:figures/full_fig_p015_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: An overview of major training and learning paradigms for world models. [PITH_FULL_IMAGE:figures/full_fig_p017_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: World model application on scientific discovery [PITH_FULL_IMAGE:figures/full_fig_p030_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: A staged roadmap for developing next-generation world models. [PITH_FULL_IMAGE:figures/full_fig_p036_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: An overview of the trinity architecture. [PITH_FULL_IMAGE:figures/full_fig_p040_10.png] view at source ↗
read the original abstract

World models -- internal simulators that learn the structure and dynamics of an environment -- have become one of the most actively debated concepts in AI. From model-based reinforcement learning and video generation to embodied robotics and ultimately, physical AI, researchers across AI subfields are building systems that they call "world models", yet there is no consensus on what a world model fundamentally is, what it should predict, or how it should be built. This perspective article provides a scientific definition of world models, discussions of their key technical aspects, and a staged roadmap for developing effective world models.

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 / 7 minor

Summary. This perspective article proposes a scientific definition of world models as compression models of physical state-transition processes under finite computational constraints, and lays out a staged roadmap progressing from unified multimodal models through unified physical representations to foundation-scale interactive simulators. The paper synthesizes a broad literature spanning model-based RL, video generation, 3D representations, causal inference, and embodied AI, organizing it through a two-dimensional taxonomy (functional roles × architectural substrates) and introducing conceptual constructs including the Inverted Pyramid Workflow, the 'one state, many decoders' principle, and a Trinity Architecture for physical AGI. The mathematical formulations (POMDPs, Bayesian filtering, structural causal models, MBRL objectives) are standard and correctly applied.

Significance. The paper's primary contribution is conceptual: it provides a unifying definition and organizational framework for a field that currently lacks terminological consensus. The compression-first framing (Definition 2.1) is a reasonable and potentially useful lens that connects information-theoretic principles to architectural decisions. The two-dimensional taxonomy (Section 2.5, Figure 5) and the functional taxonomy discussion (Section 2.4) are genuinely clarifying for a fragmented literature. The paper does not present machine-checked proofs, reproducible code, or parameter-free derivations, but it does offer falsifiable predictions—particularly the claim that a single shared physical representation can simultaneously support rendering, simulation, and planning (Section 7.2)—which could in principle be tested. The breadth of literature coverage is substantial and the roadmap is concrete enough to guide near-term research. The discussion of counterfactual reasoning (Section 4.6) and physics-informed learning (Section 4.5) are well-grounded treatments that connect world models to established formal frameworks.

major comments (3)
  1. Section 7.2 articulates the 'one state, many decoders' principle as the central architectural thesis of the roadmap: a single compact internal representation should support rendering, simulation, and planning simultaneously. The paper itself acknowledges the core risk in Section 3.2: 'excessive abstraction may discard information that is visually small but decision-critical, such as gripper-object contact, thin obstacles, subtle object pose changes, or deformation.' However, no analytical or empirical argument is provided for why lossy compression would preserve exactly this class of information. The concern is sharpened by the Inverted Pyramid Workflow (Figure 1), which proposes learning primarily from internet video—a data source that lacks proprioceptive, force, and tactile channels. The paper should either (a) provide an information-theoretic or empirical argument for why the unified
  2. representation can retain contact-rich physical details when trained on data that does not contain them, or (b) explicitly acknowledge this as a fundamental limitation and scope the 'one state, many decoders' claim accordingly. As stated, the gap between the compression-first definition (which is sound) and the unified-representation roadmap claim (which is unproven) is the weakest load-bearing link in the argument.
  3. Section 8 introduces the 'Trinity Architecture' (Agent–Evaluator–World Model) as a cognitive loop for autonomous evolution toward physical AGI. This construct is presented at a conceptual level without connection to the formal POMDP/MBRL framework developed in Sections 2–4. Specifically, the relationship between the Evaluator component and the reward function R in the POMDP formulation (Section 2.2) is unclear, and the claim that the World Model component 'knows the edge of feasible tasks of the current Actor' is not grounded in any formal mechanism. The Trinity Architecture would benefit from either formalization within the existing framework or explicit acknowledgment that it is a speculative conceptual proposal.
minor comments (7)
  1. Section 2.5 references 'Figure 1' for the many-to-many mapping of systems to taxonomy categories, but Figure 1 depicts the Inverted Pyramid Workflow. The intended reference appears to be Figure 5. Please verify cross-references.
  2. Several typos throughout: 'internaml model' (Section 1), 'interchangably' (Section 2.1 footnote), 'adanced' and 'genralist' (Section 4.2), 'reies' and 'dexterous manipulation tasks' (Section 4.2), 'natually' (Section 8). Proofreading pass recommended.
  3. Section 4.2, paragraph on DreamZero: the sentence beginning 'However, DreamZero does not really resolve the covariate shift...' is grammatically fragmented and should be revised for clarity.
  4. The paper cites numerous 2026-dated works (e.g., NVIDIA Cosmos Team 2026, ByteDance Seed Team 2026, several arXiv preprints). Given the July 2026 submission date, these may be very recent preprints. Where possible, the authors should verify that cited preprints have stable identifiers and that claims about these systems are accurate as of the cited versions.
  5. Table 1 is informative but would benefit from a column indicating which specific limitations are most critical for each paradigm, rather than listing them uniformly. This would improve usability as a reference.
  6. Section 6.6 introduces 'Federated World Models' and 'Trusted Execution Environments' as proposed solutions for privacy-preserving governance. These concepts are introduced without prior context or literature grounding. A brief reference to existing federated learning or secure computation literature would help readers calibrate their expectations.
  7. Equation (4) presents scaling laws with the notation C≈κND, but the constant κ is not defined in the surrounding text. Please add a definition.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for a careful and constructive reading of our manuscript. The two major comments both identify genuine gaps in the argument that we will address in revision. On the first comment, regarding the 'one state, many decoders' principle and the Inverted Pyramid Workflow's reliance on internet video lacking proprioceptive channels, we agree that the manuscript does not adequately address how contact-rich physical details can be preserved when training on data that does not contain them. We will revise Section 7.2 to explicitly scope the claim and acknowledge this as a fundamental limitation requiring complementary data sources. On the second comment, regarding the Trinity Architecture's lack of formal connection to the POMDP/MBRL framework, we agree that Section 8 is presented at a conceptual level without grounding in the formal apparatus of Sections 2-4. We will add an explicit mapping between the Evaluator and the reward function R, clarify the World Model's 'edge of feasible tasks' claim in terms of model uncertainty and reachable-state estimation, and acknowledge the speculative nature of the proposal where formalization is not yet possible.

read point-by-point responses

  1. Referee: Section 7.2 'one state, many decoders' principle: no argument provided for why lossy compression preserves contact-rich details when trained on internet video lacking proprioceptive/force/tactile channels. Paper should either (a) provide an information-theoretic or empirical argument, or (b) acknowledge as fundamental limitation and scope the claim.

    Authors: The referee correctly identifies the weakest load-bearing link in our argument. We do not have an information-theoretic proof that lossy compression trained on internet video will preserve contact-rich physical details, and we agree that the Inverted Pyramid Workflow's reliance on video data lacking proprioceptive, force, and tactile channels creates a genuine gap between the compression-first definition (which is sound as a general principle) and the unified-representation roadmap claim (which is unproven for this data regime). We cannot honestly provide argument (a): there is no existing theoretical result guaranteeing that information absent from the training distribution can be recovered by compression, and empirical evidence from current video-based world models (as we note in Section 6.2) shows that perceptual fidelity does not imply physical precision. We will therefore adopt option (b). Specifically, we will revise Section 7.2 to: (1) explicitly state that the 'one state, many decoders' principle is an architectural thesis whose feasibility for contact-rich domains is not yet established; (2) acknowledge that internet video, while encoding broad physical priors (object permanence, rigidity, kinematic structure), does not contain proprioceptive, force, or tactile channels, and that no compression mechanism can recover information entirely absent from the training data; (3) scope the claim to the modalities present in the training distribution, noting that contact-rich and tactile information require complementary data sources (embodied interaction data, tactile sensors, physics-informed constraints as discussed in Section 4.5); and (4) add a forward reference to Section 6.1 (Data Asymmetry) and Section 6.4 (Sim-to-Real Transfer), which already discuss these bottl revision: no

  2. Referee: Section 8 Trinity Architecture presented conceptually without connection to formal POMDP/MBRL framework. Relationship between Evaluator and reward function R unclear. Claim that World Model 'knows the edge of feasible tasks' not grounded in formal mechanism. Should formalize within existing framework or acknowledge as speculative.

    Authors: We agree that Section 8 is insufficiently connected to the formal framework developed in Sections 2-4, and we will revise accordingly. We can partially formalize the mapping: (1) The Evaluator component corresponds to the reward function R in the POMDP formulation (Section 2.2), assessing trajectory quality against task objectives. We will state this mapping explicitly. (2) The Agent corresponds to the policy pi_phi, and the World Model corresponds to the learned transition model P_hat_theta, both already formalized in Section 4.2. (3) The claim that the World Model 'knows the edge of feasible tasks of the current Actor' can be partially grounded in the model uncertainty and reachable-state estimation literature discussed in Section 4.2: ensemble disagreement (Chua et al., 2018), conservative rollout termination (Yu et al., 2020; Kidambi et al., 2020), and the horizon-limited value gap (Eq. 19) all provide formal mechanisms for estimating where the model's predictions become unreliable, which operationalizes 'the edge of feasible tasks.' We will add these connections explicitly. However, we acknowledge that the Trinity Architecture's claim of autonomous curriculum generation—where the World Model proposes tasks 'just beyond' the Agent's current limits—goes beyond what existing formal mechanisms fully support. The automatic curriculum learning literature (e.g., Plan2Explore's intrinsic motivation, Sekar et al., 2020) provides partial precedents, but the closed-loop self-evolution we describe remains a speculative conceptual proposal. We will state this explicitly in the revised Section 8, marking it as a research direction rather than a formally grounded result. revision: no

Circularity Check

0 steps flagged

No circularity found: perspective article with conceptual definitions and literature synthesis, no derivation chain that reduces to inputs by construction.

full rationale

This is a perspective/roadmap article that proposes a conceptual definition (Definition 2.1: world model as compression of state transitions under finite resources) and synthesizes external literature into taxonomies and a staged roadmap. The equations presented (Eq. 1–19) are standard formulations from POMDPs, MBRL, causal inference (Pearl's SCM framework), and scaling laws — they are used to frame discussion, not to derive novel quantitative predictions from fitted parameters. The central claims (the compression-first definition, the 'one state, many decoders' principle, the Inverted Pyramid Workflow, the Trinity Architecture) are conceptual proposals, not derivations whose outputs could reduce to their inputs by construction. Self-citations (e.g., DreamZero/Ye et al. 2026b, HERMES/Zhou et al. 2025a, τ0-WM/Zhou et al. 2026a) appear as illustrative examples within the taxonomy, not as load-bearing premises that define the central argument. No step in the paper exhibits the pattern where a 'prediction' or 'first-principles result' is equivalent to its inputs by definition, fit, or self-citation chain. The paper's weaknesses (e.g., the unproven assumption that lossy compression preserves contact-rich physical details) are correctness/completeness risks, not circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 3 axioms · 2 invented entities

The paper introduces no free parameters as it contains no empirical fitting. It relies on domain assumptions about data scaling and representation learning. It invents two conceptual entities (Trinity Architecture, Inverted Pyramid Workflow) as proposed solutions without independent empirical evidence within the paper.

axioms (3)
  • domain assumption Data diversity determines the generalization ceiling of an intelligent system at fixed architecture and compute.
    Stated in the Introduction: 'At a fixed architecture and compute budget, it is data diversity that determines this ceiling, not model structure or training hyperparameters.' This is a strong assumption underlying the Inverted Pyramid Workflow.
  • ad hoc to paper A single shared physical representation can support rendering, simulation, and planning simultaneously.
    Section 7.2 proposes the 'one state, many decoders' principle, assuming that a compact internal state can preserve sufficient information for all downstream tasks without lossy translation.
  • domain assumption Internet video contains sufficient implicit physical priors for embodied control.
    The Introduction argues that internet video encodes object permanence, rigidity, kinematics, and causal logic, and that these can be extracted via compression. This is an unproven premise for the entire roadmap.
invented entities (2)
  • Trinity Architecture no independent evidence
    purpose: A three-part cognitive loop (Actor, Evaluator, World Model) for autonomous physical learning and self-evolution.
    Proposed in Section 8 as a conceptual blueprint for Physical AGI. No implementation or empirical validation is provided.
  • Inverted Pyramid Workflow no independent evidence
    purpose: A data pipeline that funnels internet video into compact, task-aligned robotic training data.
    Described in the Introduction and Figure 1. It is a proposed workflow, not a system that has been built and tested end-to-end.

pith-pipeline@v1.1.0-glm · 50774 in / 2131 out tokens · 571738 ms · 2026-07-08T06:53:03.160464+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

284 extracted references · 284 canonical work pages · 60 internal anchors

  1. [1]

    Advances in Neural Information Processing Systems (NeurIPS) , year =

    MOPO: Model-based offline policy optimization , author =. Advances in Neural Information Processing Systems (NeurIPS) , year =

    work page

  2. [2]

    Advances in Neural Information Processing Systems (NeurIPS) , volume=

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

    work page

  3. [3]

    Advances in Neural Information Processing Systems (NeurIPS) , volume=

    Model-Based Imitation Learning for Urban Driving , author=. Advances in Neural Information Processing Systems (NeurIPS) , volume=

    work page

  4. [4]

    ReSim: Reliable World Simulation for Autonomous Driving

    ReSim: Reliable World Simulation for Autonomous Driving , author=. arXiv preprint arXiv:2506.09981 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  5. [5]

    Conference on Robot Learning (CoRL) , year=

    Evaluating Real-World Robot Manipulation Policies in Simulation , author=. Conference on Robot Learning (CoRL) , year=

    work page

  6. [6]

    Conference on Robot Learning (CoRL) , year=

    DayDreamer: World Models for Physical Robot Learning , author=. Conference on Robot Learning (CoRL) , year=

    work page

  7. [7]

    International Conference on Learning Representations (ICLR) , year=

    Learning Interactive Real-World Simulators , author=. International Conference on Learning Representations (ICLR) , year=

    work page

  8. [8]

    International Conference on Machine Learning (ICML) , year=

    RoboDreamer: Learning Compositional World Models for Robot Imagination , author=. International Conference on Machine Learning (ICML) , year=

    work page

  9. [9]

    Conference on Robot Learning (CoRL) , year=

    RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control , author=. Conference on Robot Learning (CoRL) , year=

    work page

  10. [10]

    Conference on Robot Learning (CoRL) , year=

    OpenVLA: An Open-Source Vision-Language-Action Model , author=. Conference on Robot Learning (CoRL) , year=

    work page

  11. [11]

    $\pi_0$: A Vision-Language-Action Flow Model for General Robot Control

    _0 : A Vision-Language-Action Flow Model for General Robot Control , author=. arXiv preprint arXiv:2410.24164 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  12. [12]

    arXiv preprint arXiv:2509.13780 , year=

    Behavior Foundation Model for Humanoid Robots , author=. arXiv preprint arXiv:2509.13780 , year=

    work page arXiv

  13. [13]

    IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , note=

    Planning-Oriented Autonomous Driving , author=. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , note=

    work page

  14. [14]

    Advances in Neural Information Processing Systems (NeurIPS) , year=

    Vista: A Generalizable Driving World Model with High Fidelity and Versatile Controllability , author=. Advances in Neural Information Processing Systems (NeurIPS) , year=

    work page

  15. [15]

    1995 , publisher=

    Epistemic Logic for AI and Computer Science , author=. 1995 , publisher=

    work page 1995

  16. [16]

    Advanced Robotics , volume=

    World models and predictive coding for cognitive and developmental robotics: frontiers and challenges , author=. Advanced Robotics , volume=. 2023 , publisher=

    work page 2023

  17. [17]

    Proceedings of the 24th International Conference on Autonomous Agents and Multiagent Systems , pages=

    Unlocking the potential of decentralized llm-based mas: Privacy preservation and monetization in collective intelligence , author=. Proceedings of the 24th International Conference on Autonomous Agents and Multiagent Systems , pages=

    work page

  18. [18]

    The wall confronting large language models

    The wall confronting large language models , author=. arXiv preprint arXiv:2507.19703 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  19. [19]

    General Agentic Planning Through Simulative Reasoning with World Models

    General Agentic Planning Through Simulative Reasoning with World Models , author=. arXiv preprint arXiv:2507.23773 , year=. doi:10.48550/arXiv.2507.23773 , url=

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2507.23773

  20. [20]

    2022 , eprint=

    Model-based Reinforcement Learning: A Survey , author=. 2022 , eprint=

    work page 2022

  21. [21]

    ACM SIGART Bulletin , volume=

    Dyna, an integrated architecture for learning, planning, and reacting , author=. ACM SIGART Bulletin , volume=. 1991 , publisher=

    work page 1991

  22. [22]

    Lewis, David Scott and Zueco, Enrique , booktitle =

    work page

  23. [23]

    Safety, Security, and Cognitive Risks in World Models

    Safety, Security, and Cognitive Risks in World Models , author=. arXiv preprint arXiv:2604.01346 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  24. [24]

    Human factors , volume=

    Humans and automation: Use, misuse, disuse, abuse , author=. Human factors , volume=. 1997 , publisher=

    work page 1997

  25. [25]

    2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages=

    Neuromorphic force-control in an industrial task: validating energy and latency benefits , author=. 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages=. 2024 , organization=

    work page 2024

  26. [26]

    Science , volume=

    Designing and using robots for environmental sustainability , author=. Science , volume=. 2025 , publisher=

    work page 2025

  27. [27]

    Patterns , volume=

    A view of the sustainable computing landscape , author=. Patterns , volume=. 2025 , publisher=

    work page 2025

  28. [28]

    Towards World Models in Biomedical Research

    Towards World Models in Biomedical Research , author=. arXiv preprint arXiv:2606.05925 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  29. [29]

    Nature Machine Intelligence , volume=

    A roadmap for AI in robotics , author=. Nature Machine Intelligence , volume=. 2025 , publisher=

    work page 2025

  30. [30]

    , booktitle =

    Deisenroth, Marc and Rasmussen, Carl E. , booktitle =. 2011 , publisher =

    work page 2011

  31. [31]

    2018 IEEE International Conference on Robotics and Automation (ICRA) , pages =

    Neural Network Dynamics for Model-Based Deep Reinforcement Learning with Model-Free Fine-Tuning , author =. 2018 IEEE International Conference on Robotics and Automation (ICRA) , pages =. 2018 , organization =

    work page 2018

  32. [32]

    Advances in Neural Information Processing Systems (NeurIPS) , volume =

    Deep Reinforcement Learning in a Handful of Trials using Probabilistic Dynamics Models , author =. Advances in Neural Information Processing Systems (NeurIPS) , volume =

    work page

  33. [33]

    Advances in Neural Information Processing Systems (NeurIPS) , volume =

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

    work page

  34. [34]

    International Conference on Learning Representations (ICLR) , year =

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

    work page

  35. [35]

    Mastering

    Hafner, Danijar and Lillicrap, Timothy and Norouzi, Mohammad and Ba, Jimmy , booktitle =. Mastering

    work page

  36. [36]

    Proceedings of the 7th International Conference on Machine Learning (ICML) , pages =

    Integrated Architectures for Learning, Planning, and Reacting Based on Approximating Dynamic Programming , author =. Proceedings of the 7th International Conference on Machine Learning (ICML) , pages =. 1990 , publisher =

    work page 1990

  37. [37]

    and Boots, Byron and Theodorou, Evangelos A

    Williams, Grady and Wagener, Nolan and Goldfain, Brian and Drews, Paul and Rehg, James M. and Boots, Byron and Theodorou, Evangelos A. , booktitle =. Information Theoretic. 2017 , organization =

    work page 2017

  38. [38]

    Advances in Neural Information Processing Systems (NeurIPS) , volume =

    Learning Continuous Control Policies by Stochastic Value Gradients , author =. Advances in Neural Information Processing Systems (NeurIPS) , volume =

    work page

  39. [39]

    Nature , volume=

    Mastering Atari, go, chess and shogi by planning with a learned model , author=. Nature , volume=. 2020 , publisher=

    work page 2020

  40. [40]

    and Levine, Sergey and Finn, Chelsea and Ma, Tengyu , title =

    Yu, Tianhe and Thomas, Garrett and Yu, Lantao and Ermon, Stefano and Zou, James Y. and Levine, Sergey and Finn, Chelsea and Ma, Tengyu , title =. Advances in Neural Information Processing Systems (NeurIPS) , year =

    work page

  41. [41]

    Advances in Neural Information Processing Systems (NeurIPS) , year =

    Kidambi, Rahul and Rajeswaran, Aravind and Netrapalli, Praneeth and Joachims, Thorsten , title =. Advances in Neural Information Processing Systems (NeurIPS) , year =

    work page

  42. [42]

    Learning for Dynamics and Control (L4DC) , year =

    Lambert, Nathan and Amos, Brandon and Yadan, Omry and Calandra, Roberto , title =. Learning for Dynamics and Control (L4DC) , year =

    work page

  43. [43]

    Advances in Neural Information Processing Systems (NeurIPS) , year =

    Curi, Sebastian and Berkenkamp, Felix and Krause, Andreas , title =. Advances in Neural Information Processing Systems (NeurIPS) , year =

    work page

  44. [44]

    Value-Aware Loss Function for Model-Based Reinforcement Learning , booktitle =

    Farahmand, Amir-massoud and Barreto, Andr. Value-Aware Loss Function for Model-Based Reinforcement Learning , booktitle =

    work page

  45. [45]

    The Value Equivalence Principle for Model-Based Reinforcement Learning , booktitle =

    Grimm, Christopher and Barreto, Andr. The Value Equivalence Principle for Model-Based Reinforcement Learning , booktitle =

    work page

  46. [46]

    and Silver, David , title =

    Antonoglou, Ioannis and Schrittwieser, Julian and Ozair, Sherjil and Hubert, Thomas K. and Silver, David , title =. International Conference on Learning Representations (ICLR) , year =

    work page

  47. [47]

    Proceedings of the AAAI Conference on Artificial Intelligence , year =

    Talvitie, Erik , title =. Proceedings of the AAAI Conference on Artificial Intelligence , year =

    work page

  48. [48]

    Investigating Compounding Prediction Errors in Learned Dynamics Models

    Lambert, Nathan and Pister, Kristofer and Calandra, Roberto , title =. arXiv preprint arXiv:2203.09637 , year =

    work page internal anchor Pith review Pith/arXiv arXiv

  49. [49]

    Proceedings of the 38th International Conference on Machine Learning (ICML) , year =

    Hubert, Thomas and Schrittwieser, Julian and Antonoglou, Ioannis and Barekatain, Mohammadamin and Schmitt, Simon and Silver, David , title =. Proceedings of the 38th International Conference on Machine Learning (ICML) , year =

    work page

  50. [50]

    Proceedings of the 39th International Conference on Machine Learning (ICML) , year =

    Hansen, Nicklas and Wang, Xiaolong and Su, Hao , title =. Proceedings of the 39th International Conference on Machine Learning (ICML) , year =

    work page

  51. [51]

    and Darrell, Trevor , title =

    Pathak, Deepak and Agrawal, Pulkit and Efros, Alexei A. and Darrell, Trevor , title =. Proceedings of the 34th International Conference on Machine Learning (ICML) , year =

    work page

  52. [52]

    International Conference on Learning Representations (ICLR) , year =

    Burda, Yuri and Edwards, Harrison and Storkey, Amos and Klimov, Oleg , title =. International Conference on Learning Representations (ICLR) , year =

    work page

  53. [53]

    Advances in Neural Information Processing Systems (NeurIPS) , year =

    Houthooft, Rein and Chen, Xi and Duan, Yan and Schulman, John and De Turck, Filip and Abbeel, Pieter , title =. Advances in Neural Information Processing Systems (NeurIPS) , year =

    work page

  54. [54]

    Proceedings of the 37th International Conference on Machine Learning (ICML) , year =

    Sekar, Ramanan and Rybkin, Oleh and Daniilidis, Kostas and Abbeel, Pieter and Hafner, Danijar and Pathak, Deepak , title =. Proceedings of the 37th International Conference on Machine Learning (ICML) , year =

    work page

  55. [55]

    and Barto, Andrew G

    Sutton, Richard S. and Barto, Andrew G. , title =

    work page

  56. [56]

    Nature , volume =

    Silver, David and Schrittwieser, Julian and Simonyan, Karen and Antonoglou, Ioannis and Huang, Aja and Guez, Arthur and Hubert, Thomas and Baker, Lucas and Lai, Matthew and Bolton, Adrian and others , title =. Nature , volume =

    work page

  57. [57]

    and Guez, Arthur and Sifre, Laurent and van den Driessche, George and Schrittwieser, Julian and Antonoglou, Ioannis and Panneershelvam, Veda and Lanctot, Marc and others , title =

    Silver, David and Huang, Aja and Maddison, Chris J. and Guez, Arthur and Sifre, Laurent and van den Driessche, George and Schrittwieser, Julian and Antonoglou, Ioannis and Panneershelvam, Veda and Lanctot, Marc and others , title =. Nature , volume =

    work page

  58. [58]

    Kahneman, Daniel , title =

    work page

  59. [59]

    Advances in Neural Information Processing Systems (NeurIPS) , year =

    Revisiting the Linearization Trick for Zero-Shot Multi-Agent Communication , author =. Advances in Neural Information Processing Systems (NeurIPS) , year =

    work page

  60. [60]

    Proceedings of the 14th International Conference on Artificial Intelligence and Statistics (AISTATS) , pages =

    A Reduction of Imitation Learning and Structured Prediction to No-Regret Online Learning , author =. Proceedings of the 14th International Conference on Artificial Intelligence and Statistics (AISTATS) , pages =

    work page

  61. [61]

    Proceedings of the 2nd Conference on Learning for Dynamics and Control (L4DC) , pages =

    Model-Based Meta-Policy Optimization , author =. Proceedings of the 2nd Conference on Learning for Dynamics and Control (L4DC) , pages =. 2020 , publisher =

    work page 2020

  62. [62]

    and Tsitsiklis, John N

    Konda, Vijay R. and Tsitsiklis, John N. , pages =. Actor-Critic Algorithms , type =

    work page

  63. [63]

    Mastering

    Hafner, Danijar and Lillicrap, Timothy and Norouzi, Mohammad and Ba, Jimmy , booktitle =. Mastering. 2021 , url =

    work page 2021

  64. [64]

    Nature , volume =

    Mastering diverse control tasks through world models , author =. Nature , volume =. 2025 , publisher =

    work page 2025

  65. [65]

    Nature , year =

    Mastering Atari, Go, chess and shogi by planning with a learned model , author =. Nature , year =. doi:10.1038/s41586-020-03051-4 , publisher =

    work page internal anchor Pith review doi:10.1038/s41586-020-03051-4

  66. [66]

    How Far is Video Generation from World Model: A Physical Law Perspective

    How far is video generation from world model: A physical law perspective , author=. arXiv preprint arXiv:2411.02385 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  67. [67]

    arXiv e-prints , pages=

    Do generative video models learn physical principles from watching videos? , author=. arXiv e-prints , pages=

    work page

  68. [68]

    Evaluating Newtonian Mechanics in Video Generative Models with Real Physical Systems

    Evaluating Newtonian Mechanics in Video Generative Models with Real Physical Systems , author=. arXiv preprint arXiv:2504.02918 , year=. doi:10.48550/arXiv.2504.02918 , url=

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2504.02918

  69. [69]

    Towards World Simulator: Crafting Physical Commonsense-Based Benchmark for Video Generation

    Towards world simulator: Crafting physical commonsense-based benchmark for video generation , author=. arXiv preprint arXiv:2410.05363 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  70. [70]

    "PhyWorldBench": A Comprehensive Evaluation of Physical Realism in Text-to-Video Models

    " PhyWorldBench": A Comprehensive Evaluation of Physical Realism in Text-to-Video Models , author=. arXiv preprint arXiv:2507.13428 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  71. [71]

    VBench-2.0: Advancing Video Generation Benchmark Suite for Intrinsic Faithfulness

    Vbench-2.0: Advancing video generation benchmark suite for intrinsic faithfulness , author=. arXiv preprint arXiv:2503.21755 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  72. [72]

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

    RoboWM-Bench: A Benchmark for Evaluating World Models in Robotic Manipulation , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  73. [73]

    arXiv preprint arXiv:2602.08971 , year=

    Worldarena: A unified benchmark for evaluating perception and functional utility of embodied world models , author=. arXiv preprint arXiv:2602.08971 , year=

    work page arXiv

  74. [74]

    PhyWorld: Physics-Faithful World Model for Video Generation

    PhyWorld: Physics-Faithful World Model for Video Generation , author=. arXiv preprint arXiv:2605.19242 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  75. [75]

    arXiv preprint arXiv:2603.23376 , year=

    Abot-physworld: Interactive world foundation model for robotic manipulation with physics alignment , author=. arXiv preprint arXiv:2603.23376 , year=

    work page arXiv

  76. [76]

    arXiv preprint arXiv:2511.03997 , year=

    PhysCorr: Dual-Reward DPO for Physics-Constrained Text-to-Video Generation with Automated Preference Selection , author=. arXiv preprint arXiv:2511.03997 , year=

    work page arXiv

  77. [77]

    MMPhysVideo: Scaling Physical Plausibility in Video Generation via Joint Multimodal Modeling

    MMPhysVideo: Scaling Physical Plausibility in Video Generation via Joint Multimodal Modeling , author=. arXiv preprint arXiv:2604.02817 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  78. [78]

    2024 , url =

    Video Generation Models as World Simulators , author =. 2024 , url =

    work page 2024

  79. [79]

    2024 , eprint =

    Diffusion Forcing: Next-token Prediction Meets Full-Sequence Diffusion , author =. 2024 , eprint =

    work page 2024

  80. [80]

    2024 , eprint =

    VideoPoet: A Large Language Model for Zero-Shot Video Generation , author =. 2024 , eprint =

    work page 2024

Showing first 80 references.